chore: upgrade unregistry 0.3.1 -> 0.4.1 with a fix for pushing big images

* feat: Add `uc service start/stop` commands

* Add nolint:dupl for service start/stop commands

Co-authored-by: Pasha Sviderski <me@psviderski.name>

* Add --signal and --timeout options to uc service stop

* Check whether service stop --timeout flag was used to set options properly

---------

Co-authored-by: Pasha Sviderski <me@psviderski.name>

.goreleaser.yaml

@@ -16,7 +16,8 @@ builds:
     main: ./cmd/uncloud
     binary: uncloud
     env:
-      - CGO_ENABLED=0
+      # Use CGO to be able to build fsevent required by compose on macOS.
+      - CGO_ENABLED={{ if eq .Os "darwin" }}1{{ else }}0{{ end }}
     goos:
       - darwin
       - linux

AI.md

@@ -28,7 +28,7 @@ This document provides comprehensive information about the Uncloud project for A
 - **Docker Compose compatibility**: Uses familiar Docker Compose format
 - **Zero-downtime deployments**: Rolling updates without service interruption
 - **Automatic HTTPS**: Caddy reverse proxy with Let's Encrypt integration
-- **Managed DNS**: Free `*.cluster.uncloud.run` subdomains via Uncloud DNS service
+- **Managed DNS**: Free `*.xxxxxx.uncld.dev` subdomains via Uncloud DNS service
 - **Cross-machine scaling**: Run containers across multiple machines
 
 ### 🔧 Developer Experience

README.md

@@ -1,7 +1,7 @@
 <div align="center">
-  <img src="./website/landing/images/logo.svg" height="100" width="100" alt="Uncloud logo"/>
-  <h1>Uncloud</h1>
-  <p><strong>Docker simplicity. Multi-machine power.</strong></p>
+  <img src="./website/landing/images/logo-title.svg#gh-light-mode-only" alt="Uncloud logo"/>
+  <img src="./website/landing/images/logo-title-dark.svg#gh-dark-mode-only" alt="Uncloud logo"/>
+  <p><strong>▸ Docker simplicity. Multi-machine power ◂</strong></p>
 
   <p>
     <a href="https://uncloud.run/docs"><img src="https://img.shields.io/badge/Docs-blue.svg?style=for-the-badge&logo=gitbook&logoColor=white" alt="Documentation"></a>
@@ -40,7 +40,7 @@ complexity of Kubernetes.
 * **No control plane**: Fully decentralised design eliminates single points of failure and reduces operational overhead.
 * **Imperative over declarative**: Favoring imperative operations over state reconciliation simplifies both the mental
   model and troubleshooting.
-* **Managed DNS**: Automatic DNS records `*.<id>.cluster.uncloud.run` for services with public access via managed
+* **Managed DNS**: Automatic DNS records `*.xxxxxx.uncld.dev` for services with public access via managed
   [Uncloud DNS](https://github.com/psviderski/uncloud-dns) service.
 * **Automatic HTTPS**: Built-in Caddy reverse proxy handles TLS certificate provisioning and renewal using Let's
   Encrypt.
@@ -337,6 +337,10 @@ These companies and projects are helping Uncloud with their generous sponsorship
 
 ## ❤️ Contributors
 
+<a href="https://trendshift.io/repositories/14069">
+    <img alt="Trendshift" src="https://trendshift.io/api/badge/repositories/14069" />
+</a>
+
 Thank you [@cedws](https://github.com/cedws) for being the first contributor to Uncloud! 🎉
 
 <a href="https://github.com/psviderski/uncloud/graphs/contributors">

cmd/uncloud/build.go

@@ -49,6 +49,7 @@ to cluster machines or --push-registry to upload them to external registries.`,
 
 			return runBuild(cmd.Context(), uncli, opts)
 		},
+		GroupID: "service",
 	}
 
 	cmd.Flags().StringArrayVar(&opts.BuildArgs, "build-arg", nil,
@@ -74,8 +75,6 @@ to cluster machines or --push-registry to upload them to external registries.`,
 			"Use --machine to specify which machines. (default is all machines)")
 	cmd.Flags().BoolVar(&opts.PushRegistry, "push-registry", false,
 		"Upload the built images to external registries (e.g., Docker Hub) after building.")
-	cmd.Flags().StringVarP(&opts.Context, "context", "c", "",
-		"Name of the cluster context. (default is the current context)")
 
 	return cmd
 }

cmd/uncloud/caddy/config.go

@@ -7,14 +7,12 @@ import (
 
 	"github.com/alecthomas/chroma/v2/quick"
 	"github.com/psviderski/uncloud/internal/cli"
-	"github.com/psviderski/uncloud/pkg/api"
 	"github.com/spf13/cobra"
 )
 
 type configOptions struct {
 	machine string
 	noColor bool
-	context string
 }
 
 func NewConfigCommand() *cobra.Command {
@@ -34,16 +32,12 @@ func NewConfigCommand() *cobra.Command {
 		"Name or ID of the machine to get the configuration from. (default is connected machine)")
 	cmd.Flags().BoolVar(&opts.noColor, "no-color", false,
 		"Disable syntax highlighting for the output.")
-	cmd.Flags().StringVarP(
-		&opts.context, "context", "c", "",
-		"Name of the cluster context. (default is the current context)",
-	)
 
 	return cmd
 }
 
 func runConfig(ctx context.Context, uncli *cli.CLI, opts configOptions) error {
-	clusterClient, err := uncli.ConnectCluster(ctx, opts.context)
+	clusterClient, err := uncli.ConnectCluster(ctx)
 	if err != nil {
 		return fmt.Errorf("connect to cluster: %w", err)
 	}
@@ -51,7 +45,7 @@ func runConfig(ctx context.Context, uncli *cli.CLI, opts configOptions) error {
 
 	if opts.machine != "" {
 		// If a specific machine is requested, use it to get the Caddy configuration.
-		ctx, _, err = api.ProxyMachinesContext(ctx, clusterClient, []string{opts.machine})
+		ctx, _, err = clusterClient.ProxyMachinesContext(ctx, []string{opts.machine})
 		if err != nil {
 			return err
 		}

cmd/uncloud/caddy/deploy.go

@@ -22,7 +22,6 @@ type deployOptions struct {
 	caddyfile string
 	image     string
 	machines  []string
-	context   string
 }
 
 func NewDeployCommand() *cobra.Command {
@@ -46,10 +45,6 @@ func NewDeployCommand() *cobra.Command {
 	cmd.Flags().StringSliceVarP(&opts.machines, "machine", "m", nil,
 		"Machine names or IDs to deploy to. Can be specified multiple times or as a comma-separated "+
 			"list. (default is all machines)")
-	cmd.Flags().StringVarP(
-		&opts.context, "context", "c", "",
-		"Name of the cluster context to deploy to. (default is the current context)",
-	)
 
 	return cmd
 }
@@ -64,7 +59,7 @@ func runDeploy(ctx context.Context, uncli *cli.CLI, opts deployOptions) error {
 		caddyfile = strings.TrimSpace(string(data))
 	}
 
-	clusterClient, err := uncli.ConnectCluster(ctx, opts.context)
+	clusterClient, err := uncli.ConnectCluster(ctx)
 	if err != nil {
 		return fmt.Errorf("connect to cluster: %w", err)
 	}

cmd/uncloud/context/connection.go

@@ -0,0 +1,78 @@
+package context
+
+import (
+	"fmt"
+
+	"github.com/charmbracelet/huh"
+	"github.com/psviderski/uncloud/internal/cli"
+	"github.com/psviderski/uncloud/internal/cli/config"
+	"github.com/spf13/cobra"
+)
+
+func NewConnectionCommand() *cobra.Command {
+	cmd := &cobra.Command{
+		Use:   "connection",
+		Short: "Choose a new default connection for the current context.",
+		RunE: func(cmd *cobra.Command, args []string) error {
+			uncli := cmd.Context().Value("cli").(*cli.CLI)
+			return selectConnection(uncli)
+		},
+	}
+	return cmd
+}
+
+func selectConnection(uncli *cli.CLI) error {
+	if uncli.Config == nil {
+		return fmt.Errorf("connection management is not available: Uncloud configuration file is not being used")
+	}
+	if len(uncli.Config.Contexts) == 0 {
+		return fmt.Errorf("no contexts found in Uncloud config (%s)", uncli.Config.Path())
+	}
+
+	currentCtxName := uncli.Config.CurrentContext
+	currentCtx, ok := uncli.Config.Contexts[currentCtxName]
+	if !ok {
+		return fmt.Errorf("current context '%s' not found", currentCtxName)
+	}
+
+	if len(currentCtx.Connections) == 0 {
+		return fmt.Errorf("no connections found in context '%s'", currentCtxName)
+	}
+
+	var selectedConnection int
+	form := huh.NewForm(
+		huh.NewGroup(
+			huh.NewSelect[int]().
+				Title("Select a default connection").
+				Options(buildConnectionOptions(currentCtx.Connections)...).
+				Value(&selectedConnection),
+		),
+	)
+	if err := form.Run(); err != nil {
+		return fmt.Errorf("select connection: %w", err)
+	}
+
+	currentCtx.SetDefaultConnection(selectedConnection)
+	selected := currentCtx.Connections[0]
+
+	if err := uncli.Config.Save(); err != nil {
+		return fmt.Errorf("save config: %w", err)
+	}
+
+	fmt.Printf("Default connection for context '%s' is now '%s'.\n", currentCtxName, selected)
+	return nil
+}
+
+func buildConnectionOptions(connections []config.MachineConnection) []huh.Option[int] {
+	options := make([]huh.Option[int], len(connections))
+	for i, conn := range connections {
+		key := conn.String()
+		opt := huh.NewOption(key, i)
+		if i == 0 {
+			opt.Key += " (default)"
+			opt = opt.Selected(true)
+		}
+		options[i] = opt
+	}
+	return options
+}

cmd/uncloud/context/root.go

@@ -19,6 +19,7 @@ func NewRootCommand() *cobra.Command {
 	cmd.AddCommand(
 		NewListCommand(),
 		NewUseCommand(),
+		NewConnectionCommand(),
 	)
 
 	return cmd

cmd/uncloud/deploy.go

@@ -26,8 +26,6 @@ type deployOptions struct {
 	noBuild  bool
 	recreate bool
 	yes      bool
-
-	context string
 }
 
 // NewDeployCommand creates a new command to deploy services from a Compose file.
@@ -44,6 +42,7 @@ func NewDeployCommand() *cobra.Command {
 
 			return runDeploy(cmd.Context(), uncli, opts)
 		},
+		GroupID: "service",
 	}
 
 	cmd.Flags().StringArrayVar(&opts.BuildServicesOptions.BuildArgs, "build-arg", nil,
@@ -51,8 +50,6 @@ func NewDeployCommand() *cobra.Command {
 			"Can be specified multiple times. Format: --build-arg VAR=VALUE")
 	cmd.Flags().BoolVar(&opts.BuildServicesOptions.Pull, "build-pull", false,
 		"Always attempt to pull newer versions of base images before building service images.")
-	cmd.Flags().StringVarP(&opts.context, "context", "c", "",
-		"Name of the cluster context to deploy to (default is the current context)")
 	cmd.Flags().StringSliceVarP(&opts.files, "file", "f", nil,
 		"One or more Compose files to deploy services from. (default compose.yaml)")
 	cmd.Flags().BoolVar(&opts.noBuild, "no-build", false,
@@ -108,7 +105,7 @@ func runDeploy(ctx context.Context, uncli *cli.CLI, opts deployOptions) error {
 		fmt.Println()
 	}
 
-	clusterClient, err := uncli.ConnectCluster(ctx, opts.context)
+	clusterClient, err := uncli.ConnectCluster(ctx)
 	if err != nil {
 		return fmt.Errorf("connect to cluster: %w", err)
 	}

cmd/uncloud/dns/release.go

@@ -12,32 +12,21 @@ import (
 	"google.golang.org/protobuf/types/known/emptypb"
 )
 
-type releaseOptions struct {
-	context string
-}
-
 func NewReleaseCommand() *cobra.Command {
-	opts := releaseOptions{}
-
 	cmd := &cobra.Command{
 		Use:   "release",
 		Short: "Release the reserved cluster domain.",
 		RunE: func(cmd *cobra.Command, args []string) error {
 			uncli := cmd.Context().Value("cli").(*cli.CLI)
-			return release(cmd.Context(), uncli, opts)
+			return release(cmd.Context(), uncli)
 		},
 	}
 
-	cmd.Flags().StringVarP(
-		&opts.context, "context", "c", "",
-		"Name of the cluster context. (default is the current context)",
-	)
-
 	return cmd
 }
 
-func release(ctx context.Context, uncli *cli.CLI, opts releaseOptions) error {
-	client, err := uncli.ConnectCluster(ctx, opts.context)
+func release(ctx context.Context, uncli *cli.CLI) error {
+	client, err := uncli.ConnectCluster(ctx)
 	if err != nil {
 		return fmt.Errorf("connect to cluster: %w", err)
 	}

cmd/uncloud/dns/reserve.go

@@ -19,7 +19,6 @@ const DefaultUncloudDNSAPIEndpoint = "https://dns.uncloud.run/v1"
 
 type reserveOptions struct {
 	endpoint string
-	context  string
 }
 
 func NewReserveCommand() *cobra.Command {
@@ -36,16 +35,12 @@ func NewReserveCommand() *cobra.Command {
 
 	cmd.Flags().StringVar(&opts.endpoint, "endpoint", DefaultUncloudDNSAPIEndpoint,
 		"API endpoint for the Uncloud DNS service.")
-	cmd.Flags().StringVarP(
-		&opts.context, "context", "c", "",
-		"Name of the cluster context. (default is the current context)",
-	)
 
 	return cmd
 }
 
 func reserve(ctx context.Context, uncli *cli.CLI, opts reserveOptions) error {
-	clusterClient, err := uncli.ConnectCluster(ctx, opts.context)
+	clusterClient, err := uncli.ConnectCluster(ctx)
 	if err != nil {
 		return fmt.Errorf("connect to cluster: %w", err)
 	}

cmd/uncloud/dns/root.go

@@ -9,7 +9,7 @@ func NewRootCommand() *cobra.Command {
 		Use:   "dns",
 		Short: "Manage cluster domain in Uncloud DNS.",
 		Long: "Manage cluster domain in Uncloud DNS.\n" +
-			"DNS commands allow you to reserve or release a unique '<id>.cluster.uncloud.run' domain for your " +
+			"DNS commands allow you to reserve or release a unique 'xxxxxx.uncld.dev' domain for your " +
 			"cluster. When reserved, Caddy service deployments will automatically update DNS records to route " +
 			"traffic to the services in the cluster.",
 	}

cmd/uncloud/dns/show.go

@@ -10,32 +10,21 @@ import (
 	"github.com/spf13/cobra"
 )
 
-type showOptions struct {
-	context string
-}
-
 func NewShowCommand() *cobra.Command {
-	opts := showOptions{}
-
 	cmd := &cobra.Command{
 		Use:   "show",
 		Short: "Print the cluster domain name.",
 		RunE: func(cmd *cobra.Command, args []string) error {
 			uncli := cmd.Context().Value("cli").(*cli.CLI)
-			return show(cmd.Context(), uncli, opts)
+			return show(cmd.Context(), uncli)
 		},
 	}
 
-	cmd.Flags().StringVarP(
-		&opts.context, "context", "c", "",
-		"Name of the cluster context. (default is the current context)",
-	)
-
 	return cmd
 }
 
-func show(ctx context.Context, uncli *cli.CLI, opts showOptions) error {
-	clusterClient, err := uncli.ConnectCluster(ctx, opts.context)
+func show(ctx context.Context, uncli *cli.CLI) error {
+	clusterClient, err := uncli.ConnectCluster(ctx)
 	if err != nil {
 		return fmt.Errorf("connect to cluster: %w", err)
 	}

cmd/uncloud/image/ls.go

@@ -22,7 +22,6 @@ import (
 type listOptions struct {
 	machines   []string
 	nameFilter string
-	context    string
 }
 
 func NewListCommand() *cobra.Command {
@@ -60,10 +59,6 @@ func NewListCommand() *cobra.Command {
 	cmd.Flags().StringSliceVarP(&opts.machines, "machine", "m", nil,
 		"Filter images by machine name or ID. Can be specified multiple times or as a comma-separated list. "+
 			"(default is include all machines)")
-	cmd.Flags().StringVarP(
-		&opts.context, "context", "c", "",
-		"Name of the cluster context. (default is the current context)",
-	)
 
 	return cmd
 }
@@ -82,7 +77,7 @@ type imageRow struct {
 }
 
 func list(ctx context.Context, uncli *cli.CLI, opts listOptions) error {
-	clusterClient, err := uncli.ConnectCluster(ctx, opts.context)
+	clusterClient, err := uncli.ConnectCluster(ctx)
 	if err != nil {
 		return fmt.Errorf("connect to cluster: %w", err)
 	}

cmd/uncloud/image/push.go

@@ -15,7 +15,6 @@ type pushOptions struct {
 	image    string
 	machines []string
 	platform string
-	context  string
 }
 
 func NewPushCommand() *cobra.Command {
@@ -54,16 +53,12 @@ The image is uploaded to all cluster machines (default) or the specified machine
 		"Push a specific platform of a multi-platform image (e.g., linux/amd64, linux/arm64).\n"+
 			"Local Docker must be configured to use containerd image store to support multi-platform images.",
 	)
-	cmd.Flags().StringVarP(
-		&opts.context, "context", "c", "",
-		"Name of the cluster context. (default is the current context)",
-	)
 
 	return cmd
 }
 
 func push(ctx context.Context, uncli *cli.CLI, opts pushOptions) error {
-	clusterClient, err := uncli.ConnectCluster(ctx, opts.context)
+	clusterClient, err := uncli.ConnectCluster(ctx)
 	if err != nil {
 		return fmt.Errorf("connect to cluster: %w", err)
 	}

cmd/uncloud/machine/add.go

@@ -5,6 +5,7 @@ import (
 	"errors"
 	"fmt"
 	"net/netip"
+	"strings"
 	"time"
 
 	"github.com/cenkalti/backoff/v4"
@@ -25,7 +26,6 @@ type addOptions struct {
 	noInstall bool
 	publicIP  string
 	sshKey    string
-	context   string
 	version   string
 }
 
@@ -34,19 +34,31 @@ func NewAddCommand() *cobra.Command {
 	cmd := &cobra.Command{
 		Use:   "add [USER@]HOST[:PORT]",
 		Short: "Add a remote machine to a cluster.",
-		Args:  cobra.ExactArgs(1),
+		Long: `Add a new machine to an existing Uncloud cluster.
+
+Connection methods:
+  ssh://user@host       - Use built-in SSH library (default, no prefix required)
+  ssh+cli://user@host   - Use system SSH command (supports ProxyJump, SSH config)`,
+		Args: cobra.ExactArgs(1),
 		RunE: func(cmd *cobra.Command, args []string) error {
 			uncli := cmd.Context().Value("cli").(*cli.CLI)
 
-			user, host, port, err := config.SSHDestination(args[0]).Parse()
+			// Determine if SSH CLI needs to be used and strip scheme
+			destination := args[0]
+			useSSHCLI := strings.HasPrefix(destination, "ssh+cli://")
+			destination = strings.TrimPrefix(destination, "ssh+cli://")
+			destination = strings.TrimPrefix(destination, "ssh://")
+
+			user, host, port, err := config.SSHDestination(destination).Parse()
 			if err != nil {
 				return fmt.Errorf("parse remote machine: %w", err)
 			}
 			remoteMachine := &cli.RemoteMachine{
-				User:    user,
-				Host:    host,
-				Port:    port,
-				KeyPath: opts.sshKey,
+				User:      user,
+				Host:      host,
+				Port:      port,
+				KeyPath:   opts.sshKey,
+				UseSSHCLI: useSSHCLI,
 			}
 
 			return add(cmd.Context(), uncli, remoteMachine, opts)
@@ -76,10 +88,6 @@ func NewAddCommand() *cobra.Command {
 		&opts.version, "version", "latest",
 		"Version of the Uncloud daemon to install on the machine.",
 	)
-	cmd.Flags().StringVarP(
-		&opts.context, "context", "c", "",
-		"Name of the cluster context to add the machine to. (default is the current context)",
-	)
 
 	return cmd
 }
@@ -100,7 +108,6 @@ func add(ctx context.Context, uncli *cli.CLI, remoteMachine *cli.RemoteMachine,
 	}
 
 	clusterClient, machineClient, err := uncli.AddMachine(ctx, cli.AddMachineOptions{
-		Context:       opts.context,
 		MachineName:   opts.name,
 		PublicIP:      publicIP,
 		RemoteMachine: remoteMachine,

cmd/uncloud/machine/init.go

@@ -4,6 +4,7 @@ import (
 	"context"
 	"fmt"
 	"net/netip"
+	"strings"
 
 	"github.com/docker/compose/v2/pkg/progress"
 	"github.com/psviderski/uncloud/cmd/uncloud/caddy"
@@ -32,10 +33,14 @@ type initOptions struct {
 func NewInitCommand() *cobra.Command {
 	opts := initOptions{}
 	cmd := &cobra.Command{
-		Use:   "init [USER@HOST:PORT]",
+		Use:   "init [schema://]USER@HOST[:PORT]",
 		Short: "Initialise a new cluster with a remote machine as the first member.",
-		Long: "Initialise a new cluster by setting up a remote machine as the first member.\n" +
-			"This command creates a new context in your Uncloud config to manage the cluster.",
+		Long: `Initialise a new cluster by setting up a remote machine as the first member.
+This command creates a new context in your Uncloud config to manage the cluster.
+
+Connection methods:
+  ssh://user@host       - Use built-in SSH library (default, no prefix required)
+  ssh+cli://user@host   - Use system SSH command (supports ProxyJump, SSH config)`,
 		Example: `  # Initialise a new cluster with default settings.
   uc machine init root@<your-server-ip>
 
@@ -45,7 +50,7 @@ func NewInitCommand() *cobra.Command {
   # Initialise with a non-root user and custom SSH port and key.
   uc machine init ubuntu@<your-server-ip>:2222 -i ~/.ssh/mykey
 
-  # Initialise without Caddy (no reverse proxy) and without an automatically managed domain name (xxxxxx.cluster.uncloud.run).
+  # Initialise without Caddy (no reverse proxy) and without an automatically managed domain name (xxxxxx.uncld.dev).
   # You can deploy Caddy with 'uc caddy deploy' and reserve a domain with 'uc dns reserve' later.
   uc machine init root@<your-server-ip> --no-caddy --no-dns`,
 		// TODO: support initialising a cluster on the local machine.
@@ -55,15 +60,22 @@ func NewInitCommand() *cobra.Command {
 
 			var remoteMachine *cli.RemoteMachine
 			if len(args) > 0 {
-				user, host, port, err := config.SSHDestination(args[0]).Parse()
+				// Determine if SSH CLI is requested and strip scheme
+				destination := args[0]
+				useSSHCLI := strings.HasPrefix(destination, "ssh+cli://")
+				destination = strings.TrimPrefix(destination, "ssh+cli://")
+				destination = strings.TrimPrefix(destination, "ssh://")
+
+				user, host, port, err := config.SSHDestination(destination).Parse()
 				if err != nil {
 					return fmt.Errorf("parse remote machine: %w", err)
 				}
 				remoteMachine = &cli.RemoteMachine{
-					User:    user,
-					Host:    host,
-					Port:    port,
-					KeyPath: opts.sshKey,
+					User:      user,
+					Host:      host,
+					Port:      port,
+					KeyPath:   opts.sshKey,
+					UseSSHCLI: useSSHCLI,
 				}
 			}
 
@@ -116,6 +128,11 @@ func NewInitCommand() *cobra.Command {
 }
 
 func initCluster(ctx context.Context, uncli *cli.CLI, remoteMachine *cli.RemoteMachine, opts initOptions) error {
+	if uncli.Config == nil {
+		// Config is nil when connecting directly to a remote machine (--connect) without using Uncloud config.
+		return fmt.Errorf("do not specify --connect when initialising a new cluster")
+	}
+
 	netPrefix, err := netip.ParsePrefix(opts.network)
 	if err != nil {
 		return fmt.Errorf("parse network CIDR: %w", err)

cmd/uncloud/machine/ls.go

@@ -14,25 +14,20 @@ import (
 )
 
 func NewListCommand() *cobra.Command {
-	var contextName string
 	cmd := &cobra.Command{
 		Use:     "ls",
 		Aliases: []string{"list"},
 		Short:   "List machines in a cluster.",
 		RunE: func(cmd *cobra.Command, args []string) error {
 			uncli := cmd.Context().Value("cli").(*cli.CLI)
-			return list(cmd.Context(), uncli, contextName)
+			return list(cmd.Context(), uncli)
 		},
 	}
-	cmd.Flags().StringVarP(
-		&contextName, "context", "c", "",
-		"Name of the cluster context. (default is the current context)",
-	)
 	return cmd
 }
 
-func list(ctx context.Context, uncli *cli.CLI, clusterName string) error {
-	client, err := uncli.ConnectCluster(ctx, clusterName)
+func list(ctx context.Context, uncli *cli.CLI) error {
+	client, err := uncli.ConnectCluster(ctx)
 	if err != nil {
 		return fmt.Errorf("connect to cluster: %w", err)
 	}

cmd/uncloud/machine/rename.go

@@ -9,7 +9,6 @@ import (
 )
 
 func NewRenameCommand() *cobra.Command {
-	var contextName string
 	cmd := &cobra.Command{
 		Use:   "rename OLD_NAME NEW_NAME",
 		Short: "Rename a machine in the cluster.",
@@ -20,18 +19,14 @@ configuration including network settings, public IP, and cluster membership.`,
 		Args: cobra.ExactArgs(2),
 		RunE: func(cmd *cobra.Command, args []string) error {
 			uncli := cmd.Context().Value("cli").(*cli.CLI)
-			return rename(cmd.Context(), uncli, contextName, args[0], args[1])
+			return rename(cmd.Context(), uncli, args[0], args[1])
 		},
 	}
-	cmd.Flags().StringVarP(
-		&contextName, "context", "c", "",
-		"Name of the cluster context. (default is the current context)",
-	)
 	return cmd
 }
 
-func rename(ctx context.Context, uncli *cli.CLI, contextName, oldName, newName string) error {
-	client, err := uncli.ConnectCluster(ctx, contextName)
+func rename(ctx context.Context, uncli *cli.CLI, oldName, newName string) error {
+	client, err := uncli.ConnectCluster(ctx)
 	if err != nil {
 		return err
 	}

cmd/uncloud/machine/rm.go

@@ -22,7 +22,6 @@ import (
 type removeOptions struct {
 	noReset bool
 	yes     bool
-	context string
 }
 
 func NewRmCommand() *cobra.Command {
@@ -39,8 +38,6 @@ func NewRmCommand() *cobra.Command {
 		},
 	}
 
-	cmd.Flags().StringVarP(&opts.context, "context", "c", "",
-		"Name of the cluster context. (default is the current context)")
 	cmd.Flags().BoolVarP(&opts.yes, "yes", "y", false,
 		"Do not prompt for confirmation before removing the machine.")
 	cmd.Flags().BoolVar(&opts.noReset, "no-reset", false,
@@ -51,14 +48,14 @@ func NewRmCommand() *cobra.Command {
 
 func remove(ctx context.Context, uncli *cli.CLI, nameOrID string, opts removeOptions) error {
 	// TODO: automatically choose a connection to the machine that is not being removed.
-	client, err := uncli.ConnectCluster(ctx, opts.context)
+	client, err := uncli.ConnectCluster(ctx)
 	if err != nil {
 		return fmt.Errorf("connect to cluster: %w", err)
 	}
 	defer client.Close()
 
 	// Verify the machine exists and list all service containers on it including stopped ones.
-	mctx, machines, err := api.ProxyMachinesContext(ctx, client, []string{nameOrID})
+	mctx, machines, err := client.ProxyMachinesContext(ctx, []string{nameOrID})
 	if err != nil {
 		return err
 	}
@@ -155,8 +152,21 @@ func remove(ctx context.Context, uncli *cli.CLI, nameOrID string, opts removeOpt
 		}
 	}
 
-	// TODO: remove the connection to the machine from the uncloud config if it exists. We need a way to associate
-	//  the machine with its connection in the config, e.g. by storing the machine name in the connection metadata.
+	// Remove the connection to the machine from the uncloud config if it exists.
+	if uncli.Config != nil {
+		contextName := uncli.GetContextOverrideOrCurrent()
+		if context, ok := uncli.Config.Contexts[contextName]; ok {
+			for i, c := range context.Connections {
+				if c.MachineID == m.Id {
+					context.Connections = slices.Delete(context.Connections, i, i+1)
+					break
+				}
+			}
+			if err := uncli.Config.Save(); err != nil {
+				return fmt.Errorf("save config: %w", err)
+			}
+		}
+	}
 
 	// TODO: If Caddy was running on this machine and a cluster domain is reserved,
 	//  let the user know that the DNS records should be updated.

cmd/uncloud/machine/root.go

@@ -8,7 +8,7 @@ func NewRootCommand() *cobra.Command {
 	cmd := &cobra.Command{
 		Use:     "machine",
 		Aliases: []string{"m"},
-		Short:   "Manage machines in an Uncloud cluster.",
+		Short:   "Manage machines in the cluster.",
 	}
 	cmd.AddCommand(
 		NewAddCommand(),

cmd/uncloud/machine/update.go

@@ -13,7 +13,6 @@ import (
 type updateOptions struct {
 	name     string
 	publicIP string
-	context  string
 }
 
 func NewUpdateCommand() *cobra.Command {
@@ -43,10 +42,6 @@ At least one flag must be specified to perform an update operation.`,
 		&opts.publicIP, "public-ip", "",
 		fmt.Sprintf("Public IP address of the machine for ingress configuration. Use '%s' or '' to remove the public IP.", PublicIPNone),
 	)
-	cmd.Flags().StringVarP(
-		&opts.context, "context", "c", "",
-		"Name of the cluster context. (default is the current context)",
-	)
 
 	return cmd
 }
@@ -57,7 +52,7 @@ func update(ctx context.Context, uncli *cli.CLI, cmd *cobra.Command, opts update
 		return fmt.Errorf("at least one update flag must be specified (--name, --public-ip)")
 	}
 
-	client, err := uncli.ConnectCluster(ctx, opts.context)
+	client, err := uncli.ConnectCluster(ctx)
 	if err != nil {
 		return err
 	}

cmd/uncloud/main.go

@@ -6,6 +6,7 @@ import (
 	"net/netip"
 	"strings"
 
+	"github.com/charmbracelet/lipgloss"
 	"github.com/psviderski/uncloud/cmd/uncloud/caddy"
 	cmdcontext "github.com/psviderski/uncloud/cmd/uncloud/context"
 	"github.com/psviderski/uncloud/cmd/uncloud/dns"
@@ -24,6 +25,7 @@ import (
 type globalOptions struct {
 	configPath string
 	connect    string
+	context    string
 }
 
 func main() {
@@ -38,6 +40,7 @@ func main() {
 		SilenceErrors: true,
 		PersistentPreRunE: func(cmd *cobra.Command, args []string) error {
 			cli.BindEnvToFlag(cmd, "connect", "UNCLOUD_CONNECT")
+			cli.BindEnvToFlag(cmd, "context", "UNCLOUD_CONTEXT")
 			cli.BindEnvToFlag(cmd, "uncloud-config", "UNCLOUD_CONFIG")
 
 			var conn *config.MachineConnection
@@ -55,11 +58,12 @@ func main() {
 					conn = &config.MachineConnection{
 						SSHCLI: config.SSHDestination(dest),
 					}
-				} else {
-					dest := opts.connect
-					if strings.HasPrefix(dest, "ssh://") {
-						dest = dest[len("ssh://"):]
+				} else if strings.HasPrefix(opts.connect, "unix://") {
+					conn = &config.MachineConnection{
+						Unix: opts.connect[len("unix://"):],
 					}
+				} else {
+					dest := strings.TrimPrefix(opts.connect, "ssh://")
 					conn = &config.MachineConnection{
 						SSH: config.SSHDestination(dest),
 					}
@@ -67,7 +71,7 @@ func main() {
 			}
 
 			configPath := fs.ExpandHomeDir(opts.configPath)
-			uncli, err := cli.New(configPath, conn)
+			uncli, err := cli.New(configPath, conn, opts.context)
 			if err != nil {
 				return fmt.Errorf("initialise CLI: %w", err)
 			}
@@ -78,29 +82,57 @@ func main() {
 
 	cmd.PersistentFlags().StringVar(&opts.connect, "connect", "",
 		"Connect to a remote cluster machine without using the Uncloud configuration file. [$UNCLOUD_CONNECT]\n"+
-			"Format: [ssh://]user@host[:port], ssh+cli://user@host[:port], or tcp://host:port")
+			"Format: [ssh://]user@host[:port], ssh+cli://user@host[:port], tcp://host:port, or unix:///path/to/uncloud.sock")
 	cmd.PersistentFlags().StringVar(&opts.configPath, "uncloud-config", "~/.config/uncloud/config.yaml",
 		"Path to the Uncloud configuration file. [$UNCLOUD_CONFIG]")
 	_ = cmd.MarkPersistentFlagFilename("uncloud-config", "yaml", "yml")
-	// TODO: make --context a global flag and pass it as a value of the command context.
+	cmd.PersistentFlags().StringVarP(&opts.context, "context", "c", "",
+		"Name of the cluster context to use (default is the current context). [$UNCLOUD_CONTEXT]")
+
+	// Set custom help function to show links to docs and Discord only for the root 'uc' command.
+	defaultHelpFunc := cmd.HelpFunc()
+	cmd.SetHelpFunc(func(c *cobra.Command, args []string) {
+		defaultHelpFunc(c, args)
+		// Only show links for the root 'uc' command.
+		if c.Name() == "uc" {
+			urlStyle := lipgloss.NewStyle().
+				Underline(true).
+				Foreground(lipgloss.Color("12")) // light blue
+
+			fmt.Fprintln(c.OutOrStdout())
+			fmt.Fprintf(c.OutOrStdout(), "Learn more about Uncloud:       %s\n",
+				urlStyle.Render("https://uncloud.run/docs"))
+			fmt.Fprintf(c.OutOrStdout(), "Join our Discord community:     %s\n",
+				urlStyle.Render("https://uncloud.run/discord"))
+		}
+	})
+
+	cmd.AddGroup(&cobra.Group{
+		ID:    "service",
+		Title: "Deploy and manage services:",
+	})
 
 	cmd.AddCommand(
+		NewBuildCommand(),
 		NewDeployCommand(),
 		NewDocsCommand(),
-		NewBuildCommand(),
 		NewImagesCommand(),
+		NewPsCommand(),
 		caddy.NewRootCommand(),
 		cmdcontext.NewRootCommand(),
 		dns.NewRootCommand(),
 		image.NewRootCommand(),
 		machine.NewRootCommand(),
 		service.NewRootCommand(),
-		service.NewExecCommand(),
-		service.NewInspectCommand(),
-		service.NewListCommand(),
-		service.NewRmCommand(),
-		service.NewRunCommand(),
-		service.NewScaleCommand(),
+		service.NewExecCommand("service"),
+		service.NewInspectCommand("service"),
+		service.NewListCommand("service"),
+		service.NewLogsCommand("service"),
+		service.NewRmCommand("service"),
+		service.NewRunCommand("service"),
+		service.NewScaleCommand("service"),
+		service.NewStartCommand("service"),
+		service.NewStopCommand("service"),
 		volume.NewRootCommand(),
 	)
 	cobra.CheckErr(cmd.Execute())

cmd/uncloud/ps.go

@@ -0,0 +1,284 @@
+package main
+
+import (
+	"context"
+	"fmt"
+	"sort"
+	"time"
+
+	"github.com/charmbracelet/huh/spinner"
+	"github.com/charmbracelet/lipgloss"
+	"github.com/charmbracelet/lipgloss/table"
+	"github.com/docker/docker/api/types/container"
+	"github.com/docker/go-units"
+	"github.com/spf13/cobra"
+
+	"github.com/psviderski/uncloud/internal/cli"
+	"github.com/psviderski/uncloud/pkg/client"
+)
+
+const (
+	sortByService = "service"
+	sortByMachine = "machine"
+	sortByHealth  = "health"
+)
+
+type containerHighlight int
+
+const (
+	highlightDanger containerHighlight = iota
+	highlightWarning
+	highlightSuccess
+	highlightNormal
+)
+
+type psOptions struct {
+	sortBy string
+}
+
+func NewPsCommand() *cobra.Command {
+	opts := psOptions{}
+	cmd := &cobra.Command{
+		Use:   "ps",
+		Short: "List all service containers.",
+		Long: `List all service containers across all machines in the cluster.
+
+This command provides a comprehensive overview of all running containers that are part of a service,
+making it easy to see the distribution and status of containers across the cluster.`,
+		RunE: func(cmd *cobra.Command, args []string) error {
+			uncli := cmd.Context().Value("cli").(*cli.CLI)
+
+			if opts.sortBy != sortByService && opts.sortBy != sortByMachine && opts.sortBy != sortByHealth {
+				return fmt.Errorf("invalid value for --sort: %q, must be one of '%s', '%s' or '%s'", opts.sortBy,
+					sortByService, sortByMachine, sortByHealth)
+			}
+
+			return runPs(cmd.Context(), uncli, opts)
+		},
+		GroupID: "service",
+	}
+	cmd.Flags().StringVarP(&opts.sortBy, "sort", "s", sortByService,
+		"Sort containers by 'service', 'machine', or 'health'.")
+	return cmd
+}
+
+type containerInfo struct {
+	serviceName string
+	machineName string
+	id          string
+	name        string
+	image       string
+	status      string
+	highlight   containerHighlight
+	created     time.Time
+	ip          string
+}
+
+func runPs(ctx context.Context, uncli *cli.CLI, opts psOptions) error {
+	clusterClient, err := uncli.ConnectCluster(ctx)
+	if err != nil {
+		return fmt.Errorf("connect to cluster: %w", err)
+	}
+	defer clusterClient.Close()
+
+	var containers []containerInfo
+	err = spinner.New().
+		Title(" Collecting container info...").
+		Type(spinner.MiniDot).
+		Style(lipgloss.NewStyle().Foreground(lipgloss.Color("3"))).
+		ActionWithErr(func(ctx context.Context) error {
+			containers, err = collectContainers(ctx, clusterClient)
+			return err
+		}).
+		Run()
+	if err != nil {
+		return fmt.Errorf("collect containers: %w", err)
+	}
+
+	// Sort the containers based on the sorting option.
+	sort.SliceStable(containers, func(i, j int) bool {
+		a, b := containers[i], containers[j]
+		switch opts.sortBy {
+		case sortByHealth:
+			if a.highlight != b.highlight {
+				return a.highlight < b.highlight
+			}
+			if a.serviceName != b.serviceName {
+				return a.serviceName < b.serviceName
+			}
+		case sortByMachine:
+			if a.machineName != b.machineName {
+				return a.machineName < b.machineName
+			}
+			if a.serviceName != b.serviceName {
+				return a.serviceName < b.serviceName
+			}
+		default: // sortByService
+			if a.serviceName != b.serviceName {
+				return a.serviceName < b.serviceName
+			}
+		}
+		// Fallback to creation time (newest first).
+		return a.created.After(b.created)
+	})
+
+	return printContainers(containers)
+}
+
+func printContainers(containers []containerInfo) error {
+	t := table.New().
+		// Remove the default border.
+		Border(lipgloss.Border{}).
+		BorderTop(false).
+		BorderBottom(false).
+		BorderLeft(false).
+		BorderRight(false).
+		BorderHeader(false).
+		BorderColumn(false).
+		StyleFunc(func(row, col int) lipgloss.Style {
+			if row == table.HeaderRow {
+				return lipgloss.NewStyle().Bold(true).PaddingRight(3)
+			}
+			// Regular style for data rows with padding.
+			return lipgloss.NewStyle().PaddingRight(3)
+		})
+
+	t.Headers("SERVICE", "CONTAINER ID", "CONTAINER NAME", "IMAGE", "CREATED", "STATUS", "IP ADDRESS", "MACHINE")
+
+	for _, ctr := range containers {
+		id := ctr.id
+		if len(id) > 12 {
+			id = id[:12]
+		}
+
+		created := units.HumanDuration(time.Now().UTC().Sub(ctr.created)) + " ago"
+
+		var statusStyle lipgloss.Style
+		switch ctr.highlight {
+		case highlightSuccess:
+			statusStyle = lipgloss.NewStyle().Foreground(lipgloss.Color("2")) // Green
+		case highlightDanger:
+			statusStyle = lipgloss.NewStyle().Foreground(lipgloss.Color("1")) // Red
+		case highlightWarning:
+			statusStyle = lipgloss.NewStyle().Foreground(lipgloss.Color("3")) // Yellow
+		default:
+			statusStyle = lipgloss.NewStyle() // Default
+		}
+
+		t.Row(
+			ctr.serviceName,
+			id,
+			ctr.name,
+			ctr.image,
+			created,
+			statusStyle.Render(ctr.status),
+			ctr.ip,
+			ctr.machineName,
+		)
+	}
+
+	fmt.Println(t)
+	return nil
+}
+
+func collectContainers(ctx context.Context, cli *client.Client) ([]containerInfo, error) {
+	listCtx, machines, err := cli.ProxyMachinesContext(ctx, nil)
+	if err != nil {
+		return nil, fmt.Errorf("proxy machines context: %w", err)
+	}
+
+	// Create a map of IP to machine name for resolving response metadata
+	machinesNamesByIP := make(map[string]string)
+	for _, m := range machines {
+		if addr, err := m.Machine.Network.ManagementIp.ToAddr(); err == nil {
+			machinesNamesByIP[addr.String()] = m.Machine.Name
+		}
+	}
+
+	// List all service containers across all machines in the cluster.
+	machineContainers, err := cli.Docker.ListServiceContainers(
+		listCtx, "", container.ListOptions{All: true},
+	)
+	if err != nil {
+		return nil, fmt.Errorf("list service containers: %w", err)
+	}
+
+	var containers []containerInfo
+	for _, msc := range machineContainers {
+		machineName := "unknown"
+
+		// Metadata can be nil if the request was broadcasted to only one machine.
+		if msc.Metadata == nil && len(machineContainers) > 1 {
+			return nil, fmt.Errorf("something went wrong with gRPC proxy: metadata is missing for a machine response")
+		}
+		if msc.Metadata != nil && msc.Metadata.Error != "" {
+			client.PrintWarning(fmt.Sprintf("failed to list containers on machine %s: %s", machineName,
+				msc.Metadata.Error))
+			continue
+		}
+
+		if msc.Metadata != nil {
+			var ok bool
+			machineName, ok = machinesNamesByIP[msc.Metadata.Machine]
+			if !ok {
+				// Fallback to machine's IP as name.
+				machineName = msc.Metadata.Machine
+			}
+		} else {
+			// Fallback to the first available machine name.
+			if len(machines) > 0 {
+				machineName = machines[0].Machine.Name
+			}
+		}
+
+		for _, ctr := range msc.Containers {
+			if ctr.Container.State == nil || ctr.Container.Config == nil {
+				continue
+			}
+
+			status, err := ctr.Container.HumanState()
+			if err != nil {
+				return nil, fmt.Errorf("get human state for container %s: %w", ctr.Container.ID, err)
+			}
+
+			var highlight containerHighlight
+			healthStatus := ""
+			if ctr.Container.State.Health != nil {
+				healthStatus = ctr.Container.State.Health.Status
+			}
+
+			if healthStatus == container.Unhealthy || ctr.Container.State.Status == "dead" || ctr.Container.State.OOMKilled || ctr.Container.State.Dead {
+				highlight = highlightDanger
+			} else if healthStatus == container.Healthy {
+				highlight = highlightSuccess
+			} else if ctr.Container.State.Status == "running" {
+				highlight = highlightNormal
+			} else { // Other non-critical but noteworthy states
+				highlight = highlightWarning
+			}
+
+			created, _ := time.Parse(time.RFC3339Nano, ctr.Container.Created)
+
+			ip := ctr.Container.UncloudNetworkIP()
+			ipStr := ""
+			// The container might not have an IP if it's not running or uses the host network.
+			if ip.IsValid() {
+				ipStr = ip.String()
+			}
+
+			info := containerInfo{
+				serviceName: ctr.ServiceName(),
+				machineName: machineName,
+				id:          ctr.Container.ID,
+				name:        ctr.Container.Name,
+				image:       ctr.Container.Config.Image,
+				status:      status,
+				highlight:   highlight,
+				created:     created,
+				ip:          ipStr,
+			}
+			containers = append(containers, info)
+		}
+	}
+	return containers, nil
+}

cmd/uncloud/ps_test.go

@@ -0,0 +1,442 @@
+package main
+
+import (
+	"context"
+	"encoding/json"
+	"net/netip"
+	"testing"
+
+	"github.com/psviderski/uncloud/internal/machine/api/pb"
+	"github.com/psviderski/uncloud/internal/machine/docker"
+	"github.com/psviderski/uncloud/pkg/client"
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+	"google.golang.org/grpc"
+	"google.golang.org/protobuf/types/known/emptypb"
+)
+
+// mockDockerClient implements pb.DockerClient
+type mockDockerClient struct {
+	pb.DockerClient // Embed to avoid implementing all methods
+	listResp        *pb.ListServiceContainersResponse
+	listErr         error
+}
+
+func (m *mockDockerClient) ListServiceContainers(ctx context.Context, in *pb.ListServiceContainersRequest, opts ...grpc.CallOption) (*pb.ListServiceContainersResponse, error) {
+	return m.listResp, m.listErr
+}
+
+// mockClusterClient implements pb.ClusterClient
+type mockClusterClient struct {
+	pb.ClusterClient // Embed to avoid implementing all methods
+	machinesResp     *pb.ListMachinesResponse
+	machinesErr      error
+}
+
+func (m *mockClusterClient) ListMachines(ctx context.Context, in *emptypb.Empty, opts ...grpc.CallOption) (*pb.ListMachinesResponse, error) {
+	return m.machinesResp, m.machinesErr
+}
+
+func TestCollectContainers_NilMetadata(t *testing.T) {
+	// Setup container data
+	containerData := map[string]interface{}{
+		"Id":   "container1",
+		"Name": "test-container",
+		"Config": map[string]interface{}{
+			"Image": "test-image",
+		},
+		"State": map[string]interface{}{
+			"Status":     "running",
+			"StartedAt":  "2023-01-01T12:00:00Z",
+			"FinishedAt": "0001-01-01T00:00:00Z",
+		},
+		"NetworkSettings": map[string]interface{}{
+			"Networks": map[string]interface{}{},
+		},
+	}
+	containerJSON, err := json.Marshal(containerData)
+	require.NoError(t, err)
+
+	serviceSpecJSON, err := json.Marshal(map[string]interface{}{})
+	require.NoError(t, err)
+
+	// Setup mocks
+	mockDocker := &mockDockerClient{
+		listResp: &pb.ListServiceContainersResponse{
+			Messages: []*pb.MachineServiceContainers{
+				{
+					Metadata: nil, // Simulating the issue: nil metadata
+					Containers: []*pb.ServiceContainer{
+						{
+							Container:   containerJSON,
+							ServiceSpec: serviceSpecJSON,
+						},
+					},
+				},
+			},
+		},
+	}
+
+	machineIP := "10.0.0.1"
+	mockCluster := &mockClusterClient{
+		machinesResp: &pb.ListMachinesResponse{
+			Machines: []*pb.MachineMember{
+				{
+					Machine: &pb.MachineInfo{
+						Name: "machine-1",
+						Network: &pb.NetworkConfig{
+							ManagementIp: pb.NewIP(netip.MustParseAddr(machineIP)),
+						},
+					},
+					State: pb.MachineMember_UP,
+				},
+			},
+		},
+	}
+
+	// Construct client with mocks
+	cli := &client.Client{
+		Docker: &docker.Client{
+			GRPCClient: mockDocker,
+		},
+		ClusterClient: mockCluster,
+	}
+
+	// Execute
+	containers, err := collectContainers(context.Background(), cli)
+	require.NoError(t, err)
+
+	// Verify
+	assert.Len(t, containers, 1)
+	if len(containers) > 0 {
+		c := containers[0]
+		assert.Equal(t, "container1", c.id)
+		assert.Equal(t, "test-container", c.name)
+		assert.Equal(t, "machine-1", c.machineName, "Should fall back to the single machine name when metadata is nil")
+	}
+}
+
+func TestCollectContainers_NilMetadata_MultipleMachines_Error(t *testing.T) {
+	// If we have multiple machines but receive nil metadata, it should return an error as it is ambiguous
+
+	// Setup container data
+	containerData1 := map[string]interface{}{
+		"Id": "container1",
+	}
+	containerJSON1, _ := json.Marshal(containerData1)
+
+	containerData2 := map[string]interface{}{
+		"Id": "container2",
+		"Config": map[string]interface{}{
+			"Image": "test-image",
+		},
+		"State": map[string]interface{}{
+			"Status":     "running",
+			"StartedAt":  "2023-01-01T12:00:00Z",
+			"FinishedAt": "0001-01-01T00:00:00Z",
+		},
+		"NetworkSettings": map[string]interface{}{
+			"Networks": map[string]interface{}{},
+		},
+	}
+	containerJSON2, _ := json.Marshal(containerData2)
+
+	serviceSpecJSON, _ := json.Marshal(map[string]interface{}{})
+
+	// Setup mocks
+	mockDocker := &mockDockerClient{
+		listResp: &pb.ListServiceContainersResponse{
+			Messages: []*pb.MachineServiceContainers{
+				{
+					Metadata: nil, // Nil metadata
+					Containers: []*pb.ServiceContainer{
+						{
+							Container:   containerJSON1,
+							ServiceSpec: serviceSpecJSON,
+						},
+					},
+				},
+				{
+					Metadata: &pb.Metadata{Machine: "10.0.0.2"},
+					Containers: []*pb.ServiceContainer{
+						{
+							Container:   containerJSON2,
+							ServiceSpec: serviceSpecJSON,
+						},
+					},
+				},
+			},
+		},
+	}
+
+	mockCluster := &mockClusterClient{
+		machinesResp: &pb.ListMachinesResponse{
+			Machines: []*pb.MachineMember{
+				{
+					Machine: &pb.MachineInfo{
+						Name: "machine-1",
+						Network: &pb.NetworkConfig{
+							ManagementIp: pb.NewIP(netip.MustParseAddr("10.0.0.1")),
+						},
+					},
+				},
+				{
+					Machine: &pb.MachineInfo{
+						Name: "machine-2",
+						Network: &pb.NetworkConfig{
+							ManagementIp: pb.NewIP(netip.MustParseAddr("10.0.0.2")),
+						},
+					},
+				},
+			},
+		},
+	}
+
+	// Construct client with mocks
+	cli := &client.Client{
+		Docker: &docker.Client{
+			GRPCClient: mockDocker,
+		},
+		ClusterClient: mockCluster,
+	}
+
+	// Execute
+	_, err := collectContainers(context.Background(), cli)
+	require.Error(t, err)
+	assert.Contains(t, err.Error(), "metadata is missing for a machine response")
+}
+
+func TestCollectContainers_MetadataPresent_MultipleMachines(t *testing.T) {
+	// Verify correct mapping of containers to machines when metadata is present
+
+	// Setup container data
+	containerData1 := map[string]interface{}{
+		"Id":   "container1",
+		"Name": "container-1",
+		"Config": map[string]interface{}{
+			"Image": "image-1",
+		},
+		"State": map[string]interface{}{
+			"Status":     "running",
+			"StartedAt":  "2023-01-01T12:00:00Z",
+			"FinishedAt": "0001-01-01T00:00:00Z",
+		},
+		"NetworkSettings": map[string]interface{}{
+			"Networks": map[string]interface{}{},
+		},
+	}
+	containerJSON1, _ := json.Marshal(containerData1)
+
+	containerData2 := map[string]interface{}{
+		"Id":   "container2",
+		"Name": "container-2",
+		"Config": map[string]interface{}{
+			"Image": "image-2",
+		},
+		"State": map[string]interface{}{
+			"Status":     "running",
+			"StartedAt":  "2023-01-01T12:00:00Z",
+			"FinishedAt": "0001-01-01T00:00:00Z",
+		},
+		"NetworkSettings": map[string]interface{}{
+			"Networks": map[string]interface{}{},
+		},
+	}
+	containerJSON2, _ := json.Marshal(containerData2)
+
+	serviceSpecJSON, _ := json.Marshal(map[string]interface{}{})
+
+	// Setup mocks
+	mockDocker := &mockDockerClient{
+		listResp: &pb.ListServiceContainersResponse{
+			Messages: []*pb.MachineServiceContainers{
+				{
+					Metadata: &pb.Metadata{Machine: "10.0.0.1"},
+					Containers: []*pb.ServiceContainer{
+						{
+							Container:   containerJSON1,
+							ServiceSpec: serviceSpecJSON,
+						},
+					},
+				},
+				{
+					Metadata: &pb.Metadata{Machine: "10.0.0.2"},
+					Containers: []*pb.ServiceContainer{
+						{
+							Container:   containerJSON2,
+							ServiceSpec: serviceSpecJSON,
+						},
+					},
+				},
+			},
+		},
+	}
+
+	mockCluster := &mockClusterClient{
+		machinesResp: &pb.ListMachinesResponse{
+			Machines: []*pb.MachineMember{
+				{
+					Machine: &pb.MachineInfo{
+						Name: "machine-1",
+						Network: &pb.NetworkConfig{
+							ManagementIp: pb.NewIP(netip.MustParseAddr("10.0.0.1")),
+						},
+					},
+				},
+				{
+					Machine: &pb.MachineInfo{
+						Name: "machine-2",
+						Network: &pb.NetworkConfig{
+							ManagementIp: pb.NewIP(netip.MustParseAddr("10.0.0.2")),
+						},
+					},
+				},
+			},
+		},
+	}
+
+	cli := &client.Client{
+		Docker:        &docker.Client{GRPCClient: mockDocker},
+		ClusterClient: mockCluster,
+	}
+
+	containers, err := collectContainers(context.Background(), cli)
+	require.NoError(t, err)
+
+	assert.Len(t, containers, 2)
+	// Order is not guaranteed by the map iteration in logic or parallel fetch (though here it's mocked sequential),
+	// but collectContainers just appends.
+	// We'll find them by ID.
+	for _, c := range containers {
+		if c.id == "container1" {
+			assert.Equal(t, "machine-1", c.machineName)
+		} else if c.id == "container2" {
+			assert.Equal(t, "machine-2", c.machineName)
+		} else {
+			t.Errorf("unexpected container id: %s", c.id)
+		}
+	}
+}
+
+func TestCollectContainers_NilMetadata_NoMachines(t *testing.T) {
+	// Case: 1 msc with nil metadata but no machines at all
+
+	containerData := map[string]interface{}{
+		"Id": "container1",
+		"Config": map[string]interface{}{
+			"Image": "test-image",
+		},
+		"State": map[string]interface{}{
+			"Status":     "running",
+			"StartedAt":  "2023-01-01T12:00:00Z",
+			"FinishedAt": "0001-01-01T00:00:00Z",
+		},
+		"NetworkSettings": map[string]interface{}{
+			"Networks": map[string]interface{}{},
+		},
+	}
+	containerJSON, _ := json.Marshal(containerData)
+	serviceSpecJSON, _ := json.Marshal(map[string]interface{}{})
+
+	mockDocker := &mockDockerClient{
+		listResp: &pb.ListServiceContainersResponse{
+			Messages: []*pb.MachineServiceContainers{
+				{
+					Metadata: nil,
+					Containers: []*pb.ServiceContainer{
+						{
+							Container:   containerJSON,
+							ServiceSpec: serviceSpecJSON,
+						},
+					},
+				},
+			},
+		},
+	}
+
+	// No machines in cluster response
+	mockCluster := &mockClusterClient{
+		machinesResp: &pb.ListMachinesResponse{
+			Machines: []*pb.MachineMember{},
+		},
+	}
+
+	cli := &client.Client{
+		Docker:        &docker.Client{GRPCClient: mockDocker},
+		ClusterClient: mockCluster,
+	}
+
+	containers, err := collectContainers(context.Background(), cli)
+	require.NoError(t, err)
+
+	assert.Len(t, containers, 1)
+	if len(containers) > 0 {
+		assert.Equal(t, "unknown", containers[0].machineName)
+	}
+}
+
+func TestCollectContainers_MetadataPresent_NotInMapping(t *testing.T) {
+	// Case: msc with metadata that is not in the IP-to-name mapping
+
+	containerData := map[string]interface{}{
+		"Id": "container1",
+		"Config": map[string]interface{}{
+			"Image": "test-image",
+		},
+		"State": map[string]interface{}{
+			"Status":     "running",
+			"StartedAt":  "2023-01-01T12:00:00Z",
+			"FinishedAt": "0001-01-01T00:00:00Z",
+		},
+		"NetworkSettings": map[string]interface{}{
+			"Networks": map[string]interface{}{},
+		},
+	}
+	containerJSON, _ := json.Marshal(containerData)
+	serviceSpecJSON, _ := json.Marshal(map[string]interface{}{})
+
+	mockDocker := &mockDockerClient{
+		listResp: &pb.ListServiceContainersResponse{
+			Messages: []*pb.MachineServiceContainers{
+				{
+					Metadata: &pb.Metadata{Machine: "10.0.0.99"}, // Unknown IP
+					Containers: []*pb.ServiceContainer{
+						{
+							Container:   containerJSON,
+							ServiceSpec: serviceSpecJSON,
+						},
+					},
+				},
+			},
+		},
+	}
+
+	mockCluster := &mockClusterClient{
+		machinesResp: &pb.ListMachinesResponse{
+			Machines: []*pb.MachineMember{
+				{
+					Machine: &pb.MachineInfo{
+						Name: "machine-1",
+						Network: &pb.NetworkConfig{
+							ManagementIp: pb.NewIP(netip.MustParseAddr("10.0.0.1")),
+						},
+					},
+				},
+			},
+		},
+	}
+
+	cli := &client.Client{
+		Docker:        &docker.Client{GRPCClient: mockDocker},
+		ClusterClient: mockCluster,
+	}
+
+	containers, err := collectContainers(context.Background(), cli)
+	require.NoError(t, err)
+
+	assert.Len(t, containers, 1)
+	if len(containers) > 0 {
+		// Should fallback to the IP/string in metadata
+		assert.Equal(t, "10.0.0.99", containers[0].machineName)
+	}
+}

cmd/uncloud/service/exec.go

@@ -15,18 +15,17 @@ type execCliOptions struct {
 	detach      bool
 	interactive bool
 	noTty       bool
-	context     string
 	containerId string
 }
 
 var DEFAULT_COMMAND = []string{"sh", "-c", "command -v bash >/dev/null 2>&1 && exec bash || exec sh"}
 
-func NewExecCommand() *cobra.Command {
+func NewExecCommand(groupID string) *cobra.Command {
 	opts := execCliOptions{}
 
 	execCmd := &cobra.Command{
 		Use:   "exec [OPTIONS] SERVICE [COMMAND ARGS...]",
-		Short: "Execute a command in a running service container",
+		Short: "Execute a command in a running service container.",
 		Long: `Execute a command (interactive shell by default) in a running container within a service.
 If the service has multiple replicas and no container ID is specified, the command will be executed in a random container.
 	`,
@@ -55,6 +54,7 @@ If the service has multiple replicas and no container ID is specified, the comma
 			}
 			return runExec(cmd.Context(), uncli, serviceName, command, opts)
 		},
+		GroupID: groupID,
 	}
 
 	execCmd.Flags().BoolVarP(&opts.detach, "detach", "d", false, "Detached mode: run command in the background")
@@ -69,9 +69,6 @@ If the service has multiple replicas and no container ID is specified, the comma
 	execCmd.Flags().BoolP("tty", "t", false, "Allocate a pseudo-TTY")
 	execCmd.Flags().MarkHidden("tty")
 
-	execCmd.Flags().StringVarP(&opts.context, "context", "c", "",
-		"Name of the cluster context. (default is the current context)")
-
 	// Common flags
 	execCmd.Flags().StringVar(&opts.containerId, "container", "",
 		"ID of the container to exec into. Accepts full ID or a unique prefix "+
@@ -95,7 +92,7 @@ func runExec(ctx context.Context, uncli *cli.CLI, serviceName string, command []
 		}
 	}
 
-	client, err := uncli.ConnectClusterWithOptions(ctx, opts.context, cli.ConnectOptions{
+	client, err := uncli.ConnectClusterWithOptions(ctx, cli.ConnectOptions{
 		ShowProgress: false,
 	})
 	if err != nil {

cmd/uncloud/service/inspect.go

@@ -4,21 +4,22 @@ import (
 	"context"
 	"fmt"
 	"os"
+	"slices"
 	"text/tabwriter"
 	"time"
 
 	"github.com/docker/docker/pkg/stringid"
 	"github.com/docker/go-units"
 	"github.com/psviderski/uncloud/internal/cli"
+	"github.com/psviderski/uncloud/pkg/api"
 	"github.com/spf13/cobra"
 )
 
 type inspectOptions struct {
 	service string
-	context string
 }
 
-func NewInspectCommand() *cobra.Command {
+func NewInspectCommand(groupID string) *cobra.Command {
 	opts := inspectOptions{}
 	cmd := &cobra.Command{
 		Use:   "inspect SERVICE",
@@ -29,16 +30,13 @@ func NewInspectCommand() *cobra.Command {
 			opts.service = args[0]
 			return inspect(cmd.Context(), uncli, opts)
 		},
+		GroupID: groupID,
 	}
-	cmd.Flags().StringVarP(
-		&opts.context, "context", "c", "",
-		"Name of the cluster context. (default is the current context)",
-	)
 	return cmd
 }
 
 func inspect(ctx context.Context, uncli *cli.CLI, opts inspectOptions) error {
-	client, err := uncli.ConnectCluster(ctx, opts.context)
+	client, err := uncli.ConnectCluster(ctx)
 	if err != nil {
 		return fmt.Errorf("connect to cluster: %w", err)
 	}
@@ -63,18 +61,26 @@ func inspect(ctx context.Context, uncli *cli.CLI, opts inspectOptions) error {
 	fmt.Printf("Mode:       %s\n", svc.Mode)
 	fmt.Println()
 
+	// Parse created times for sorting and display.
+	createdTimes := make(map[string]time.Time, len(svc.Containers))
+	for _, ctr := range svc.Containers {
+		createdTimes[ctr.Container.ID], _ = time.Parse(time.RFC3339Nano, ctr.Container.Created)
+	}
+
+	// Sort containers by created time (newest first).
+	slices.SortFunc(svc.Containers, func(a, b api.MachineServiceContainer) int {
+		return createdTimes[b.Container.ID].Compare(createdTimes[a.Container.ID])
+	})
+
 	// Print the list of containers in a table format.
 	tw := tabwriter.NewWriter(os.Stdout, 0, 0, 3, ' ', 0)
-	if _, err = fmt.Fprintln(tw, "CONTAINER ID\tIMAGE\tCREATED\tSTATUS\tMACHINE"); err != nil {
+	if _, err = fmt.Fprintln(tw, "CONTAINER ID\tIMAGE\tCREATED\tSTATUS\tIP ADDRESS\tMACHINE"); err != nil {
 		return fmt.Errorf("write header: %w", err)
 	}
 
+	now := time.Now().UTC()
 	for _, ctr := range svc.Containers {
-		createdAt, err := time.Parse(time.RFC3339Nano, ctr.Container.Created)
-		if err != nil {
-			return fmt.Errorf("parse created time: %w", err)
-		}
-		created := units.HumanDuration(time.Now().UTC().Sub(createdAt)) + " ago"
+		created := units.HumanDuration(now.Sub(createdTimes[ctr.Container.ID])) + " ago"
 
 		machine := machinesNamesByID[ctr.MachineID]
 		if machine == "" {
@@ -85,13 +91,21 @@ func inspect(ctx context.Context, uncli *cli.CLI, opts inspectOptions) error {
 			return fmt.Errorf("get human state: %w", err)
 		}
 
+		ip := ctr.Container.UncloudNetworkIP()
+		ipStr := ""
+		// The container might not have an IP if it's not running or uses the host network.
+		if ip.IsValid() {
+			ipStr = ip.String()
+		}
+
 		_, err = fmt.Fprintf(
 			tw,
-			"%s\t%s\t%s\t%s\t%s\n",
+			"%s\t%s\t%s\t%s\t%s\t%s\n",
 			stringid.TruncateID(ctr.Container.ID),
 			ctr.Container.Config.Image,
 			created,
 			state,
+			ipStr,
 			machine,
 		)
 		if err != nil {

cmd/uncloud/service/logs.go

@@ -0,0 +1,336 @@
+package service
+
+import (
+	"context"
+	"errors"
+	"fmt"
+	"os"
+	"slices"
+	"strconv"
+	"strings"
+	"time"
+
+	"github.com/charmbracelet/lipgloss"
+	mapset "github.com/deckarep/golang-set/v2"
+	"github.com/docker/docker/pkg/stringid"
+	"github.com/psviderski/uncloud/internal/cli"
+	"github.com/psviderski/uncloud/pkg/api"
+	"github.com/psviderski/uncloud/pkg/client"
+	"github.com/psviderski/uncloud/pkg/client/compose"
+	"github.com/spf13/cobra"
+)
+
+type logsOptions struct {
+	files    []string
+	follow   bool
+	tail     string
+	since    string
+	until    string
+	utc      bool
+	machines []string
+}
+
+func NewLogsCommand(groupID string) *cobra.Command {
+	var options logsOptions
+
+	cmd := &cobra.Command{
+		Use:     "logs [SERVICE...]",
+		Aliases: []string{"log"},
+		Short:   "View service logs.",
+		Long: `View logs from all replicas of the specified service(s) across all machines in the cluster.
+
+If no services are specified, streams logs from all services defined in the Compose file
+(compose.yaml by default or the file(s) specified with --file).`,
+		Example: `  # View recent logs for a service.
+  uc logs web
+
+  # Stream logs in real-time (follow mode).
+  uc logs -f web
+
+  # View logs from multiple services.
+  uc logs web api db
+
+  # View logs from all services in compose.yaml.
+  uc logs
+
+  # Show last 20 lines per replica (default is 100).
+  uc logs -n 20 web
+
+  # Show all logs without line limit.
+  uc logs -n all web
+
+  # View logs from a specific time range.
+  uc logs --since 3h --until 1h30m web
+
+  # View logs only from replicas running on specific machines.
+  uc logs -m machine1,machine2 web api`,
+		RunE: func(cmd *cobra.Command, args []string) error {
+			uncli := cmd.Context().Value("cli").(*cli.CLI)
+			return runLogs(cmd.Context(), uncli, args, options)
+		},
+		GroupID: groupID,
+	}
+
+	cmd.Flags().StringSliceVar(&options.files, "file", nil,
+		"One or more Compose files to load service names from when no services are specified. (default compose.yaml)")
+	cmd.Flags().BoolVarP(&options.follow, "follow", "f", false,
+		"Continually stream new logs.")
+	cmd.Flags().StringSliceVarP(&options.machines, "machine", "m", nil,
+		"Filter logs by machine name or ID. Can be specified multiple times or as a comma-separated list.")
+	cmd.Flags().StringVar(&options.since, "since", "",
+		"Show logs generated on or after the given timestamp. Accepts relative duration, RFC 3339 date, or Unix timestamp.\n"+
+			"Examples:\n"+
+			"  --since 2m30s                      Relative duration (2 minutes 30 seconds ago)\n"+
+			"  --since 1h                         Relative duration (1 hour ago)\n"+
+			"  --since 2025-11-24                 RFC 3339 date only (midnight using local timezone)\n"+
+			"  --since 2024-05-14T22:50:00        RFC 3339 date/time using local timezone\n"+
+			"  --since 2024-01-31T10:30:00Z       RFC 3339 date/time in UTC\n"+
+			"  --since 1763953966                 Unix timestamp (seconds since January 1, 1970)")
+	cmd.Flags().StringVarP(&options.tail, "tail", "n", "100",
+		"Show the most recent logs and limit the number of lines shown per replica. Use 'all' to show all logs.")
+	cmd.Flags().StringVar(&options.until, "until", "",
+		"Show logs generated before the given timestamp. Accepts relative duration, RFC 3339 date, or Unix timestamp.\n"+
+			"See --since for examples.")
+	cmd.Flags().BoolVar(&options.utc, "utc", false,
+		"Print timestamps in UTC instead of local timezone.")
+
+	return cmd
+}
+
+func runLogs(ctx context.Context, uncli *cli.CLI, serviceNames []string, opts logsOptions) error {
+	// If no services specified, try to load them from the Compose file(s).
+	if len(serviceNames) == 0 {
+		project, err := compose.LoadProject(ctx, opts.files)
+		if err != nil {
+			return fmt.Errorf("load compose file(s): %w", err)
+		}
+		// View logs for all services, including disabled by inactive profiles.
+		serviceNames = append(project.ServiceNames(), project.DisabledServiceNames()...)
+		if len(serviceNames) == 0 {
+			return errors.New("no services found in compose file(s)")
+		}
+	}
+
+	// Parse tail option.
+	tail := -1
+	if opts.tail != "all" {
+		tailInt, err := strconv.Atoi(opts.tail)
+		if err != nil {
+			return fmt.Errorf("invalid --tail value '%s': %w", opts.tail, err)
+		}
+		tail = tailInt
+	}
+
+	c, err := uncli.ConnectCluster(ctx)
+	if err != nil {
+		return fmt.Errorf("connect to cluster: %w", err)
+	}
+	defer c.Close()
+
+	logsOpts := api.ServiceLogsOptions{
+		Follow:   opts.follow,
+		Tail:     tail,
+		Since:    opts.since,
+		Until:    opts.until,
+		Machines: cli.ExpandCommaSeparatedValues(opts.machines),
+	}
+
+	// Collect log streams from all services.
+	machineIDsSet := mapset.NewSet[string]()
+	svcStreams := make([]<-chan api.ServiceLogEntry, 0, len(serviceNames))
+	for _, serviceName := range serviceNames {
+		svc, ch, err := c.ServiceLogs(ctx, serviceName, logsOpts)
+		if err != nil {
+			return fmt.Errorf("stream logs for service '%s': %w", serviceName, err)
+		}
+		svcStreams = append(svcStreams, ch)
+
+		machineIDs := svc.MachineIDs()
+		machineIDsSet.Append(machineIDs...)
+	}
+
+	var stream <-chan api.ServiceLogEntry
+	if len(serviceNames) == 1 {
+		stream = svcStreams[0]
+	} else {
+		// Merge all service streams into a single sorted stream without stall detection as its handled per-service.
+		merger := client.NewLogMerger(svcStreams, client.LogMergerOptions{})
+		stream = merger.Stream()
+	}
+
+	// Fetch machine names for all machines (machineIDsSet) service containers are running on.
+	machines, err := c.ListMachines(ctx, &api.MachineFilter{NamesOrIDs: machineIDsSet.ToSlice()})
+	if err != nil {
+		return fmt.Errorf("list machines: %w", err)
+	}
+	machineNames := make([]string, 0, len(machines))
+	for _, m := range machines {
+		machineNames = append(machineNames, m.Machine.Name)
+	}
+
+	formatter := newLogFormatter(machineNames, serviceNames, opts.utc)
+
+	// Print merged logs.
+	for entry := range stream {
+		if entry.Err != nil {
+			formatter.printError(entry)
+			continue
+		}
+		formatter.printEntry(entry)
+	}
+
+	return nil
+}
+
+// Available colors for machine/service differentiation.
+var colorPalette = []lipgloss.Color{
+	lipgloss.Color("10"), // Bright green
+	lipgloss.Color("11"), // Bright yellow
+	lipgloss.Color("12"), // Bright blue
+	lipgloss.Color("13"), // Bright magenta
+	lipgloss.Color("14"), // Bright cyan
+	lipgloss.Color("2"),  // Green
+	lipgloss.Color("3"),  // Yellow
+	lipgloss.Color("4"),  // Blue
+	lipgloss.Color("5"),  // Magenta
+	lipgloss.Color("6"),  // Cyan
+}
+
+// logFormatter handles formatting and printing of log entries with dynamic column alignment.
+type logFormatter struct {
+	machineNames []string
+	serviceNames []string
+
+	maxMachineWidth int
+	maxServiceWidth int
+
+	utc bool
+}
+
+func newLogFormatter(machineNames, serviceNames []string, utc bool) *logFormatter {
+	slices.Sort(machineNames)
+	slices.Sort(serviceNames)
+
+	maxMachineWidth := 0
+	for _, name := range machineNames {
+		if len(name) > maxMachineWidth {
+			maxMachineWidth = len(name)
+		}
+	}
+
+	maxServiceWidth := 0
+	for _, name := range serviceNames {
+		if len(name) > maxServiceWidth {
+			maxServiceWidth = len(name)
+		}
+	}
+
+	return &logFormatter{
+		machineNames:    machineNames,
+		serviceNames:    serviceNames,
+		maxMachineWidth: maxMachineWidth,
+		maxServiceWidth: maxServiceWidth,
+		utc:             utc,
+	}
+}
+
+// formatTimestamp formats timestamp using local timezone or UTC if configured.
+func (f *logFormatter) formatTimestamp(t time.Time) string {
+	if f.utc {
+		t = t.UTC()
+	} else {
+		t = t.In(time.Local)
+	}
+	dimStyle := lipgloss.NewStyle().Faint(true)
+
+	return dimStyle.Render(t.Format(time.StampMilli))
+}
+
+func (f *logFormatter) formatMachine(name string) string {
+	style := lipgloss.NewStyle().Bold(true).PaddingRight(f.maxMachineWidth - len(name))
+
+	if len(f.serviceNames) == 1 {
+		// Machine name is coloured for single-service logs.
+		i := slices.Index(f.machineNames, name)
+		if i == -1 {
+			f.machineNames = append(f.machineNames, name)
+			i = len(f.machineNames) - 1
+		}
+
+		style = style.Foreground(colorPalette[i%len(colorPalette)])
+	}
+
+	return style.Render(name)
+}
+
+func (f *logFormatter) formatServiceContainer(serviceName, containerID string) string {
+	styleService := lipgloss.NewStyle().Bold(true).PaddingRight(f.maxServiceWidth - len(serviceName))
+	styleContainer := lipgloss.NewStyle().Faint(true)
+
+	if len(f.serviceNames) > 1 {
+		// Service name is coloured for multi-service logs.
+		i := slices.Index(f.serviceNames, serviceName)
+		if i == -1 {
+			f.serviceNames = append(f.serviceNames, serviceName)
+			i = len(f.serviceNames) - 1
+		}
+
+		styleService = styleService.Foreground(colorPalette[i%len(colorPalette)])
+	}
+
+	return styleService.Render(serviceName) + styleContainer.Render("["+containerID[:5]+"]")
+}
+
+// printEntry prints a single log entry with proper formatting.
+func (f *logFormatter) printEntry(entry api.ServiceLogEntry) {
+	if entry.Stream != api.LogStreamStdout && entry.Stream != api.LogStreamStderr {
+		return
+	}
+
+	var output strings.Builder
+
+	// Timestamp
+	output.WriteString(f.formatTimestamp(entry.Timestamp))
+	output.WriteString(" ")
+
+	// Machine name
+	output.WriteString(f.formatMachine(entry.Metadata.MachineName))
+	output.WriteString(" ")
+
+	// Service[container_id]
+	output.WriteString(f.formatServiceContainer(entry.Metadata.ServiceName, entry.Metadata.ContainerID))
+	output.WriteString(" ")
+
+	// Message
+	output.Write(entry.Message)
+
+	// Print to appropriate stream.
+	if entry.Stream == api.LogStreamStderr {
+		fmt.Fprint(os.Stderr, output.String())
+	} else {
+		fmt.Print(output.String())
+	}
+}
+
+// printError prints an error entry (e.g., stalled stream warning).
+func (f *logFormatter) printError(entry api.ServiceLogEntry) {
+	if entry.Metadata.ContainerID != "" {
+		msg := fmt.Sprintf("WARNING: log stream from %s[%s] on machine '%s'",
+			entry.Metadata.ServiceName,
+			stringid.TruncateID(entry.Metadata.ContainerID),
+			entry.Metadata.MachineName)
+
+		if errors.Is(entry.Err, api.ErrLogStreamStalled) {
+			msg += " stopped responding"
+		} else {
+			msg += fmt.Sprintf(": %v", entry.Err)
+		}
+
+		style := lipgloss.NewStyle().Bold(true).Foreground(lipgloss.Color("11")) // Bold bright yellow
+		fmt.Fprintln(os.Stderr, style.Render(msg))
+	} else {
+		msg := fmt.Sprintf("ERROR: %v", entry.Err)
+		style := lipgloss.NewStyle().Bold(true).Foreground(lipgloss.Color("9")) // Bold bright red
+		fmt.Fprintln(os.Stderr, style.Render(msg))
+	}
+}

cmd/uncloud/service/ls.go

@@ -4,33 +4,31 @@ import (
 	"context"
 	"fmt"
 	"os"
+	"slices"
 	"strings"
 	"text/tabwriter"
 
 	"github.com/psviderski/uncloud/internal/cli"
+	"github.com/psviderski/uncloud/pkg/api"
 	"github.com/spf13/cobra"
 )
 
-func NewListCommand() *cobra.Command {
-	var contextName string
+func NewListCommand(groupID string) *cobra.Command {
 	cmd := &cobra.Command{
 		Use:     "ls",
 		Aliases: []string{"list"},
 		Short:   "List services.",
 		RunE: func(cmd *cobra.Command, args []string) error {
 			uncli := cmd.Context().Value("cli").(*cli.CLI)
-			return list(cmd.Context(), uncli, contextName)
+			return list(cmd.Context(), uncli)
 		},
+		GroupID: groupID,
 	}
-	cmd.Flags().StringVarP(
-		&contextName, "context", "c", "",
-		"Name of the cluster context. (default is the current context)",
-	)
 	return cmd
 }
 
-func list(ctx context.Context, uncli *cli.CLI, contextName string) error {
-	client, err := uncli.ConnectCluster(ctx, contextName)
+func list(ctx context.Context, uncli *cli.CLI) error {
+	client, err := uncli.ConnectCluster(ctx)
 	if err != nil {
 		return fmt.Errorf("connect to cluster: %w", err)
 	}
@@ -41,15 +39,15 @@ func list(ctx context.Context, uncli *cli.CLI, contextName string) error {
 		return fmt.Errorf("list services: %w", err)
 	}
 
-	serviceNames := make(map[string]struct{}, len(services))
+	// Sort services by name.
 	haveDuplicateNames := false
-	for _, svc := range services {
-		if _, exists := serviceNames[svc.Name]; exists {
+	slices.SortFunc(services, func(a, b api.Service) int {
+		if a.Name == b.Name {
 			haveDuplicateNames = true
-			break
+			return strings.Compare(a.ID, b.ID)
 		}
-		serviceNames[svc.Name] = struct{}{}
-	}
+		return strings.Compare(a.Name, b.Name)
+	})
 
 	// Print the list of services in a table format.
 	tw := tabwriter.NewWriter(os.Stdout, 0, 0, 3, ' ', 0)

cmd/uncloud/service/rm.go

@@ -11,31 +11,32 @@ import (
 
 type rmOptions struct {
 	services []string
-	context  string
 }
 
-func NewRmCommand() *cobra.Command {
+func NewRmCommand(groupID string) *cobra.Command {
 	opts := rmOptions{}
 	cmd := &cobra.Command{
 		Use:     "rm SERVICE [SERVICE...]",
 		Aliases: []string{"remove", "delete"},
 		Short:   "Remove one or more services.",
-		Args:    cobra.MinimumNArgs(1),
+		Long: `Remove one or more services.
+
+The volumes used by the services are preserved and should be removed separately
+with 'uc volume rm'. Anonymous Docker volumes (automatically created from VOLUME
+directives in image Dockerfiles) are automatically removed with their containers.`,
+		Args: cobra.MinimumNArgs(1),
 		RunE: func(cmd *cobra.Command, args []string) error {
 			uncli := cmd.Context().Value("cli").(*cli.CLI)
 			opts.services = args
 			return rm(cmd.Context(), uncli, opts)
 		},
+		GroupID: groupID,
 	}
-	cmd.Flags().StringVarP(
-		&opts.context, "context", "c", "",
-		"Name of the cluster context. (default is the current context)",
-	)
 	return cmd
 }
 
 func rm(ctx context.Context, uncli *cli.CLI, opts rmOptions) error {
-	client, err := uncli.ConnectCluster(ctx, opts.context)
+	client, err := uncli.ConnectCluster(ctx)
 	if err != nil {
 		return fmt.Errorf("connect to cluster: %w", err)
 	}

cmd/uncloud/service/root.go

@@ -8,15 +8,18 @@ func NewRootCommand() *cobra.Command {
 	cmd := &cobra.Command{
 		Use:     "service",
 		Aliases: []string{"svc"},
-		Short:   "Manage services in an Uncloud cluster.",
+		Short:   "Manage services in the cluster.",
 	}
 	cmd.AddCommand(
-		NewInspectCommand(),
-		NewListCommand(),
-		NewRmCommand(),
-		NewRunCommand(),
-		NewScaleCommand(),
-		NewExecCommand(),
+		NewExecCommand(""),
+		NewInspectCommand(""),
+		NewListCommand(""),
+		NewLogsCommand(""),
+		NewRmCommand(""),
+		NewRunCommand(""),
+		NewScaleCommand(""),
+		NewStartCommand(""),
+		NewStopCommand(""),
 	)
 	return cmd
 }

cmd/uncloud/service/run.go

@@ -34,11 +34,9 @@ type runOptions struct {
 	replicas          uint
 	user              string
 	volumes           []string
-
-	context string
 }
 
-func NewRunCommand() *cobra.Command {
+func NewRunCommand(groupID string) *cobra.Command {
 	opts := runOptions{}
 
 	cmd := &cobra.Command{
@@ -56,6 +54,7 @@ func NewRunCommand() *cobra.Command {
 
 			return run(cmd.Context(), uncli, opts)
 		},
+		GroupID: groupID,
 	}
 
 	cmd.Flags().StringVar(&opts.caddyfile, "caddyfile", "",
@@ -112,11 +111,6 @@ func NewRunCommand() *cobra.Command {
 			"  -v /data/uploads:/app/uploads         	 Bind mount /data/uploads host directory to /app/uploads in container\n"+
 			"  -v /host/path:/container/path:ro 		 Bind mount a host directory or file as read-only")
 
-	cmd.Flags().StringVarP(
-		&opts.context, "context", "c", "",
-		"Name of the cluster context to run the service in. (default is the current context)",
-	)
-
 	return cmd
 }
 
@@ -126,7 +120,7 @@ func run(ctx context.Context, uncli *cli.CLI, opts runOptions) error {
 		return err
 	}
 
-	clusterClient, err := uncli.ConnectCluster(ctx, opts.context)
+	clusterClient, err := uncli.ConnectCluster(ctx)
 	if err != nil {
 		return fmt.Errorf("connect to cluster: %w", err)
 	}

cmd/uncloud/service/scale.go

@@ -15,10 +15,9 @@ import (
 type scaleOptions struct {
 	service  string
 	replicas uint
-	context  string
 }
 
-func NewScaleCommand() *cobra.Command {
+func NewScaleCommand(groupID string) *cobra.Command {
 	opts := scaleOptions{}
 	cmd := &cobra.Command{
 		Use:   "scale SERVICE REPLICAS",
@@ -37,13 +36,9 @@ func NewScaleCommand() *cobra.Command {
 
 			return scale(cmd.Context(), uncli, opts)
 		},
+		GroupID: groupID,
 	}
 
-	cmd.Flags().StringVarP(
-		&opts.context, "context", "c", "",
-		"Name of the cluster context. (default is the current context)",
-	)
-
 	return cmd
 }
 
@@ -57,7 +52,7 @@ func scale(ctx context.Context, uncli *cli.CLI, opts scaleOptions) error {
 		)
 	}
 
-	clusterClient, err := uncli.ConnectCluster(ctx, opts.context)
+	clusterClient, err := uncli.ConnectCluster(ctx)
 	if err != nil {
 		return fmt.Errorf("connect to cluster: %w", err)
 	}

cmd/uncloud/service/start.go

@@ -0,0 +1,54 @@
+//nolint:dupl
+package service
+
+import (
+	"context"
+	"fmt"
+
+	"github.com/docker/compose/v2/pkg/progress"
+	"github.com/psviderski/uncloud/internal/cli"
+	"github.com/spf13/cobra"
+)
+
+type startOptions struct {
+	services []string
+}
+
+func NewStartCommand(groupID string) *cobra.Command {
+	opts := startOptions{}
+	cmd := &cobra.Command{
+		Use:   "start SERVICE [SERVICE...]",
+		Short: "Start one or more services.",
+		Long: `Start one or more previously stopped services.
+
+Starts all containers of the specified service(s) across all machines in the cluster.
+Services can be specified by name or ID.`,
+		Args: cobra.MinimumNArgs(1),
+		RunE: func(cmd *cobra.Command, args []string) error {
+			uncli := cmd.Context().Value("cli").(*cli.CLI)
+			opts.services = args
+			return start(cmd.Context(), uncli, opts)
+		},
+		GroupID: groupID,
+	}
+	return cmd
+}
+
+func start(ctx context.Context, uncli *cli.CLI, opts startOptions) error {
+	client, err := uncli.ConnectCluster(ctx)
+	if err != nil {
+		return fmt.Errorf("connect to cluster: %w", err)
+	}
+	defer client.Close()
+
+	for _, s := range opts.services {
+		err = progress.RunWithTitle(ctx, func(ctx context.Context) error {
+			if err = client.StartService(ctx, s); err != nil {
+				return fmt.Errorf("start service '%s': %w", s, err)
+			}
+			return nil
+		}, uncli.ProgressOut(), "Starting service "+s)
+	}
+
+	return err
+}

cmd/uncloud/service/stop.go

@@ -0,0 +1,68 @@
+//nolint:dupl
+package service
+
+import (
+	"context"
+	"fmt"
+
+	"github.com/docker/compose/v2/pkg/progress"
+	"github.com/docker/docker/api/types/container"
+	"github.com/psviderski/uncloud/internal/cli"
+	"github.com/spf13/cobra"
+)
+
+type stopOptions struct {
+	services []string
+	signal   string
+	timeout  int
+}
+
+func NewStopCommand(groupID string) *cobra.Command {
+	opts := stopOptions{}
+	cmd := &cobra.Command{
+		Use:   "stop SERVICE [SERVICE...]",
+		Short: "Stop one or more services.",
+		Long: `Stop one or more running services.
+
+Gracefully stops all containers of the specified service(s) across all machines in the cluster.
+Services can be specified by name or ID. Stopped services can be restarted with 'uc start'.`,
+		Args: cobra.MinimumNArgs(1),
+		RunE: func(cmd *cobra.Command, args []string) error {
+			uncli := cmd.Context().Value("cli").(*cli.CLI)
+			opts.services = args
+			return stop(cmd.Context(), uncli, opts)
+		},
+		GroupID: groupID,
+	}
+	cmd.Flags().StringVarP(&opts.signal, "signal", "s", "",
+		"Signal to send to each container's main process.\n"+
+			"Can be a signal name (SIGTERM, SIGINT, SIGHUP, etc.) or a number. (default SIGTERM)")
+	cmd.Flags().IntVarP(&opts.timeout, "timeout", "t", 10,
+		"Seconds to wait for each container to stop gracefully before forcibly killing it with SIGKILL.\n"+
+			"Use -1 to wait indefinitely.")
+	return cmd
+}
+
+func stop(ctx context.Context, uncli *cli.CLI, opts stopOptions) error {
+	client, err := uncli.ConnectCluster(ctx)
+	if err != nil {
+		return fmt.Errorf("connect to cluster: %w", err)
+	}
+	defer client.Close()
+
+	stopOpts := container.StopOptions{
+		Signal:  opts.signal,
+		Timeout: &opts.timeout,
+	}
+
+	for _, s := range opts.services {
+		err = progress.RunWithTitle(ctx, func(ctx context.Context) error {
+			if err = client.StopService(ctx, s, stopOpts); err != nil {
+				return fmt.Errorf("stop service '%s': %w", s, err)
+			}
+			return nil
+		}, uncli.ProgressOut(), "Stopping service "+s)
+	}
+
+	return err
+}

cmd/uncloud/volume/create.go

@@ -19,7 +19,6 @@ type createOptions struct {
 	driverOpts []string
 	labels     []string
 	machine    string
-	context    string
 }
 
 func NewCreateCommand() *cobra.Command {
@@ -50,14 +49,12 @@ func NewCreateCommand() *cobra.Command {
 		"Labels to assign to the volume in the form of 'key=value' pairs. Can be specified multiple times.")
 	cmd.Flags().StringVarP(&opts.machine, "machine", "m", "",
 		"Name or ID of the machine to create the volume on.")
-	cmd.Flags().StringVarP(&opts.context, "context", "c", "",
-		"Name of the cluster context. (default is the current context)")
 
 	return cmd
 }
 
 func create(ctx context.Context, uncli *cli.CLI, name string, opts createOptions) error {
-	client, err := uncli.ConnectCluster(ctx, opts.context)
+	client, err := uncli.ConnectCluster(ctx)
 	if err != nil {
 		return fmt.Errorf("connect to cluster: %w", err)
 	}

cmd/uncloud/volume/inspect.go

@@ -13,7 +13,6 @@ import (
 
 type inspectOptions struct {
 	machine string
-	context string
 }
 
 func NewInspectCommand() *cobra.Command {
@@ -32,14 +31,12 @@ func NewInspectCommand() *cobra.Command {
 	cmd.Flags().StringVarP(&opts.machine, "machine", "m", "",
 		"Name or ID of the machine where the volume is located. "+
 			"If not specified, the volume will be searched across all machines.")
-	cmd.Flags().StringVarP(&opts.context, "context", "c", "",
-		"Name of the cluster context. (default is the current context)")
 
 	return cmd
 }
 
 func inspect(ctx context.Context, uncli *cli.CLI, name string, opts inspectOptions) error {
-	client, err := uncli.ConnectCluster(ctx, opts.context)
+	client, err := uncli.ConnectCluster(ctx)
 	if err != nil {
 		return fmt.Errorf("connect to cluster: %w", err)
 	}

cmd/uncloud/volume/ls.go

@@ -16,7 +16,6 @@ import (
 type listOptions struct {
 	machines []string
 	quiet    bool
-	context  string
 }
 
 func NewListCommand() *cobra.Command {
@@ -38,14 +37,11 @@ func NewListCommand() *cobra.Command {
 	cmd.Flags().BoolVarP(&opts.quiet, "quiet", "q", false,
 		"Only display volume names.")
 
-	cmd.Flags().StringVarP(&opts.context, "context", "c", "",
-		"Name of the cluster context. (default is the current context)")
-
 	return cmd
 }
 
 func list(ctx context.Context, uncli *cli.CLI, opts listOptions) error {
-	client, err := uncli.ConnectCluster(ctx, opts.context)
+	client, err := uncli.ConnectCluster(ctx)
 	if err != nil {
 		return fmt.Errorf("connect to cluster: %w", err)
 	}

cmd/uncloud/volume/rm.go

@@ -14,7 +14,6 @@ type removeOptions struct {
 	force    bool
 	machines []string
 	yes      bool
-	context  string
 }
 
 func NewRemoveCommand() *cobra.Command {
@@ -41,14 +40,11 @@ func NewRemoveCommand() *cobra.Command {
 	cmd.Flags().BoolVarP(&opts.yes, "yes", "y", false,
 		"Do not prompt for confirmation before removing the volume(s).")
 
-	cmd.Flags().StringVarP(&opts.context, "context", "c", "",
-		"Name of the cluster context. (default is the current context)")
-
 	return cmd
 }
 
 func remove(ctx context.Context, uncli *cli.CLI, names []string, opts removeOptions) error {
-	client, err := uncli.ConnectCluster(ctx, opts.context)
+	client, err := uncli.ConnectCluster(ctx)
 	if err != nil {
 		return fmt.Errorf("connect to cluster: %w", err)
 	}

cmd/uncloud/volume/root.go

@@ -7,7 +7,7 @@ import (
 func NewRootCommand() *cobra.Command {
 	cmd := &cobra.Command{
 		Use:   "volume",
-		Short: "Manage volumes in an Uncloud cluster.",
+		Short: "Manage volumes in the cluster.",
 	}
 	cmd.AddCommand(
 		NewCreateCommand(),

go.mod

@@ -9,9 +9,10 @@ require (
 	github.com/alecthomas/chroma/v2 v2.20.0
 	github.com/caddyserver/caddy/v2 v2.8.4
 	github.com/cenkalti/backoff/v4 v4.3.0
-	github.com/charmbracelet/bubbles v0.20.0
-	github.com/charmbracelet/bubbletea v1.3.9
+	github.com/charmbracelet/bubbles v0.21.0
+	github.com/charmbracelet/bubbletea v1.3.10
 	github.com/charmbracelet/huh v0.6.0
+	github.com/charmbracelet/huh/spinner v0.0.0-20251110114415-25888d17260b
 	github.com/charmbracelet/lipgloss v1.1.0
 	github.com/compose-spec/compose-go/v2 v2.9.0
 	github.com/containerd/errdefs v1.0.0
@@ -45,7 +46,7 @@ require (
 	github.com/muesli/termenv v0.16.0
 	github.com/opencontainers/go-digest v1.0.0
 	github.com/opencontainers/image-spec v1.1.1
-	github.com/psviderski/unregistry v0.3.1
+	github.com/psviderski/unregistry v0.4.1
 	github.com/siderolabs/discovery-api v0.1.4
 	github.com/siderolabs/discovery-client v0.1.9
 	github.com/siderolabs/grpc-proxy v0.5.1
@@ -149,7 +150,7 @@ require (
 	github.com/eiannone/keyboard v0.0.0-20220611211555-0d226195f203 // indirect
 	github.com/emicklei/go-restful/v3 v3.11.0 // indirect
 	github.com/erikgeiser/coninput v0.0.0-20211004153227-1c3628e74d0f // indirect
-	github.com/fatih/color v1.17.0 // indirect
+	github.com/fatih/color v1.18.0 // indirect
 	github.com/felixge/httpsnoop v1.0.4 // indirect
 	github.com/fsnotify/fsevents v0.2.0 // indirect
 	github.com/fvbommel/sortorder v1.1.0 // indirect

go.sum

@@ -168,22 +168,24 @@ github.com/cespare/xxhash/v2 v2.1.1/go.mod h1:VGX0DQ3Q6kWi7AoAeZDth3/j3BFtOZR5XL
 github.com/cespare/xxhash/v2 v2.2.0/go.mod h1:VGX0DQ3Q6kWi7AoAeZDth3/j3BFtOZR5XLFGgcrjCOs=
 github.com/cespare/xxhash/v2 v2.3.0 h1:UL815xU9SqsFlibzuggzjXhog7bL6oX9BbNZnL2UFvs=
 github.com/cespare/xxhash/v2 v2.3.0/go.mod h1:VGX0DQ3Q6kWi7AoAeZDth3/j3BFtOZR5XLFGgcrjCOs=
-github.com/charmbracelet/bubbles v0.20.0 h1:jSZu6qD8cRQ6k9OMfR1WlM+ruM8fkPWkHvQWD9LIutE=
-github.com/charmbracelet/bubbles v0.20.0/go.mod h1:39slydyswPy+uVOHZ5x/GjwVAFkCsV8IIVy+4MhzwwU=
-github.com/charmbracelet/bubbletea v1.3.9 h1:OBYdfRo6QnlIcXNmcoI2n1NNS65Nk6kI2L2FO1puS/4=
-github.com/charmbracelet/bubbletea v1.3.9/go.mod h1:ORQfo0fk8U+po9VaNvnV95UPWA1BitP1E0N6xJPlHr4=
+github.com/charmbracelet/bubbles v0.21.0 h1:9TdC97SdRVg/1aaXNVWfFH3nnLAwOXr8Fn6u6mfQdFs=
+github.com/charmbracelet/bubbles v0.21.0/go.mod h1:HF+v6QUR4HkEpz62dx7ym2xc71/KBHg+zKwJtMw+qtg=
+github.com/charmbracelet/bubbletea v1.3.10 h1:otUDHWMMzQSB0Pkc87rm691KZ3SWa4KUlvF9nRvCICw=
+github.com/charmbracelet/bubbletea v1.3.10/go.mod h1:ORQfo0fk8U+po9VaNvnV95UPWA1BitP1E0N6xJPlHr4=
 github.com/charmbracelet/colorprofile v0.2.3-0.20250311203215-f60798e515dc h1:4pZI35227imm7yK2bGPcfpFEmuY1gc2YSTShr4iJBfs=
 github.com/charmbracelet/colorprofile v0.2.3-0.20250311203215-f60798e515dc/go.mod h1:X4/0JoqgTIPSFcRA/P6INZzIuyqdFY5rm8tb41s9okk=
 github.com/charmbracelet/huh v0.6.0 h1:mZM8VvZGuE0hoDXq6XLxRtgfWyTI3b2jZNKh0xWmax8=
 github.com/charmbracelet/huh v0.6.0/go.mod h1:GGNKeWCeNzKpEOh/OJD8WBwTQjV3prFAtQPpLv+AVwU=
+github.com/charmbracelet/huh/spinner v0.0.0-20251110114415-25888d17260b h1:oefmPctgff7OB4cDz3Ndp+ewwrh7i+fwth70qvZCieI=
+github.com/charmbracelet/huh/spinner v0.0.0-20251110114415-25888d17260b/go.mod h1:OMqKat/mm9a/qOnpuNOPyYO9bPzRNnmzLnRZT5KYltg=
 github.com/charmbracelet/lipgloss v1.1.0 h1:vYXsiLHVkK7fp74RkV7b2kq9+zDLoEU4MZoFqR/noCY=
 github.com/charmbracelet/lipgloss v1.1.0/go.mod h1:/6Q8FR2o+kj8rz4Dq0zQc3vYf7X+B0binUUBwA0aL30=
 github.com/charmbracelet/x/ansi v0.10.1 h1:rL3Koar5XvX0pHGfovN03f5cxLbCF2YvLeyz7D2jVDQ=
 github.com/charmbracelet/x/ansi v0.10.1/go.mod h1:3RQDQ6lDnROptfpWuUVIUG64bD2g2BgntdxH0Ya5TeE=
 github.com/charmbracelet/x/cellbuf v0.0.13-0.20250311204145-2c3ea96c31dd h1:vy0GVL4jeHEwG5YOXDmi86oYw2yuYUGqz6a8sLwg0X8=
 github.com/charmbracelet/x/cellbuf v0.0.13-0.20250311204145-2c3ea96c31dd/go.mod h1:xe0nKWGd3eJgtqZRaN9RjMtK7xUYchjzPr7q6kcvCCs=
-github.com/charmbracelet/x/exp/golden v0.0.0-20240815200342-61de596daa2b h1:MnAMdlwSltxJyULnrYbkZpp4k58Co7Tah3ciKhSNo0Q=
-github.com/charmbracelet/x/exp/golden v0.0.0-20240815200342-61de596daa2b/go.mod h1:wDlXFlCrmJ8J+swcL/MnGUuYnqgQdW9rhSD61oNMb6U=
+github.com/charmbracelet/x/exp/golden v0.0.0-20241011142426-46044092ad91 h1:payRxjMjKgx2PaCWLZ4p3ro9y97+TVLZNaRZgJwSVDQ=
+github.com/charmbracelet/x/exp/golden v0.0.0-20241011142426-46044092ad91/go.mod h1:wDlXFlCrmJ8J+swcL/MnGUuYnqgQdW9rhSD61oNMb6U=
 github.com/charmbracelet/x/exp/strings v0.0.0-20240919170804-a4978c8e603a h1:JMdM89Udp/cOl5tC3MuUJXTPE/nAdU1oyt9jRU44qq8=
 github.com/charmbracelet/x/exp/strings v0.0.0-20240919170804-a4978c8e603a/go.mod h1:pBhA0ybfXv6hDjQUZ7hk1lVxBiUbupdw5R31yPUViVQ=
 github.com/charmbracelet/x/term v0.2.1 h1:AQeHeLZ1OqSXhrAWpYUtZyX1T3zVxfpZuEQMIQaGIAQ=
@@ -344,8 +346,8 @@ github.com/erikgeiser/coninput v0.0.0-20211004153227-1c3628e74d0f/go.mod h1:vw97
 github.com/erikstmartin/go-testdb v0.0.0-20160219214506-8d10e4a1bae5/go.mod h1:a2zkGnVExMxdzMo3M0Hi/3sEU+cWnZpSni0O6/Yb/P0=
 github.com/fatih/color v1.7.0/go.mod h1:Zm6kSWBoL9eyXnKyktHP6abPY2pDugNf5KwzbycvMj4=
 github.com/fatih/color v1.9.0/go.mod h1:eQcE1qtQxscV5RaZvpXrrb8Drkc3/DdQ+uUYCNjL+zU=
-github.com/fatih/color v1.17.0 h1:GlRw1BRJxkpqUCBKzKOw098ed57fEsKeNjpTe3cSjK4=
-github.com/fatih/color v1.17.0/go.mod h1:YZ7TlrGPkiz6ku9fK3TLD/pl3CpsiFyu8N92HLgmosI=
+github.com/fatih/color v1.18.0 h1:S8gINlzdQ840/4pfAwic/ZE0djQEH3wM94VfqLTZcOM=
+github.com/fatih/color v1.18.0/go.mod h1:4FelSpRwEGDpQ12mAdzqdOukCy4u8WUtOY6lkT/6HfU=
 github.com/felixge/httpsnoop v1.0.4 h1:NFTV2Zj1bL4mc9sqWACXbQFVBBg2W3GPvqp8/ESS2Wg=
 github.com/felixge/httpsnoop v1.0.4/go.mod h1:m8KPJKqk1gH5J9DgRY2ASl2lWCfGKXixSwevea8zH2U=
 github.com/flynn/noise v1.1.0 h1:KjPQoQCEFdZDiP03phOvGi11+SVVhBG2wOWAorLsstg=
@@ -1038,8 +1040,8 @@ github.com/prometheus/procfs v0.0.3/go.mod h1:4A/X28fw3Fc593LaREMrKMqOKvUAntwMDa
 github.com/prometheus/procfs v0.0.8/go.mod h1:7Qr8sr6344vo1JqZ6HhLceV9o3AJ1Ff+GxbHq6oeK9A=
 github.com/prometheus/procfs v0.15.1 h1:YagwOFzUgYfKKHX6Dr+sHT7km/hxC76UB0learggepc=
 github.com/prometheus/procfs v0.15.1/go.mod h1:fB45yRUv8NstnjriLhBQLuOUt+WW4BsoGhij/e3PBqk=
-github.com/psviderski/unregistry v0.3.1 h1:2xQ3uMNMyXAASqq9+kGUgNMIFOvPSMIFGdz47uDfjhQ=
-github.com/psviderski/unregistry v0.3.1/go.mod h1:DXLnzrskO5NuRlzZhpbyo1E6KpirnctUvq6WtCGkpUQ=
+github.com/psviderski/unregistry v0.4.1 h1:KedQTjyj8hHbzygK9uw5YS10amMZ5cjSYHqSjUUBhoE=
+github.com/psviderski/unregistry v0.4.1/go.mod h1:DXLnzrskO5NuRlzZhpbyo1E6KpirnctUvq6WtCGkpUQ=
 github.com/quic-go/qpack v0.5.1 h1:giqksBPnT/HDtZ6VhtFKgoLOWmlyo9Ei6u9PqzIMbhI=
 github.com/quic-go/qpack v0.5.1/go.mod h1:+PC4XFrEskIVkcLzpEkbLqq1uCoxPhQuvK5rH1ZgaEg=
 github.com/quic-go/quic-go v0.48.2 h1:wsKXZPeGWpMpCGSWqOcqpW2wZYic/8T3aqiOID0/KWE=

internal/cli/build.go

@@ -40,8 +40,6 @@ type BuildServicesOptions struct {
 	PushRegistry bool
 
 	// Cluster-specific options (only used if PushCluster is true).
-	// Context is the name of the cluster context.
-	Context string
 	// Machines is a list of machine names or IDs to push the image to. If empty, images are pushed to all machines.
 	Machines []string
 }
@@ -95,7 +93,7 @@ func (cli *CLI) BuildServices(ctx context.Context, project *composetypes.Project
 	// Add a line break after the build output.
 	fmt.Fprintln(cli.ProgressOut())
 
-	clusterClient, err := cli.ConnectCluster(ctx, opts.Context)
+	clusterClient, err := cli.ConnectCluster(ctx)
 	if err != nil {
 		return fmt.Errorf("connect to cluster: %w", err)
 	}

internal/cli/cli.go

@@ -28,14 +28,16 @@ const (
 )
 
 type CLI struct {
-	Config *config.Config
-	conn   *config.MachineConnection
+	Config          *config.Config
+	conn            *config.MachineConnection
+	contextOverride string
 }
 
 // New creates a new CLI instance with the given config path or remote machine connection.
 // If the connection is provided, the config is ignored for all operations which is useful for interacting with
 // a cluster without creating a config.
-func New(configPath string, conn *config.MachineConnection) (*CLI, error) {
+// If a non-empty context name is given, it will override the current default.
+func New(configPath string, conn *config.MachineConnection, contextName string) (*CLI, error) {
 	if conn != nil {
 		return &CLI{conn: conn}, nil
 	}
@@ -46,7 +48,8 @@ func New(configPath string, conn *config.MachineConnection) (*CLI, error) {
 	}
 
 	return &CLI{
-		Config: cfg,
+		Config:          cfg,
+		contextOverride: contextName,
 	}, nil
 }
 
@@ -68,19 +71,28 @@ func (cli *CLI) SetCurrentContext(name string) error {
 	return cli.Config.Save()
 }
 
-// ConnectCluster connects to a cluster using the given context name or the current context if not specified.
+func (cli *CLI) GetContextOverrideOrCurrent() string {
+	contextName := cli.contextOverride
+	if contextName == "" {
+		contextName = cli.Config.CurrentContext
+	}
+	return contextName
+}
+
+// ConnectCluster connects to a cluster using the context override or the current context if not specified.
 // If the CLI was initialised with a machine connection, the config is ignored and the connection is used instead.
-func (cli *CLI) ConnectCluster(ctx context.Context, contextName string) (*client.Client, error) {
-	return cli.ConnectClusterWithOptions(ctx, contextName, ConnectOptions{
+func (cli *CLI) ConnectCluster(ctx context.Context) (*client.Client, error) {
+	return cli.ConnectClusterWithOptions(ctx, ConnectOptions{
 		// Default to showing progress for CLI usage.
 		ShowProgress: true,
 	})
 }
 
-// ConnectClusterWithOptions connects to a cluster using the given context name and options.
+// ConnectClusterWithOptions connects to a cluster with the given options.
 // If the CLI was initialised with a machine connection, the config is ignored and the connection is used instead.
+// If the CLI has an override context, it is used instead of the current default.
 // Options are useful when using the CLI as a library where you may want to disable visual feedback.
-func (cli *CLI) ConnectClusterWithOptions(ctx context.Context, contextName string, opts ConnectOptions) (*client.Client, error) {
+func (cli *CLI) ConnectClusterWithOptions(ctx context.Context, opts ConnectOptions) (*client.Client, error) {
 	if cli.conn != nil {
 		return ConnectCluster(ctx, *cli.conn, opts)
 	}
@@ -92,6 +104,8 @@ func (cli *CLI) ConnectClusterWithOptions(ctx context.Context, contextName strin
 			cli.Config.Path(),
 		)
 	}
+
+	contextName := cli.contextOverride
 	if contextName == "" {
 		// If the cluster is not specified, use the current cluster if set.
 		if cli.Config.CurrentContext == "" {
@@ -157,7 +171,7 @@ func (cli *CLI) InitCluster(ctx context.Context, opts InitClusterOptions) (*clie
 		return cli.initRemoteMachine(ctx, opts)
 	}
 	// TODO: implement local machine initialisation
-	return nil, fmt.Errorf("local machine initialisation is not implemented yet")
+	return nil, fmt.Errorf("local machine initialisation is not implemented yet. Please specify a remote machine")
 }
 
 func (cli *CLI) initRemoteMachine(ctx context.Context, opts InitClusterOptions) (*client.Client, error) {
@@ -228,8 +242,21 @@ func (cli *CLI) initRemoteMachine(ctx context.Context, opts InitClusterOptions)
 
 	// Save the machine's SSH connection details in the context config.
 	connCfg := config.MachineConnection{
-		SSH:        config.NewSSHDestination(opts.RemoteMachine.User, opts.RemoteMachine.Host, opts.RemoteMachine.Port),
 		SSHKeyFile: opts.RemoteMachine.KeyPath,
+		MachineID:  resp.Machine.Id,
+	}
+	if opts.RemoteMachine.UseSSHCLI {
+		connCfg.SSHCLI = config.NewSSHDestination(
+			opts.RemoteMachine.User,
+			opts.RemoteMachine.Host,
+			opts.RemoteMachine.Port,
+		)
+	} else {
+		connCfg.SSH = config.NewSSHDestination(
+			opts.RemoteMachine.User,
+			opts.RemoteMachine.Host,
+			opts.RemoteMachine.Port,
+		)
 	}
 	cli.Config.Contexts[contextName].Connections = append(cli.Config.Contexts[contextName].Connections, connCfg)
 	if err = cli.Config.Save(); err != nil {
@@ -265,7 +292,6 @@ func (cli *CLI) newContextName(name string) (string, error) {
 }
 
 type AddMachineOptions struct {
-	Context       string
 	MachineName   string
 	PublicIP      *netip.Addr
 	RemoteMachine *RemoteMachine
@@ -278,11 +304,8 @@ type AddMachineOptions struct {
 // cluster. The machine client is connected to the new machine and can be used to interact with it.
 // Both client should be closed after use by the caller.
 func (cli *CLI) AddMachine(ctx context.Context, opts AddMachineOptions) (*client.Client, *client.Client, error) {
-	contextName := opts.Context
-	if contextName == "" {
-		contextName = cli.Config.CurrentContext
-	}
-	c, err := cli.ConnectCluster(ctx, contextName)
+	contextName := cli.GetContextOverrideOrCurrent()
+	c, err := cli.ConnectCluster(ctx)
 	if err != nil {
 		return nil, nil, fmt.Errorf("connect to cluster (context '%s'): %w", contextName, err)
 	}
@@ -396,11 +419,21 @@ func (cli *CLI) AddMachine(ctx context.Context, opts AddMachineOptions) (*client
 
 	// Save the machine's SSH connection details in the context config.
 	connCfg := config.MachineConnection{
-		SSH:        config.NewSSHDestination(opts.RemoteMachine.User, opts.RemoteMachine.Host, opts.RemoteMachine.Port),
 		SSHKeyFile: opts.RemoteMachine.KeyPath,
+		MachineID:  addResp.Machine.Id,
 	}
-	if contextName == "" {
-		contextName = cli.Config.CurrentContext
+	if opts.RemoteMachine.UseSSHCLI {
+		connCfg.SSHCLI = config.NewSSHDestination(
+			opts.RemoteMachine.User,
+			opts.RemoteMachine.Host,
+			opts.RemoteMachine.Port,
+		)
+	} else {
+		connCfg.SSH = config.NewSSHDestination(
+			opts.RemoteMachine.User,
+			opts.RemoteMachine.Host,
+			opts.RemoteMachine.Port,
+		)
 	}
 	cli.Config.Contexts[contextName].Connections = append(cli.Config.Contexts[contextName].Connections, connCfg)
 	if err = cli.Config.Save(); err != nil {
@@ -420,6 +453,35 @@ func (cli *CLI) AddMachine(ctx context.Context, opts AddMachineOptions) (*client
 func provisionOrConnectRemoteMachine(
 	ctx context.Context, remoteMachine *RemoteMachine, skipInstall bool, version string,
 ) (*client.Client, error) {
+	// Use SSH CLI
+	if remoteMachine.UseSSHCLI {
+		exec := sshexec.NewSSHCLIRemote(
+			remoteMachine.User,
+			remoteMachine.Host,
+			remoteMachine.Port,
+			remoteMachine.KeyPath,
+		)
+
+		if !skipInstall {
+			if err := provisionMachine(ctx, exec, version); err != nil {
+				return nil, fmt.Errorf("provision machine: %w", err)
+			}
+		}
+
+		sshConfig := &connector.SSHConnectorConfig{
+			User:    remoteMachine.User,
+			Host:    remoteMachine.Host,
+			Port:    remoteMachine.Port,
+			KeyPath: remoteMachine.KeyPath,
+		}
+		machineClient, err := client.New(ctx, connector.NewSSHCLIConnector(sshConfig))
+		if err != nil {
+			return nil, fmt.Errorf("connect to remote machine: %w", err)
+		}
+		return machineClient, nil
+	}
+
+	// Use Go SSH
 	sshClient, err := sshexec.Connect(remoteMachine.User, remoteMachine.Host, remoteMachine.Port, remoteMachine.KeyPath)
 	// If the SSH connection using SSH agent fails and no key path is provided, try to use the default SSH key.
 	if err != nil && remoteMachine.KeyPath == "" {

internal/cli/config/connection.go

@@ -22,9 +22,12 @@ type MachineConnection struct {
 	SSHKeyFile string         `yaml:"ssh_key_file,omitempty"`
 	// TCP is the address and port of the machine's API server.
 	// The pointer is used to omit the field when not set. Otherwise, yaml marshalling includes an empty object.
-	TCP       *netip.AddrPort `yaml:"tcp,omitempty"`
-	Host      string          `yaml:"host,omitempty"`
-	PublicKey secret.Secret   `yaml:"public_key,omitempty"`
+	TCP *netip.AddrPort `yaml:"tcp,omitempty"`
+	// Unix is the path to the machine's API unix socket.
+	Unix      string        `yaml:"unix,omitempty"`
+	Host      string        `yaml:"host,omitempty"`
+	PublicKey secret.Secret `yaml:"public_key,omitempty"`
+	MachineID string        `yaml:"machine_id,omitempty"`
 }
 
 func (c MachineConnection) String() string {
@@ -34,6 +37,8 @@ func (c MachineConnection) String() string {
 		return "ssh+cli://" + string(c.SSHCLI)
 	} else if c.TCP != nil && c.TCP.IsValid() {
 		return fmt.Sprintf("tcp://%s", c.TCP)
+	} else if c.Unix != "" {
+		return fmt.Sprintf("unix://%s", c.Unix)
 	}
 	return "unknown connection"
 }
@@ -49,12 +54,15 @@ func (c *MachineConnection) Validate() error {
 	if c.TCP != nil && c.TCP.IsValid() {
 		setCount++
 	}
+	if c.Unix != "" {
+		setCount++
+	}
 
 	if setCount == 0 {
-		return errors.New("no connection method specified (ssh, ssh_cli, or tcp required)")
+		return errors.New("no connection method specified (ssh, ssh_cli, tcp, or unix required)")
 	}
 	if setCount > 1 {
-		return errors.New("only one connection method allowed per connection (ssh, ssh_cli, or tcp)")
+		return errors.New("only one connection method allowed per connection (ssh, ssh_cli, tcp, or unix)")
 	}
 
 	return nil

internal/cli/config/connection_test.go

@@ -53,11 +53,27 @@ func TestMachineConnection_String(t *testing.T) {
 			},
 			want: "tcp://10.0.0.1:8080",
 		},
+		{
+			name: "unix connection",
+			conn: MachineConnection{
+				Unix: "/run/uncloud/uncloud.sock",
+			},
+			want: "unix:///run/uncloud/uncloud.sock",
+		},
 		{
 			name: "no connection",
 			conn: MachineConnection{},
 			want: "unknown connection",
 		},
+		{
+			name: "ssh connection with machine id",
+			conn: MachineConnection{
+				SSH:       "user@host.com",
+				MachineID: "ed98b9f7575308c340263cd279e3b568",
+			},
+			// String() does not use MachineID; just verifying entry with it is valid
+			want: "ssh://user@host.com",
+		},
 	}
 
 	for _, tt := range tests {
@@ -103,6 +119,13 @@ func TestMachineConnection_Validate(t *testing.T) {
 			},
 			wantErr: false,
 		},
+		{
+			name: "unix only - valid",
+			conn: MachineConnection{
+				Unix: "/path/to/socket",
+			},
+			wantErr: false,
+		},
 		{
 			name:    "no connection method - error",
 			conn:    MachineConnection{},
@@ -118,6 +141,15 @@ func TestMachineConnection_Validate(t *testing.T) {
 			wantErr: true,
 			errMsg:  "only one connection method allowed",
 		},
+		{
+			name: "ssh and unix - error",
+			conn: MachineConnection{
+				SSH:  "user@host",
+				Unix: "/path/to/socket",
+			},
+			wantErr: true,
+			errMsg:  "only one connection method allowed",
+		},
 		{
 			name: "ssh and tcp - error",
 			conn: MachineConnection{

internal/cli/config/context.go

@@ -4,3 +4,21 @@ type Context struct {
 	Name        string              `yaml:"-"`
 	Connections []MachineConnection `yaml:"connections"`
 }
+
+func (c *Context) SetDefaultConnection(index int) {
+	// Do nothing if the index is out of range.
+	if index < 0 || index >= len(c.Connections) {
+		return
+	}
+
+	selected := c.Connections[index]
+	newConnections := make([]MachineConnection, 0, len(c.Connections))
+	newConnections = append(newConnections, selected)
+
+	for i, conn := range c.Connections {
+		if i != index {
+			newConnections = append(newConnections, conn)
+		}
+	}
+	c.Connections = newConnections
+}

internal/cli/connect.go

@@ -73,6 +73,8 @@ func connectCluster(ctx context.Context, conn config.MachineConnection) (*client
 		useSSHCLI = true
 	} else if conn.TCP != nil && conn.TCP.IsValid() {
 		return client.New(ctx, connector.NewTCPConnector(*conn.TCP))
+	} else if conn.Unix != "" {
+		return client.New(ctx, connector.NewUnixConnector(conn.Unix))
 	} else {
 		return nil, errors.New("connection configuration is invalid")
 	}

internal/cli/machine.go

@@ -21,10 +21,11 @@ const (
 )
 
 type RemoteMachine struct {
-	User    string
-	Host    string
-	Port    int
-	KeyPath string
+	User      string
+	Host      string
+	Port      int
+	KeyPath   string
+	UseSSHCLI bool // indicates ssh+cli:// should be used
 }
 
 func installCmd(user string, version string) string {

internal/machine/api/pb/docker.proto

@@ -5,6 +5,7 @@ package api;
 option go_package = "github.com/psviderski/uncloud/internal/machine/api/pb";
 
 import "google/protobuf/empty.proto";
+import "google/protobuf/timestamp.proto";
 import "internal/machine/api/pb/common.proto";
 
 service Docker {
@@ -15,6 +16,9 @@ service Docker {
   rpc ListContainers(ListContainersRequest) returns (ListContainersResponse);
   rpc RemoveContainer(RemoveContainerRequest) returns (google.protobuf.Empty);
 
+  rpc ExecContainer(stream ExecContainerRequest) returns (stream ExecContainerResponse);
+  rpc ContainerLogs(ContainerLogsRequest) returns (stream ContainerLogEntry);
+
   rpc PullImage(PullImageRequest) returns (stream JSONMessage);
   rpc InspectImage(InspectImageRequest) returns (InspectImageResponse);
   // InspectRemoteImage returns the image metadata for an image in a remote registry using the machine's
@@ -30,8 +34,6 @@ service Docker {
   rpc InspectServiceContainer(InspectContainerRequest) returns (ServiceContainer);
   rpc ListServiceContainers(ListServiceContainersRequest) returns (ListServiceContainersResponse);
   rpc RemoveServiceContainer(RemoveContainerRequest) returns (google.protobuf.Empty);
-
-  rpc ExecContainer(stream ExecContainerRequest) returns (stream ExecContainerResponse);
 }
 
 message CreateContainerRequest {
@@ -94,6 +96,64 @@ message RemoveContainerRequest {
   bytes options = 2;
 }
 
+message ExecContainerRequest {
+  oneof payload {
+    // Initial configuration for the exec session. Must be sent as the first message.
+    ExecConfig config = 1;
+    // Raw stdin data to be written to the exec process.
+    bytes stdin = 2;
+    // TTY resize event (only used when TTY is enabled).
+    ResizeEvent resize = 3;
+  }
+}
+
+message ExecConfig {
+  // Container ID to execute the command in.
+  string container_id = 1;
+  // JSON serialised ExecOptions
+  bytes options = 2;
+}
+
+message ResizeEvent {
+  uint32 height = 1;
+  uint32 width = 2;
+}
+
+message ExecContainerResponse {
+  oneof payload {
+    // Exec instance ID returned after creating the exec.
+    string exec_id = 1;
+    // Raw stdout data from the exec process.
+    bytes stdout = 2;
+    // Raw stderr data from the exec process (only when TTY is disabled).
+    bytes stderr = 3;
+    // Exit code of the exec process. Sent as the final message.
+    int32 exit_code = 4;
+  }
+}
+
+message ContainerLogsRequest {
+  string container_id = 1;
+  // Options for logs retrieval.
+  bool follow = 2;
+  int32 tail = 3; // -1 means all
+  string since = 4; // https://www.rfc-editor.org/rfc/rfc3339.html timestamp or Go duration string
+  string until = 5; // https://www.rfc-editor.org/rfc/rfc3339.html timestamp or Go duration string
+}
+
+message ContainerLogEntry {
+  enum StreamType {
+    UNKNOWN = 0;
+    STDOUT = 1;
+    STDERR = 2;
+    HEARTBEAT = 3;
+  }
+  StreamType stream = 1;
+  google.protobuf.Timestamp timestamp = 2;
+  // Log line content. Empty for heartbeat entries.
+  bytes message = 3;
+}
+
 message PullImageRequest {
   string image = 1;
   // JSON serialised image.PullOptions.
@@ -216,39 +276,3 @@ message MachineServiceContainers {
   Metadata metadata = 1;
   repeated ServiceContainer containers = 2;
 }
-
-message ExecContainerRequest {
-  oneof payload {
-    // Initial configuration for the exec session. Must be sent as the first message.
-    ExecConfig config = 1;
-    // Raw stdin data to be written to the exec process.
-    bytes stdin = 2;
-    // TTY resize event (only used when TTY is enabled).
-    ResizeEvent resize = 3;
-  }
-}
-
-message ExecConfig {
-  // Container ID to execute the command in.
-  string container_id = 1;
-  // JSON serialised ExecOptions
-  bytes options = 2;
-}
-
-message ResizeEvent {
-  uint32 height = 1;
-  uint32 width = 2;
-}
-
-message ExecContainerResponse {
-  oneof payload {
-    // Exec instance ID returned after creating the exec.
-    string exec_id = 1;
-    // Raw stdout data from the exec process.
-    bytes stdout = 2;
-    // Raw stderr data from the exec process (only when TTY is disabled).
-    bytes stderr = 3;
-    // Exit code of the exec process. Sent as the final message.
-    int32 exit_code = 4;
-  }
-}

internal/machine/api/pb/docker_grpc.pb.go

@@ -26,6 +26,8 @@ const (
 	Docker_StopContainer_FullMethodName           = "/api.Docker/StopContainer"
 	Docker_ListContainers_FullMethodName          = "/api.Docker/ListContainers"
 	Docker_RemoveContainer_FullMethodName         = "/api.Docker/RemoveContainer"
+	Docker_ExecContainer_FullMethodName           = "/api.Docker/ExecContainer"
+	Docker_ContainerLogs_FullMethodName           = "/api.Docker/ContainerLogs"
 	Docker_PullImage_FullMethodName               = "/api.Docker/PullImage"
 	Docker_InspectImage_FullMethodName            = "/api.Docker/InspectImage"
 	Docker_InspectRemoteImage_FullMethodName      = "/api.Docker/InspectRemoteImage"
@@ -37,7 +39,6 @@ const (
 	Docker_InspectServiceContainer_FullMethodName = "/api.Docker/InspectServiceContainer"
 	Docker_ListServiceContainers_FullMethodName   = "/api.Docker/ListServiceContainers"
 	Docker_RemoveServiceContainer_FullMethodName  = "/api.Docker/RemoveServiceContainer"
-	Docker_ExecContainer_FullMethodName           = "/api.Docker/ExecContainer"
 )
 
 // DockerClient is the client API for Docker service.
@@ -50,6 +51,8 @@ type DockerClient interface {
 	StopContainer(ctx context.Context, in *StopContainerRequest, opts ...grpc.CallOption) (*emptypb.Empty, error)
 	ListContainers(ctx context.Context, in *ListContainersRequest, opts ...grpc.CallOption) (*ListContainersResponse, error)
 	RemoveContainer(ctx context.Context, in *RemoveContainerRequest, opts ...grpc.CallOption) (*emptypb.Empty, error)
+	ExecContainer(ctx context.Context, opts ...grpc.CallOption) (grpc.BidiStreamingClient[ExecContainerRequest, ExecContainerResponse], error)
+	ContainerLogs(ctx context.Context, in *ContainerLogsRequest, opts ...grpc.CallOption) (grpc.ServerStreamingClient[ContainerLogEntry], error)
 	PullImage(ctx context.Context, in *PullImageRequest, opts ...grpc.CallOption) (grpc.ServerStreamingClient[JSONMessage], error)
 	InspectImage(ctx context.Context, in *InspectImageRequest, opts ...grpc.CallOption) (*InspectImageResponse, error)
 	// InspectRemoteImage returns the image metadata for an image in a remote registry using the machine's
@@ -63,7 +66,6 @@ type DockerClient interface {
 	InspectServiceContainer(ctx context.Context, in *InspectContainerRequest, opts ...grpc.CallOption) (*ServiceContainer, error)
 	ListServiceContainers(ctx context.Context, in *ListServiceContainersRequest, opts ...grpc.CallOption) (*ListServiceContainersResponse, error)
 	RemoveServiceContainer(ctx context.Context, in *RemoveContainerRequest, opts ...grpc.CallOption) (*emptypb.Empty, error)
-	ExecContainer(ctx context.Context, opts ...grpc.CallOption) (grpc.BidiStreamingClient[ExecContainerRequest, ExecContainerResponse], error)
 }
 
 type dockerClient struct {
@@ -134,9 +136,41 @@ func (c *dockerClient) RemoveContainer(ctx context.Context, in *RemoveContainerR
 	return out, nil
 }
 
+func (c *dockerClient) ExecContainer(ctx context.Context, opts ...grpc.CallOption) (grpc.BidiStreamingClient[ExecContainerRequest, ExecContainerResponse], error) {
+	cOpts := append([]grpc.CallOption{grpc.StaticMethod()}, opts...)
+	stream, err := c.cc.NewStream(ctx, &Docker_ServiceDesc.Streams[0], Docker_ExecContainer_FullMethodName, cOpts...)
+	if err != nil {
+		return nil, err
+	}
+	x := &grpc.GenericClientStream[ExecContainerRequest, ExecContainerResponse]{ClientStream: stream}
+	return x, nil
+}
+
+// This type alias is provided for backwards compatibility with existing code that references the prior non-generic stream type by name.
+type Docker_ExecContainerClient = grpc.BidiStreamingClient[ExecContainerRequest, ExecContainerResponse]
+
+func (c *dockerClient) ContainerLogs(ctx context.Context, in *ContainerLogsRequest, opts ...grpc.CallOption) (grpc.ServerStreamingClient[ContainerLogEntry], error) {
+	cOpts := append([]grpc.CallOption{grpc.StaticMethod()}, opts...)
+	stream, err := c.cc.NewStream(ctx, &Docker_ServiceDesc.Streams[1], Docker_ContainerLogs_FullMethodName, cOpts...)
+	if err != nil {
+		return nil, err
+	}
+	x := &grpc.GenericClientStream[ContainerLogsRequest, ContainerLogEntry]{ClientStream: stream}
+	if err := x.ClientStream.SendMsg(in); err != nil {
+		return nil, err
+	}
+	if err := x.ClientStream.CloseSend(); err != nil {
+		return nil, err
+	}
+	return x, nil
+}
+
+// This type alias is provided for backwards compatibility with existing code that references the prior non-generic stream type by name.
+type Docker_ContainerLogsClient = grpc.ServerStreamingClient[ContainerLogEntry]
+
 func (c *dockerClient) PullImage(ctx context.Context, in *PullImageRequest, opts ...grpc.CallOption) (grpc.ServerStreamingClient[JSONMessage], error) {
 	cOpts := append([]grpc.CallOption{grpc.StaticMethod()}, opts...)
-	stream, err := c.cc.NewStream(ctx, &Docker_ServiceDesc.Streams[0], Docker_PullImage_FullMethodName, cOpts...)
+	stream, err := c.cc.NewStream(ctx, &Docker_ServiceDesc.Streams[2], Docker_PullImage_FullMethodName, cOpts...)
 	if err != nil {
 		return nil, err
 	}
@@ -253,19 +287,6 @@ func (c *dockerClient) RemoveServiceContainer(ctx context.Context, in *RemoveCon
 	return out, nil
 }
 
-func (c *dockerClient) ExecContainer(ctx context.Context, opts ...grpc.CallOption) (grpc.BidiStreamingClient[ExecContainerRequest, ExecContainerResponse], error) {
-	cOpts := append([]grpc.CallOption{grpc.StaticMethod()}, opts...)
-	stream, err := c.cc.NewStream(ctx, &Docker_ServiceDesc.Streams[1], Docker_ExecContainer_FullMethodName, cOpts...)
-	if err != nil {
-		return nil, err
-	}
-	x := &grpc.GenericClientStream[ExecContainerRequest, ExecContainerResponse]{ClientStream: stream}
-	return x, nil
-}
-
-// This type alias is provided for backwards compatibility with existing code that references the prior non-generic stream type by name.
-type Docker_ExecContainerClient = grpc.BidiStreamingClient[ExecContainerRequest, ExecContainerResponse]
-
 // DockerServer is the server API for Docker service.
 // All implementations must embed UnimplementedDockerServer
 // for forward compatibility.
@@ -276,6 +297,8 @@ type DockerServer interface {
 	StopContainer(context.Context, *StopContainerRequest) (*emptypb.Empty, error)
 	ListContainers(context.Context, *ListContainersRequest) (*ListContainersResponse, error)
 	RemoveContainer(context.Context, *RemoveContainerRequest) (*emptypb.Empty, error)
+	ExecContainer(grpc.BidiStreamingServer[ExecContainerRequest, ExecContainerResponse]) error
+	ContainerLogs(*ContainerLogsRequest, grpc.ServerStreamingServer[ContainerLogEntry]) error
 	PullImage(*PullImageRequest, grpc.ServerStreamingServer[JSONMessage]) error
 	InspectImage(context.Context, *InspectImageRequest) (*InspectImageResponse, error)
 	// InspectRemoteImage returns the image metadata for an image in a remote registry using the machine's
@@ -289,7 +312,6 @@ type DockerServer interface {
 	InspectServiceContainer(context.Context, *InspectContainerRequest) (*ServiceContainer, error)
 	ListServiceContainers(context.Context, *ListServiceContainersRequest) (*ListServiceContainersResponse, error)
 	RemoveServiceContainer(context.Context, *RemoveContainerRequest) (*emptypb.Empty, error)
-	ExecContainer(grpc.BidiStreamingServer[ExecContainerRequest, ExecContainerResponse]) error
 	mustEmbedUnimplementedDockerServer()
 }
 
@@ -318,6 +340,12 @@ func (UnimplementedDockerServer) ListContainers(context.Context, *ListContainers
 func (UnimplementedDockerServer) RemoveContainer(context.Context, *RemoveContainerRequest) (*emptypb.Empty, error) {
 	return nil, status.Errorf(codes.Unimplemented, "method RemoveContainer not implemented")
 }
+func (UnimplementedDockerServer) ExecContainer(grpc.BidiStreamingServer[ExecContainerRequest, ExecContainerResponse]) error {
+	return status.Errorf(codes.Unimplemented, "method ExecContainer not implemented")
+}
+func (UnimplementedDockerServer) ContainerLogs(*ContainerLogsRequest, grpc.ServerStreamingServer[ContainerLogEntry]) error {
+	return status.Errorf(codes.Unimplemented, "method ContainerLogs not implemented")
+}
 func (UnimplementedDockerServer) PullImage(*PullImageRequest, grpc.ServerStreamingServer[JSONMessage]) error {
 	return status.Errorf(codes.Unimplemented, "method PullImage not implemented")
 }
@@ -351,9 +379,6 @@ func (UnimplementedDockerServer) ListServiceContainers(context.Context, *ListSer
 func (UnimplementedDockerServer) RemoveServiceContainer(context.Context, *RemoveContainerRequest) (*emptypb.Empty, error) {
 	return nil, status.Errorf(codes.Unimplemented, "method RemoveServiceContainer not implemented")
 }
-func (UnimplementedDockerServer) ExecContainer(grpc.BidiStreamingServer[ExecContainerRequest, ExecContainerResponse]) error {
-	return status.Errorf(codes.Unimplemented, "method ExecContainer not implemented")
-}
 func (UnimplementedDockerServer) mustEmbedUnimplementedDockerServer() {}
 func (UnimplementedDockerServer) testEmbeddedByValue()                {}
 
@@ -483,6 +508,24 @@ func _Docker_RemoveContainer_Handler(srv interface{}, ctx context.Context, dec f
 	return interceptor(ctx, in, info, handler)
 }
 
+func _Docker_ExecContainer_Handler(srv interface{}, stream grpc.ServerStream) error {
+	return srv.(DockerServer).ExecContainer(&grpc.GenericServerStream[ExecContainerRequest, ExecContainerResponse]{ServerStream: stream})
+}
+
+// This type alias is provided for backwards compatibility with existing code that references the prior non-generic stream type by name.
+type Docker_ExecContainerServer = grpc.BidiStreamingServer[ExecContainerRequest, ExecContainerResponse]
+
+func _Docker_ContainerLogs_Handler(srv interface{}, stream grpc.ServerStream) error {
+	m := new(ContainerLogsRequest)
+	if err := stream.RecvMsg(m); err != nil {
+		return err
+	}
+	return srv.(DockerServer).ContainerLogs(m, &grpc.GenericServerStream[ContainerLogsRequest, ContainerLogEntry]{ServerStream: stream})
+}
+
+// This type alias is provided for backwards compatibility with existing code that references the prior non-generic stream type by name.
+type Docker_ContainerLogsServer = grpc.ServerStreamingServer[ContainerLogEntry]
+
 func _Docker_PullImage_Handler(srv interface{}, stream grpc.ServerStream) error {
 	m := new(PullImageRequest)
 	if err := stream.RecvMsg(m); err != nil {
@@ -674,13 +717,6 @@ func _Docker_RemoveServiceContainer_Handler(srv interface{}, ctx context.Context
 	return interceptor(ctx, in, info, handler)
 }
 
-func _Docker_ExecContainer_Handler(srv interface{}, stream grpc.ServerStream) error {
-	return srv.(DockerServer).ExecContainer(&grpc.GenericServerStream[ExecContainerRequest, ExecContainerResponse]{ServerStream: stream})
-}
-
-// This type alias is provided for backwards compatibility with existing code that references the prior non-generic stream type by name.
-type Docker_ExecContainerServer = grpc.BidiStreamingServer[ExecContainerRequest, ExecContainerResponse]
-
 // Docker_ServiceDesc is the grpc.ServiceDesc for Docker service.
 // It's only intended for direct use with grpc.RegisterService,
 // and not to be introspected or modified (even as a copy)
@@ -754,17 +790,22 @@ var Docker_ServiceDesc = grpc.ServiceDesc{
 		},
 	},
 	Streams: []grpc.StreamDesc{
-		{
-			StreamName:    "PullImage",
-			Handler:       _Docker_PullImage_Handler,
-			ServerStreams: true,
-		},
 		{
 			StreamName:    "ExecContainer",
 			Handler:       _Docker_ExecContainer_Handler,
 			ServerStreams: true,
 			ClientStreams: true,
 		},
+		{
+			StreamName:    "ContainerLogs",
+			Handler:       _Docker_ContainerLogs_Handler,
+			ServerStreams: true,
+		},
+		{
+			StreamName:    "PullImage",
+			Handler:       _Docker_PullImage_Handler,
+			ServerStreams: true,
+		},
 	},
 	Metadata: "internal/machine/api/pb/docker.proto",
 }

internal/machine/docker/client_exec.go

@@ -201,10 +201,10 @@ func handleClientOutputStream(ctx context.Context, stream pb.Docker_ExecContaine
 // ExecContainer executes a command in a running container with bidirectional streaming.
 // TODO: This can be merged with pkg/client as it's an unnecessary logic split.
 func (c *Client) ExecContainer(ctx context.Context, opts ExecConfig) (exitCode int, err error) {
-	// TODO: We need to handle Ctrl-C and other signals here to forward them to the container process.
-	// Right now, Ctrl-C will just terminate the client process, which is not ideal.
-	// We should catch the signal, send it to the container process, and only exit
-	// when the container process exits.
+	// Note: In non-interactive mode (without TTY), signals like SIGTERM/SIGINT will terminate the client
+	// process without being forwarded to the remote container process. We could catch and forward these
+	// signals (useful for long-running commands), but for example "kubectl exec" doesn't do this either (as of November 2025),
+	// so we keep it simple for now.
 
 	slog.Debug("starting ExecContainer", "containerID", opts.ContainerID, "options", opts.Options)
 

internal/machine/docker/server.go

@@ -47,6 +47,7 @@ import (
 	"google.golang.org/grpc/codes"
 	"google.golang.org/grpc/status"
 	"google.golang.org/protobuf/types/known/emptypb"
+	"google.golang.org/protobuf/types/known/timestamppb"
 )
 
 var fullDockerIDRegex = regexp.MustCompile(`^[a-f0-9]{64}$`)
@@ -1015,6 +1016,97 @@ func (s *Server) RemoveServiceContainer(ctx context.Context, req *pb.RemoveConta
 	return resp, nil
 }
 
+// logsHeartbeatInterval is the interval at which heartbeat entries are sent when there are no logs to stream.
+const logsHeartbeatInterval = 200 * time.Millisecond
+
+// ContainerLogs streams logs from a container.
+func (s *Server) ContainerLogs(
+	req *pb.ContainerLogsRequest, stream grpc.ServerStreamingServer[pb.ContainerLogEntry],
+) error {
+	// Stream context is cancelled when the client has disconnected or the stream has ended.
+	ctx := stream.Context()
+
+	opts := ContainerLogsOptions{
+		ContainerID: req.ContainerId,
+		Follow:      req.Follow,
+		Tail:        int(req.Tail),
+		Since:       req.Since,
+		Until:       req.Until,
+	}
+
+	logsCh, err := s.service.ContainerLogs(ctx, opts)
+	if err != nil {
+		if errdefs.IsNotFound(err) {
+			return status.Error(codes.NotFound, err.Error())
+		}
+		return status.Errorf(codes.Internal, "get container logs: %v", err)
+	}
+
+	log := slog.With("container_id", req.ContainerId, "stream_id", fmt.Sprintf("%p", stream)[2:])
+	log.Debug("Starting container logs streaming.",
+		"follow", req.Follow, "tail", req.Tail, "since", req.Since, "until", req.Until)
+
+	// Heartbeats are needed only when following logs to let the client know when there are no new log entries
+	// to allow it to advance the watermark of last received log timestamp.
+	var heartbeatCh <-chan time.Time
+	if req.Follow {
+		heartbeatTicker := time.NewTicker(logsHeartbeatInterval)
+		defer heartbeatTicker.Stop()
+		heartbeatCh = heartbeatTicker.C
+	}
+
+	started := time.Now()
+	lastSent := time.Time{}
+
+	for {
+		select {
+		case entry, ok := <-logsCh:
+			if !ok {
+				// Channel closed, no more log entries.
+				return nil
+			}
+
+			if entry.Err != nil {
+				return status.Error(codes.Internal, entry.Err.Error())
+			}
+
+			pbEntry := &pb.ContainerLogEntry{
+				Stream:    api.LogStreamTypeToProto(entry.Stream),
+				Timestamp: timestamppb.New(entry.Timestamp),
+				Message:   entry.Message,
+			}
+			if err = stream.Send(pbEntry); err != nil {
+				return status.Errorf(codes.Internal, "send log entry: %v", err)
+			}
+			lastSent = entry.Timestamp
+
+		case now := <-heartbeatCh:
+			// Only send heartbeat if no log entries have been sent since the last heartbeat interval or
+			// if no log entries have been sent at all for at least a heartbeat interval since starting.
+			if now.Sub(lastSent) < logsHeartbeatInterval ||
+				(lastSent.IsZero() && now.Sub(started) < logsHeartbeatInterval) {
+				continue
+			}
+
+			// Use the timestamp one heartbeat in the past to be conservative. This reduces the chance of sending
+			// a timestamp that is greater than a log entry currently being parsed but not yet sent, which would
+			// cause the client to incorrectly believe it has received all logs up to that point.
+			heartbeat := &pb.ContainerLogEntry{
+				Stream:    pb.ContainerLogEntry_HEARTBEAT,
+				Timestamp: timestamppb.New(now.Add(-logsHeartbeatInterval)),
+			}
+			if err = stream.Send(heartbeat); err != nil {
+				return status.Errorf(codes.Internal, "send log stream heartbeat: %v", err)
+			}
+			lastSent = heartbeat.Timestamp.AsTime()
+			log.Debug("Sent log stream heartbeat.", "timestamp", lastSent)
+
+		case <-ctx.Done():
+			return status.Error(codes.Canceled, ctx.Err().Error())
+		}
+	}
+}
+
 // receiveExecConfig receives and validates the initial exec configuration from the stream.
 func (s *Server) receiveExecConfig(stream pb.Docker_ExecContainerServer) (*pb.ExecConfig, api.ExecOptions, error) {
 	req, err := stream.Recv()

internal/machine/docker/service.go

@@ -1,18 +1,22 @@
 package docker
 
 import (
+	"bytes"
 	"context"
 	"database/sql"
 	"encoding/json"
 	"errors"
 	"fmt"
 	"log/slog"
+	"strconv"
 	"strings"
+	"time"
 
 	"github.com/docker/docker/api/types/container"
 	"github.com/docker/docker/api/types/filters"
 	"github.com/docker/docker/api/types/image"
 	"github.com/docker/docker/client"
+	"github.com/docker/docker/pkg/stdcopy"
 	"github.com/jmoiron/sqlx"
 	"github.com/psviderski/uncloud/pkg/api"
 	"google.golang.org/grpc/codes"
@@ -150,3 +154,104 @@ func (s *Service) ListImages(ctx context.Context, opts image.ListOptions) (Image
 
 	return imagesResp, nil
 }
+
+// ContainerLogsOptions specifies parameters for ContainerLogs.
+type ContainerLogsOptions struct {
+	ContainerID string
+	Follow      bool
+	Tail        int
+	Since       string
+	Until       string
+}
+
+// ContainerLogs streams logs from a container and returns demultiplexed entries via a channel.
+// The channel is closed when streaming completes or context is cancelled.
+func (s *Service) ContainerLogs(ctx context.Context, opts ContainerLogsOptions) (<-chan api.ContainerLogEntry, error) {
+	dockerOpts := container.LogsOptions{
+		ShowStdout: true,
+		ShowStderr: true,
+		Follow:     opts.Follow,
+		Tail:       strconv.FormatInt(int64(opts.Tail), 10),
+		Since:      opts.Since,
+		Until:      opts.Until,
+		Timestamps: true,
+	}
+
+	reader, err := s.Client.ContainerLogs(ctx, opts.ContainerID, dockerOpts)
+	if err != nil {
+		return nil, err
+	}
+
+	outCh := make(chan api.ContainerLogEntry)
+	stdoutWriter := &logsChannelWriter{ctx: ctx, ch: outCh, isStderr: false}
+	stderrWriter := &logsChannelWriter{ctx: ctx, ch: outCh, isStderr: true}
+
+	// Wrap the context in a cancellable one to unblock the second goroutine below when StdCopy completes.
+	ctx, cancel := context.WithCancel(ctx)
+
+	// Run StdCopy in a goroutine to be able to handle context cancellation.
+	go func() {
+		defer close(outCh)
+		defer cancel()
+
+		// StdCopy is blocking and will return when the reader is closed in another goroutine below or on error.
+		if _, err := stdcopy.StdCopy(stdoutWriter, stderrWriter, reader); err != nil {
+			// Send error as the last entry.
+			select {
+			case outCh <- api.ContainerLogEntry{Err: fmt.Errorf("demultiplex container logs: %w", err)}:
+			case <-ctx.Done():
+			}
+		}
+	}()
+
+	// Close the reader when the context is done to cancel StdCopy if it's still running.
+	go func() {
+		<-ctx.Done()
+		reader.Close()
+	}()
+
+	return outCh, nil
+}
+
+// logsChannelWriter is a writer for stdcopy.StdCopy that sends demultiplexed container logs to a channel.
+type logsChannelWriter struct {
+	ctx      context.Context
+	ch       chan<- api.ContainerLogEntry
+	isStderr bool
+}
+
+func (w *logsChannelWriter) Write(data []byte) (n int, err error) {
+	// Parse timestamp and message from the demultiplexed Docker log payload if the data looks like it contains one.
+	// Format: 2025-01-01T00:00:00.000000000Z message
+	timestamp := time.Time{}
+	message := data
+	if len(data) > 30 && data[4] == '-' && data[7] == '-' && data[10] == 'T' {
+		timestampPart, messagePart, found := bytes.Cut(data, []byte(" "))
+		if found {
+			timestamp, err = time.Parse(time.RFC3339Nano, string(timestampPart))
+			if err != nil {
+				timestamp = time.Time{}
+			}
+			message = messagePart
+		}
+	}
+
+	entry := api.ContainerLogEntry{
+		Timestamp: timestamp,
+		// Clone is required because message is a slice into data, which stdcopy.StdCopy may reuse
+		// after Write returns but before the entry is consumed from the channel.
+		Message: bytes.Clone(message),
+	}
+	if w.isStderr {
+		entry.Stream = api.LogStreamStderr
+	} else {
+		entry.Stream = api.LogStreamStdout
+	}
+
+	select {
+	case w.ch <- entry:
+		return len(data), nil
+	case <-w.ctx.Done():
+		return 0, w.ctx.Err()
+	}
+}

internal/sshexec/sshcli.go

@@ -0,0 +1,72 @@
+package sshexec
+
+import (
+	"context"
+	"fmt"
+	"io"
+	"os/exec"
+	"strconv"
+	"strings"
+)
+
+type SSHCLIRemote struct {
+	user    string
+	host    string
+	port    int
+	keyPath string
+}
+
+func NewSSHCLIRemote(user, host string, port int, keyPath string) *SSHCLIRemote {
+	return &SSHCLIRemote{
+		user:    user,
+		host:    host,
+		port:    port,
+		keyPath: keyPath,
+	}
+}
+
+// TODO: Refactor and reuse this with buildDialArgs and buildSSHArgs from
+// SSHCLI Connector.
+func (r *SSHCLIRemote) buildSSHArgs() []string {
+	args := []string{"-o", "ConnectTimeout=5"}
+
+	if r.port != 0 && r.port != 22 {
+		args = append(args, "-p", strconv.Itoa(r.port))
+	}
+
+	if r.keyPath != "" {
+		args = append(args, "-i", r.keyPath)
+	}
+
+	args = append(args, r.user+"@"+r.host)
+	return args
+}
+
+func (r *SSHCLIRemote) Run(ctx context.Context, cmd string) (string, error) {
+	args := r.buildSSHArgs()
+	args = append(args, cmd)
+
+	execCmd := exec.CommandContext(ctx, "ssh", args...)
+	output, err := execCmd.CombinedOutput()
+	if err != nil {
+		return strings.TrimSpace(string(output)),
+			fmt.Errorf("run command on remote host: %w: %s", err, string(output))
+	}
+	return strings.TrimSpace(string(output)), nil
+}
+
+func (r *SSHCLIRemote) Stream(ctx context.Context, cmd string, stdout, stderr io.Writer) error {
+	args := r.buildSSHArgs()
+	args = append(args, cmd)
+
+	execCmd := exec.CommandContext(ctx, "ssh", args...)
+	execCmd.Stdout = stdout
+	execCmd.Stderr = stderr
+
+	return execCmd.Run()
+}
+
+// no-op as there is no persistent connection.
+func (r *SSHCLIRemote) Close() error {
+	return nil
+}

internal/sshexec/sshcli_test.go

@@ -0,0 +1,76 @@
+package sshexec
+
+import (
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+)
+
+func TestSSHCLIRemote_buildSSHArgs(t *testing.T) {
+	t.Parallel()
+
+	tests := []struct {
+		name     string
+		user     string
+		host     string
+		port     int
+		keyPath  string
+		expected []string
+	}{
+		{
+			name:     "default port, no key",
+			user:     "root",
+			host:     "example.com",
+			port:     22,
+			keyPath:  "",
+			expected: []string{"-o", "ConnectTimeout=5", "root@example.com"},
+		},
+		{
+			name:     "custom port",
+			user:     "ubuntu",
+			host:     "192.168.1.10",
+			port:     2222,
+			keyPath:  "",
+			expected: []string{"-o", "ConnectTimeout=5", "-p", "2222", "ubuntu@192.168.1.10"},
+		},
+		{
+			name:     "with key path",
+			user:     "root",
+			host:     "example.com",
+			port:     22,
+			keyPath:  "/path/to/key",
+			expected: []string{"-o", "ConnectTimeout=5", "-i", "/path/to/key", "root@example.com"},
+		},
+		{
+			name:     "all options",
+			user:     "admin",
+			host:     "server.local",
+			port:     2222,
+			keyPath:  "~/.ssh/id_rsa",
+			expected: []string{"-o", "ConnectTimeout=5", "-p", "2222", "-i", "~/.ssh/id_rsa", "admin@server.local"},
+		},
+		{
+			name:     "port zero (default)",
+			user:     "root",
+			host:     "example.com",
+			port:     0,
+			keyPath:  "",
+			expected: []string{"-o", "ConnectTimeout=5", "root@example.com"},
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			t.Parallel()
+
+			remote := &SSHCLIRemote{
+				user:    tt.user,
+				host:    tt.host,
+				port:    tt.port,
+				keyPath: tt.keyPath,
+			}
+			result := remote.buildSSHArgs()
+			assert.Equal(t, tt.expected, result)
+		})
+	}
+}

misc/user_guide.md

@@ -22,7 +22,7 @@ Uncloud stores its configuration in `~/.config/uncloud/config.yaml`. If you wish
 
 Begin by initialising the first node in your cluster with `uc machine init [USER@HOST:PORT]`. If you do not have a need for Caddy reverse proxy, you may disable this feature with `--no-caddy`. If you want to avoid using uncloud's managed DNS service, add the `--no-dns` flag.
 
-This command will idempotently install Docker, uncloudd, uncloud-corrosion. If Caddy is enabled, it will set up a reverse proxy. If Uncloud DNS is enabled, it will create a DNS A record for the machine's public IP address under `*.[CLUSTER ID].cluster.uncloud.run`.
+This command will idempotently install Docker, uncloudd, uncloud-corrosion. If Caddy is enabled, it will set up a reverse proxy. If Uncloud DNS is enabled, it will create a DNS A record for the machine's public IP address under `*.[CLUSTER ID].uncld.dev`.
 
 If you wish to uninstall Uncloud and its components, run `uncloud-uninstall`.
 

pkg/api/client.go

@@ -2,13 +2,10 @@ package api
 
 import (
 	"context"
-	"fmt"
-	"strings"
 
 	"github.com/docker/docker/api/types/container"
 	"github.com/docker/docker/api/types/volume"
 	"github.com/psviderski/uncloud/internal/machine/api/pb"
-	"google.golang.org/grpc/metadata"
 )
 
 type Client interface {
@@ -51,6 +48,8 @@ type ServiceClient interface {
 	RunService(ctx context.Context, spec ServiceSpec) (RunServiceResponse, error)
 	InspectService(ctx context.Context, id string) (Service, error)
 	RemoveService(ctx context.Context, id string) error
+	StopService(ctx context.Context, id string, opts container.StopOptions) error
+	StartService(ctx context.Context, id string) error
 }
 
 type VolumeClient interface {
@@ -58,41 +57,3 @@ type VolumeClient interface {
 	ListVolumes(ctx context.Context, filter *VolumeFilter) ([]MachineVolume, error)
 	RemoveVolume(ctx context.Context, machineNameOrID, volumeName string, force bool) error
 }
-
-// ProxyMachinesContext returns a new context that proxies gRPC requests to the specified machines.
-// If namesOrIDs is nil, all machines are included.
-func ProxyMachinesContext(
-	ctx context.Context, cli MachineClient, namesOrIDs []string,
-) (context.Context, MachineMembersList, error) {
-	// TODO: move the machine IP resolution to the proxy router to allow setting machine names and IDs in the metadata.
-	machines, err := cli.ListMachines(ctx, nil)
-	if err != nil {
-		return nil, nil, fmt.Errorf("list machines: %w", err)
-	}
-
-	var proxiedMachines MachineMembersList
-	var notFound []string
-	for _, nameOrID := range namesOrIDs {
-		if m := machines.FindByNameOrID(nameOrID); m != nil {
-			proxiedMachines = append(proxiedMachines, m)
-		} else {
-			notFound = append(notFound, nameOrID)
-		}
-	}
-
-	if len(notFound) > 0 {
-		return nil, nil, fmt.Errorf("machines not found: %s", strings.Join(notFound, ", "))
-	}
-
-	if len(namesOrIDs) == 0 {
-		proxiedMachines = machines
-	}
-
-	md := metadata.New(nil)
-	for _, m := range proxiedMachines {
-		machineIP, _ := m.Machine.Network.ManagementIp.ToAddr()
-		md.Append("machines", machineIP.String())
-	}
-
-	return metadata.NewOutgoingContext(ctx, md), proxiedMachines, nil
-}

pkg/api/logs.go

@@ -0,0 +1,87 @@
+package api
+
+import (
+	"errors"
+	"time"
+
+	"github.com/psviderski/uncloud/internal/machine/api/pb"
+)
+
+const (
+	LogStreamUnknown LogStreamType = iota
+	LogStreamStdout
+	LogStreamStderr
+	// LogStreamHeartbeat represents a heartbeat log entry with a timestamp indicating that
+	// there are no older logs than this timestamp.
+	LogStreamHeartbeat
+)
+
+type LogStreamType int
+
+// LogStreamTypeFromProto converts a protobuf ContainerLogEntry.StreamType to the internal LogStreamType.
+func LogStreamTypeFromProto(s pb.ContainerLogEntry_StreamType) LogStreamType {
+	switch s {
+	case pb.ContainerLogEntry_STDOUT:
+		return LogStreamStdout
+	case pb.ContainerLogEntry_STDERR:
+		return LogStreamStderr
+	case pb.ContainerLogEntry_HEARTBEAT:
+		return LogStreamHeartbeat
+	default:
+		return LogStreamUnknown
+	}
+}
+
+// LogStreamTypeToProto converts LogStreamType to protobuf ContainerLogEntry.StreamType.
+func LogStreamTypeToProto(s LogStreamType) pb.ContainerLogEntry_StreamType {
+	switch s {
+	case LogStreamStdout:
+		return pb.ContainerLogEntry_STDOUT
+	case LogStreamStderr:
+		return pb.ContainerLogEntry_STDERR
+	case LogStreamHeartbeat:
+		return pb.ContainerLogEntry_HEARTBEAT
+	default:
+		return pb.ContainerLogEntry_UNKNOWN
+	}
+}
+
+// ServiceLogsOptions specifies parameters for ServiceLogs.
+type ServiceLogsOptions struct {
+	Follow bool
+	Tail   int
+	Since  string
+	Until  string
+	// Machines filters logs to only include containers running on the specified machines (names or IDs).
+	// If empty, logs from all machines are included.
+	Machines []string
+}
+
+// ServiceLogEntry represents a single log entry from a service container.
+type ServiceLogEntry struct {
+	// Metadata may not be set if an error occurred (Err is not nil).
+	Metadata ServiceLogEntryMetadata
+	ContainerLogEntry
+}
+
+// ServiceLogEntryMetadata contains metadata about the source of a log entry.
+type ServiceLogEntryMetadata struct {
+	ServiceID   string
+	ServiceName string
+	ContainerID string
+	MachineID   string
+	MachineName string
+}
+
+// ContainerLogEntry represents a single log entry from a container.
+type ContainerLogEntry struct {
+	Stream    LogStreamType
+	Timestamp time.Time
+	Message   []byte
+	// Err indicates that an error occurred while streaming logs from a container.
+	// Other fields are not set if Err is not nil.
+	Err error
+}
+
+// ErrLogStreamStalled indicates that a log stream stopped sending data and may be unresponsive.
+var ErrLogStreamStalled = errors.New("log stream stopped responding")

pkg/api/service.go

@@ -8,6 +8,7 @@ import (
 	"slices"
 	"strings"
 
+	mapset "github.com/deckarep/golang-set/v2"
 	"github.com/distribution/reference"
 	"github.com/google/go-cmp/cmp"
 	"github.com/google/go-cmp/cmp/cmpopts"
@@ -366,6 +367,16 @@ type MachineServiceContainer struct {
 	Container ServiceContainer
 }
 
+// MachineIDs returns a list of unique machine IDs where the service containers are running.
+func (s *Service) MachineIDs() []string {
+	ids := mapset.NewSet[string]()
+	for _, mc := range s.Containers {
+		ids.Add(mc.MachineID)
+	}
+
+	return ids.ToSlice()
+}
+
 // Images returns a sorted list of unique images used by the service containers.
 func (s *Service) Images() []string {
 	images := make(map[string]struct{})

pkg/api/volume.go

@@ -83,6 +83,9 @@ func (v *VolumeSpec) SetDefaults() VolumeSpec {
 		if spec.VolumeOptions.Name == "" {
 			spec.VolumeOptions.Name = spec.Name
 		}
+		if spec.VolumeOptions.Driver != nil && spec.VolumeOptions.Driver.Name == "" {
+			spec.VolumeOptions.Driver.Name = VolumeDriverLocal
+		}
 	}
 	// TODO: set explicit default values for Propagation and Recursive for bind mounts?
 

pkg/api/volume_test.go

@@ -0,0 +1,142 @@
+package api
+
+import (
+	"testing"
+
+	"github.com/docker/docker/api/types/mount"
+	"github.com/docker/docker/api/types/volume"
+	"github.com/stretchr/testify/assert"
+)
+
+func TestVolumeSpec_MatchesDockerVolume(t *testing.T) {
+	t.Parallel()
+
+	tests := []struct {
+		name     string
+		spec     VolumeSpec
+		vol      volume.Volume
+		expected bool
+	}{
+		{
+			name: "match with explicit local driver",
+			spec: VolumeSpec{
+				Type: VolumeTypeVolume,
+				VolumeOptions: &VolumeOptions{
+					Driver: &mount.Driver{
+						Name: "local",
+						Options: map[string]string{
+							"foo": "bar",
+						},
+					},
+				},
+			},
+			vol: volume.Volume{
+				Driver: "local",
+				Options: map[string]string{
+					"foo": "bar",
+				},
+			},
+			expected: true,
+		},
+		{
+			name: "match with empty driver name in spec (implicit local)",
+			spec: VolumeSpec{
+				Type: VolumeTypeVolume,
+				VolumeOptions: &VolumeOptions{
+					Driver: &mount.Driver{
+						Name: "", // Implicitly local
+						Options: map[string]string{
+							"foo": "bar",
+						},
+					},
+				},
+			},
+			vol: volume.Volume{
+				Driver: "local",
+				Options: map[string]string{
+					"foo": "bar",
+				},
+			},
+			expected: true,
+		},
+		{
+			name: "non-match with different driver options",
+			spec: VolumeSpec{
+				Type: VolumeTypeVolume,
+				VolumeOptions: &VolumeOptions{
+					Driver: &mount.Driver{
+						Name: "local",
+						Options: map[string]string{
+							"foo": "baz",
+						},
+					},
+				},
+			},
+			vol: volume.Volume{
+				Driver: "local",
+				Options: map[string]string{
+					"foo": "bar",
+				},
+			},
+			expected: false,
+		},
+		{
+			name: "non-match with different driver name",
+			spec: VolumeSpec{
+				Type: VolumeTypeVolume,
+				VolumeOptions: &VolumeOptions{
+					Driver: &mount.Driver{
+						Name: "custom",
+						Options: map[string]string{
+							"foo": "bar",
+						},
+					},
+				},
+			},
+			vol: volume.Volume{
+				Driver: "local",
+				Options: map[string]string{
+					"foo": "bar",
+				},
+			},
+			expected: false,
+		},
+		{
+			name: "match external volume without driver by name only",
+			spec: VolumeSpec{
+				Name: "external",
+				Type: VolumeTypeVolume,
+			},
+			vol: volume.Volume{
+				Name:   "external",
+				Driver: "local",
+				Options: map[string]string{
+					"foo": "bar",
+				},
+			},
+			expected: true,
+		},
+		{
+			name: "non-match external volume by name",
+			spec: VolumeSpec{
+				Name: "unknown",
+				Type: VolumeTypeVolume,
+			},
+			vol: volume.Volume{
+				Name:   "external",
+				Driver: "local",
+				Options: map[string]string{
+					"foo": "bar",
+				},
+			},
+			expected: false,
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			matches := tt.spec.MatchesDockerVolume(tt.vol)
+			assert.Equal(t, tt.expected, matches)
+		})
+	}
+}

pkg/client/caddy.go

@@ -71,6 +71,13 @@ func (cli *Client) NewCaddyDeployment(image, config string, placement api.Placem
 				Protocol:      api.ProtocolTCP,
 				Mode:          api.PortModeHost,
 			},
+			// Needed for HTTP/3 (QUIC)
+			{
+				PublishedPort: 443,
+				ContainerPort: 443,
+				Protocol:      api.ProtocolUDP,
+				Mode:          api.PortModeHost,
+			},
 		},
 		Volumes: []api.VolumeSpec{
 			{

pkg/client/client.go

@@ -5,6 +5,7 @@ import (
 	"errors"
 	"fmt"
 	"os"
+	"strings"
 
 	"github.com/docker/cli/cli/streams"
 	"github.com/psviderski/uncloud/internal/machine/api/pb"
@@ -76,3 +77,41 @@ func proxyToMachine(ctx context.Context, machine *pb.MachineInfo) context.Contex
 	md := metadata.Pairs("machines", machineIP.String())
 	return metadata.NewOutgoingContext(ctx, md)
 }
+
+// ProxyMachinesContext returns a new context that proxies gRPC requests to the specified machines.
+// If namesOrIDs is nil, all machines are included.
+func (cli *Client) ProxyMachinesContext(
+	ctx context.Context, namesOrIDs []string,
+) (context.Context, api.MachineMembersList, error) {
+	// TODO: move the machine IP resolution to the proxy router to allow setting machine names and IDs in the metadata.
+	machines, err := cli.ListMachines(ctx, nil)
+	if err != nil {
+		return nil, nil, fmt.Errorf("list machines: %w", err)
+	}
+
+	var proxiedMachines api.MachineMembersList
+	var notFound []string
+	for _, nameOrID := range namesOrIDs {
+		if m := machines.FindByNameOrID(nameOrID); m != nil {
+			proxiedMachines = append(proxiedMachines, m)
+		} else {
+			notFound = append(notFound, nameOrID)
+		}
+	}
+
+	if len(notFound) > 0 {
+		return nil, nil, fmt.Errorf("machines not found: %s", strings.Join(notFound, ", "))
+	}
+
+	if len(namesOrIDs) == 0 {
+		proxiedMachines = machines
+	}
+
+	md := metadata.New(nil)
+	for _, m := range proxiedMachines {
+		machineIP, _ := m.Machine.Network.ManagementIp.ToAddr()
+		md.Append("machines", machineIP.String())
+	}
+
+	return metadata.NewOutgoingContext(ctx, md), proxiedMachines, nil
+}

pkg/client/compose/caddy_test.go

@@ -1,8 +1,11 @@
 package compose
 
 import (
+	"context"
+	"os"
 	"testing"
 
+	composecli "github.com/compose-spec/compose-go/v2/cli"
 	"github.com/stretchr/testify/assert"
 	"github.com/stretchr/testify/require"
 )
@@ -178,7 +181,15 @@ services:
 
 	for _, tt := range tests {
 		t.Run(tt.name, func(t *testing.T) {
-			project, err := loadProjectFromContent(t, tt.composeYAML)
+			// Get current working directory for relative path resolution to testdata directory.
+			wd, err := os.Getwd()
+			require.NoError(t, err)
+
+			project, err := LoadProjectFromContent(
+				context.Background(),
+				tt.composeYAML,
+				composecli.WithWorkingDirectory(wd),
+			)
 
 			if tt.wantErr != "" {
 				require.ErrorContains(t, err, tt.wantErr)

pkg/client/compose/deploy.go

@@ -49,10 +49,6 @@ func NewDeploymentWithStrategy(ctx context.Context, cli Client, project *types.P
 		ClusterDomain: domain,
 	}
 
-	if strategy == nil {
-		strategy = &deploy.RollingStrategy{State: state}
-	}
-
 	return &Deployment{
 		Client:       cli,
 		Project:      project,
@@ -88,6 +84,7 @@ func (d *Deployment) Plan(ctx context.Context) (deploy.SequenceOperation, error)
 	}
 
 	// Check external volumes and plan the creation of missing volumes before deploying services.
+	// Updates the cluster state (d.state) with the scheduled volumes.
 	volumeOps, err := d.planVolumes(serviceSpecs)
 	if err != nil {
 		return plan, err
@@ -98,8 +95,8 @@ func (d *Deployment) Plan(ctx context.Context) (deploy.SequenceOperation, error)
 
 	for _, spec := range serviceSpecs {
 		// TODO: properly handle depends_on conditions in the service deployment plan as the first operation.
-		// Pass the update cluster state with scheduled volumes to the deployment.
-		deployment := deploy.NewDeployment(d.Client, spec, d.Strategy)
+		// Pass the updated cluster state with the scheduled volumes to the deployment.
+		deployment := deploy.NewDeploymentWithClusterState(d.Client, spec, d.Strategy, d.state)
 		servicePlan, err := deployment.Plan(ctx)
 		if err != nil {
 			return plan, fmt.Errorf("create deployment plan for service '%s': %w", spec.Name, err)

pkg/client/compose/port_test.go

@@ -1,6 +1,7 @@
 package compose
 
 import (
+	"context"
 	"net/netip"
 	"testing"
 
@@ -301,7 +302,7 @@ services:
 		t.Run(tt.name, func(t *testing.T) {
 			t.Parallel()
 
-			project, err := loadProjectFromContent(t, tt.content)
+			project, err := LoadProjectFromContent(context.Background(), tt.content)
 			if tt.wantErr != "" {
 				require.Error(t, err)
 				assert.Contains(t, err.Error(), tt.wantErr)
@@ -375,7 +376,7 @@ services:
 		t.Run(tt.name, func(t *testing.T) {
 			t.Parallel()
 
-			project, err := loadProjectFromContent(t, tt.content)
+			project, err := LoadProjectFromContent(context.Background(), tt.content)
 			require.NoError(t, err)
 
 			service, err := project.GetService("web")
@@ -402,7 +403,7 @@ services:
       - "9090:9090/tcp@host"
 `
 
-	project, err := loadProjectFromContent(t, content)
+	project, err := LoadProjectFromContent(context.Background(), content)
 	require.NoError(t, err)
 
 	service, err := project.GetService("web")

pkg/client/compose/project.go

@@ -3,12 +3,15 @@ package compose
 import (
 	"context"
 	"fmt"
+	"os"
+	"path/filepath"
 	"strings"
 
 	composecli "github.com/compose-spec/compose-go/v2/cli"
 	"github.com/compose-spec/compose-go/v2/types"
 )
 
+// LoadProject loads a Compose project from the default locations or the given paths.
 func LoadProject(ctx context.Context, paths []string, opts ...composecli.ProjectOptionsFn) (*types.Project, error) {
 	defaultOpts := []composecli.ProjectOptionsFn{
 		// First apply os.Environment, always wins.
@@ -61,6 +64,26 @@ func LoadProject(ctx context.Context, paths []string, opts ...composecli.Project
 	return project, nil
 }
 
+// LoadProjectFromContent loads a Compose project from the given YAML content.
+func LoadProjectFromContent(
+	ctx context.Context, content string, opts ...composecli.ProjectOptionsFn,
+) (*types.Project, error) {
+	// Create a temporary directory for the compose file.
+	tmpDir, err := os.MkdirTemp("", "uncloud-compose-*")
+	if err != nil {
+		return nil, fmt.Errorf("create temporary directory: %w", err)
+	}
+	defer os.RemoveAll(tmpDir)
+
+	// Write the YAML content to compose.yaml in the temporary directory.
+	composePath := filepath.Join(tmpDir, "compose.yaml")
+	if err := os.WriteFile(composePath, []byte(content), 0o644); err != nil {
+		return nil, fmt.Errorf("write compose file: %w", err)
+	}
+
+	return LoadProject(ctx, []string{composePath}, opts...)
+}
+
 // removeProjectPrefixFromNames removes the project name prefix from volume names.
 func removeProjectPrefixFromNames(project *types.Project) {
 	prefix := project.Name + "_"

pkg/client/compose/service.go

@@ -249,7 +249,7 @@ func dockerVolumeSpecFromCompose(serviceVolume types.ServiceVolumeConfig, volume
 	}
 
 	if !volume.External {
-		if volume.Driver != "" {
+		if volume.Driver != "" || len(volume.DriverOpts) > 0 {
 			spec.VolumeOptions.Driver = &mount.Driver{
 				Name:    volume.Driver,
 				Options: volume.DriverOpts,

pkg/client/compose/service_test.go

@@ -3,13 +3,13 @@ package compose
 import (
 	"context"
 	"net/netip"
+	"os"
 	"path/filepath"
 	"slices"
 	"strings"
 	"testing"
 
-	"github.com/compose-spec/compose-go/v2/loader"
-	"github.com/compose-spec/compose-go/v2/types"
+	composecli "github.com/compose-spec/compose-go/v2/cli"
 	"github.com/docker/docker/api/types/container"
 	"github.com/docker/docker/api/types/mount"
 	"github.com/docker/go-units"
@@ -20,57 +20,6 @@ import (
 	"github.com/stretchr/testify/require"
 )
 
-// loadProjectFromContent loads a compose project from YAML content.
-// Keep the implementation in sync with LoadProject.
-func loadProjectFromContent(t *testing.T, content string) (*types.Project, error) {
-	t.Helper()
-	ctx := context.Background()
-
-	configDetails := types.ConfigDetails{
-		ConfigFiles: []types.ConfigFile{
-			{
-				Filename: "docker-compose.yml",
-				Content:  []byte(content),
-			},
-		},
-	}
-
-	project, err := loader.LoadWithContext(ctx, configDetails, func(o *loader.Options) {
-		o.SetProjectName("test", true)
-		// Register our custom extensions
-		if o.KnownExtensions == nil {
-			o.KnownExtensions = map[string]any{}
-		}
-		o.KnownExtensions[CaddyExtensionKey] = Caddy{}
-		o.KnownExtensions[PortsExtensionKey] = PortsSource{}
-		o.KnownExtensions[MachinesExtensionKey] = MachinesSource{}
-	})
-	if err != nil {
-		return nil, err
-	}
-
-	removeProjectPrefixFromNames(project)
-	// Apply extension transformations since we're not using LoadProject.
-	if project, err = transformServicesCaddyExtension(project); err != nil {
-		return nil, err
-	}
-	if project, err = transformServicesPortsExtension(project); err != nil {
-		return nil, err
-	}
-
-	// Validate extension combinations after all transformations.
-	if err = validateServicesExtensions(project); err != nil {
-		return nil, err
-	}
-
-	// Process image templates in services to expand Go template expressions using git repo state.
-	if project, err = ProcessImageTemplates(project); err != nil {
-		return nil, err
-	}
-
-	return project, nil
-}
-
 func TestServiceSpecFromCompose(t *testing.T) {
 	t.Parallel()
 
@@ -431,7 +380,15 @@ services:
 
 	for _, tt := range tests {
 		t.Run(tt.name, func(t *testing.T) {
-			project, err := loadProjectFromContent(t, tt.composeYAML)
+			// Get current working directory for relative path resolution to testdata directory.
+			wd, err := os.Getwd()
+			require.NoError(t, err)
+
+			project, err := LoadProjectFromContent(
+				context.Background(),
+				tt.composeYAML,
+				composecli.WithWorkingDirectory(wd),
+			)
 			require.NoError(t, err)
 
 			spec, err := ServiceSpecFromCompose(project, "web")
@@ -617,7 +574,7 @@ services:
 
 	for _, tt := range tests {
 		t.Run(tt.name, func(t *testing.T) {
-			project, err := loadProjectFromContent(t, tt.composeYAML)
+			project, err := LoadProjectFromContent(context.Background(), tt.composeYAML)
 			require.NoError(t, err)
 
 			spec, err := ServiceSpecFromCompose(project, "ai")
@@ -629,6 +586,114 @@ services:
 	}
 }
 
+func TestServiceSpecFromCompose_VolumeDriverOpts(t *testing.T) {
+	tests := []struct {
+		name           string
+		composeYAML    string
+		expectedVolume api.VolumeSpec
+	}{
+		{
+			name: "volume with driver_opts only (no driver specified)",
+			composeYAML: `
+services:
+  test:
+    image: nginx
+    volumes:
+      - nfsmount:/data/nfs
+volumes:
+  nfsmount:
+    driver_opts:
+      type: "nfs"
+      o: "addr=192.168.1.100,nolock,soft"
+      device: ":/mnt/share"
+`,
+			expectedVolume: api.VolumeSpec{
+				Name: "nfsmount",
+				Type: api.VolumeTypeVolume,
+				VolumeOptions: &api.VolumeOptions{
+					Name: "nfsmount",
+					Driver: &mount.Driver{
+						Options: map[string]string{
+							"type":   "nfs",
+							"o":      "addr=192.168.1.100,nolock,soft",
+							"device": ":/mnt/share",
+						},
+					},
+				},
+			},
+		},
+		{
+			name: "volume with driver and driver_opts",
+			composeYAML: `
+services:
+  test:
+    image: nginx
+    volumes:
+      - nfsmount:/data/nfs
+volumes:
+  nfsmount:
+    driver: local
+    driver_opts:
+      type: "nfs"
+      o: "addr=192.168.1.100,nolock,soft"
+      device: ":/mnt/share"
+`,
+			expectedVolume: api.VolumeSpec{
+				Name: "nfsmount",
+				Type: api.VolumeTypeVolume,
+				VolumeOptions: &api.VolumeOptions{
+					Name: "nfsmount",
+					Driver: &mount.Driver{
+						Name: "local",
+						Options: map[string]string{
+							"type":   "nfs",
+							"o":      "addr=192.168.1.100,nolock,soft",
+							"device": ":/mnt/share",
+						},
+					},
+				},
+			},
+		},
+		{
+			name: "external volume ignores driver_opts",
+			composeYAML: `
+services:
+  test:
+    image: nginx
+    volumes:
+      - extvolume:/data
+volumes:
+  extvolume:
+    external: true
+`,
+			expectedVolume: api.VolumeSpec{
+				Name: "extvolume",
+				Type: api.VolumeTypeVolume,
+				VolumeOptions: &api.VolumeOptions{
+					Name: "extvolume",
+				},
+			},
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			project, err := LoadProjectFromContent(context.Background(), tt.composeYAML)
+			require.NoError(t, err)
+
+			spec, err := ServiceSpecFromCompose(project, "test")
+			require.NoError(t, err)
+
+			require.Len(t, spec.Volumes, 1, "expected exactly one volume")
+			actualVolume := spec.Volumes[0]
+
+			cmpOpts := cmp.Options{cmpopts.EquateEmpty()}
+			assert.True(t, cmp.Equal(actualVolume, tt.expectedVolume, cmpOpts...),
+				"Volume mismatch:\n%s", cmp.Diff(actualVolume, tt.expectedVolume, cmpOpts...))
+		})
+	}
+}
+
 func TestServiceSpecFromCompose_XMachinesPlacement(t *testing.T) {
 	tests := []struct {
 		name        string
@@ -765,7 +830,7 @@ services:
 
 	for _, tt := range tests {
 		t.Run(tt.name, func(t *testing.T) {
-			project, err := loadProjectFromContent(t, tt.composeYAML)
+			project, err := LoadProjectFromContent(context.Background(), tt.composeYAML)
 
 			if tt.expectError {
 				assert.Error(t, err)

pkg/client/connector/unix.go

@@ -0,0 +1,41 @@
+package connector
+
+import (
+	"context"
+	"fmt"
+
+	"golang.org/x/net/proxy"
+	"google.golang.org/grpc"
+	"google.golang.org/grpc/credentials/insecure"
+)
+
+// UnixConnector establishes a connection to the machine API through a unix domain socket.
+type UnixConnector struct {
+	socketPath string
+}
+
+func NewUnixConnector(socketPath string) *UnixConnector {
+	return &UnixConnector{socketPath: socketPath}
+}
+
+func (c *UnixConnector) Connect(_ context.Context) (*grpc.ClientConn, error) {
+	// gRPC uses "unix:path" syntax for unix sockets.
+	target := "unix:" + c.socketPath
+
+	conn, err := grpc.NewClient(
+		target,
+		grpc.WithTransportCredentials(insecure.NewCredentials()),
+	)
+	if err != nil {
+		return nil, fmt.Errorf("create machine API client: %w", err)
+	}
+	return conn, nil
+}
+
+func (c *UnixConnector) Dialer() (proxy.ContextDialer, error) {
+	return nil, fmt.Errorf("proxy connections are not supported over a unix connection")
+}
+
+func (c *UnixConnector) Close() error {
+	return nil
+}

pkg/client/deploy/deploy.go

@@ -6,6 +6,7 @@ import (
 	"fmt"
 
 	"github.com/psviderski/uncloud/pkg/api"
+	"github.com/psviderski/uncloud/pkg/client/deploy/scheduler"
 )
 
 type Client interface {
@@ -25,6 +26,8 @@ type Deployment struct {
 	Strategy Strategy
 	cli      Client
 	plan     *Plan
+	// state is an optional current and planned cluster state used for scheduling decisions.
+	state *scheduler.ClusterState
 }
 
 type Plan struct {
@@ -47,6 +50,16 @@ func NewDeployment(cli Client, spec api.ServiceSpec, strategy Strategy) *Deploym
 	}
 }
 
+// NewDeploymentWithClusterState creates a new deployment like NewDeployment but also with a provided current cluster
+// state used for scheduling decisions.
+func NewDeploymentWithClusterState(
+	cli Client, spec api.ServiceSpec, strategy Strategy, state *scheduler.ClusterState,
+) *Deployment {
+	d := NewDeployment(cli, spec, strategy)
+	d.state = state
+	return d
+}
+
 // Plan returns a plan of operations to reconcile the service to the desired state.
 // If a plan has already been created, the same plan will be returned.
 func (d *Deployment) Plan(ctx context.Context) (Plan, error) {
@@ -73,7 +86,14 @@ func (d *Deployment) Plan(ctx context.Context) (Plan, error) {
 		return Plan{}, fmt.Errorf("resolve service spec: %w", err)
 	}
 
-	plan, err := d.Strategy.Plan(ctx, d.cli, d.Service, resolvedSpec)
+	if d.state == nil {
+		d.state, err = scheduler.InspectClusterState(ctx, d.cli)
+		if err != nil {
+			return Plan{}, fmt.Errorf("inspect cluster state: %w", err)
+		}
+	}
+
+	plan, err := d.Strategy.Plan(d.state, d.Service, resolvedSpec)
 	if err != nil {
 		return Plan{}, fmt.Errorf("create plan using %s strategy: %w", d.Strategy.Type(), err)
 	}

pkg/client/deploy/scheduler/volume.go

@@ -17,7 +17,7 @@ import (
 //   - If a volume already exists on a machine, it must be used instead of creating a new one.
 //   - A missing volume must only be created on one machine.
 type VolumeScheduler struct {
-	// state is the current state of machines and their resources in the cluster.
+	// state is the current and planned state of machines and their resources in the cluster.
 	state *ClusterState
 	// serviceSpecs is a list of service specifications included in the deployment.
 	serviceSpecs []api.ServiceSpec
@@ -83,7 +83,10 @@ func NewVolumeScheduler(state *ClusterState, specs []api.ServiceSpec) (*VolumeSc
 			if spec, ok := volumeSpecs[vol.Name]; ok {
 				if !spec.MatchesDockerVolume(vol) {
 					return nil, fmt.Errorf("volume '%s' specification does not match the existing volume "+
-						"on machine '%s'", vol.Name, machine.Info.Name)
+						"on machine '%s'. Use a different volume name or adjust the volume options to match "+
+						"the existing volume. You can also remove the existing volume from the machine(s) with "+
+						"'uc volume rm' (WARNING: the data will be lost) and run the deployment again to create "+
+						"a new volume with the correct specification", vol.Name, machine.Info.Name)
 				}
 
 				if _, setInitialised := existingVolumeMachines[vol.Name]; !setInitialised {
@@ -106,6 +109,7 @@ func NewVolumeScheduler(state *ClusterState, specs []api.ServiceSpec) (*VolumeSc
 // Schedule determines what missing volumes should be created and where for services in the multi-service deployment.
 // It returns a map of machine IDs to a list of api.VolumeSpec that should be created on that machine,
 // or an error if services can't be scheduled due to scheduling constraints.
+// It also updates the state of the machines in the cluster state to reflect the scheduled volumes.
 func (s *VolumeScheduler) Schedule() (map[string][]api.VolumeSpec, error) {
 	if len(s.serviceSpecs) == 0 {
 		// No services with volume mounts, nothing to schedule.

pkg/client/deploy/strategy.go

@@ -1,7 +1,6 @@
 package deploy
 
 import (
-	"context"
 	"fmt"
 	"math/rand/v2"
 	"slices"
@@ -18,31 +17,31 @@ type Strategy interface {
 	// Type returns the type of the deployment strategy, e.g. "rolling", "blue-green".
 	Type() string
 	// Plan returns the operation to reconcile the service to the desired state.
-	// If the service does not exist (new deployment), svc will be nil.
-	Plan(ctx context.Context, cli scheduler.Client, svc *api.Service, spec api.ServiceSpec) (Plan, error)
+	// If the service does not exist (new deployment), svc will be nil. state provides the current and planned state
+	// of the cluster for scheduling decisions.
+	Plan(state *scheduler.ClusterState, svc *api.Service, spec api.ServiceSpec) (Plan, error)
 }
 
 // RollingStrategy implements a rolling update deployment pattern where containers are updated one at a time
 // to minimize service disruption.
 type RollingStrategy struct {
-	State         *scheduler.ClusterState
+	// ForceRecreate indicates whether all containers should be recreated during the deployment,
+	// regardless of whether their specifications have changed.
 	ForceRecreate bool
+
+	// state is the current and planned state of the cluster used for scheduling decisions.
+	state *scheduler.ClusterState
 }
 
 func (s *RollingStrategy) Type() string {
 	return "rolling"
 }
 
-func (s *RollingStrategy) Plan(
-	ctx context.Context, cli scheduler.Client, svc *api.Service, spec api.ServiceSpec,
-) (Plan, error) {
-	if s.State == nil {
-		state, err := scheduler.InspectClusterState(ctx, cli)
-		if err != nil {
-			return Plan{}, fmt.Errorf("inspect cluster state: %w", err)
-		}
-		s.State = state
+func (s *RollingStrategy) Plan(state *scheduler.ClusterState, svc *api.Service, spec api.ServiceSpec) (Plan, error) {
+	if state == nil {
+		return Plan{}, fmt.Errorf("cluster state must be provided")
 	}
+	s.state = state
 
 	// We can assume that the spec is valid at this point because it has been validated by the deployment.
 	switch spec.Mode {
@@ -65,7 +64,7 @@ func (s *RollingStrategy) planReplicated(svc *api.Service, spec api.ServiceSpec)
 		return plan, err
 	}
 
-	sched := scheduler.NewServiceScheduler(s.State, spec)
+	sched := scheduler.NewServiceScheduler(s.state, spec)
 	// TODO: return a detailed report on required constraints and which ones are satisfied?
 	availableMachines, err := sched.EligibleMachines()
 	if err != nil {
@@ -221,7 +220,7 @@ func (s *RollingStrategy) planGlobal(svc *api.Service, spec api.ServiceSpec) (Pl
 		}
 	}
 
-	sched := scheduler.NewServiceScheduler(s.State, spec)
+	sched := scheduler.NewServiceScheduler(s.state, spec)
 	availableMachines, err := sched.EligibleMachines()
 	if err != nil {
 		return plan, err

pkg/client/image.go

@@ -48,7 +48,7 @@ func (cli *Client) InspectRemoteImage(ctx context.Context, id string) ([]api.Mac
 // it lists images on all machines.
 func (cli *Client) ListImages(ctx context.Context, filter api.ImageFilter) ([]api.MachineImages, error) {
 	// Broadcast the image list request to the specified machines or all machines if none specified.
-	listCtx, machines, err := api.ProxyMachinesContext(ctx, cli, filter.Machines)
+	listCtx, machines, err := cli.ProxyMachinesContext(ctx, filter.Machines)
 	if err != nil {
 		return nil, fmt.Errorf("create request context to broadcast to machines: %w", err)
 	}
@@ -149,18 +149,6 @@ func (cli *Client) PushImage(ctx context.Context, image string, opts PushImageOp
 			return fmt.Errorf("list machines: %w", err)
 		}
 
-		// Check if all specified machines were found.
-		if len(machineMembers) != len(opts.Machines) {
-			var notFound []string
-			for _, nameOrID := range opts.Machines {
-				if machineMembers.FindByNameOrID(nameOrID) == nil {
-					notFound = append(notFound, nameOrID)
-				}
-			}
-
-			return fmt.Errorf("machines not found: %s", strings.Join(notFound, ", "))
-		}
-
 		for _, mm := range machineMembers {
 			machines = append(machines, mm.Machine)
 		}

pkg/client/logmerger.go

@@ -0,0 +1,310 @@
+package client
+
+import (
+	"container/heap"
+	"sync"
+	"time"
+
+	"github.com/psviderski/uncloud/pkg/api"
+)
+
+// logMergerMaxInFlightPerStream limits how many entries each input stream can have in the processing queue before being
+// throttled. This ensures fair interleaving between streams and prevents one fast stream from causing unbounded
+// buffering while waiting for slower streams.
+const (
+	logMergerMaxInFlightPerStream = 100
+	// logMergerHeartbeatDebounceInterval defines the minimum interval between deduplicated heartbeat entries emitted
+	// by the LogMerger. Keep it in sync with logsHeartbeatInterval internal/machine/docker/server.go.
+	logMergerHeartbeatDebounceInterval = 200 * time.Millisecond
+)
+
+// LogMergerOptions configures the behavior of LogMerger.
+type LogMergerOptions struct {
+	// StallTimeout specifies how long a stream can go without receiving any data before it's considered
+	// stalled and excluded from watermark calculation. A zero timeout disables stall detection.
+	StallTimeout time.Duration
+	// StallCheckInterval specifies how often to check for stalled streams.
+	StallCheckInterval time.Duration
+}
+
+// DefaultLogMergerOptions provides sensible default options that enable stall detection for LogMerger.
+var DefaultLogMergerOptions = LogMergerOptions{
+	StallTimeout:       10 * time.Second,
+	StallCheckInterval: 1 * time.Second,
+}
+
+// LogMerger merges multiple log streams into a single chronologically ordered stream based on timestamps.
+// It uses a low watermark algorithm to ensure proper ordering across streams.
+// Heartbeat entries from streams advance the watermark to enable timely emission of buffered logs.
+type LogMerger struct {
+	streams []*mergerStream
+	queue   logsHeap
+	// watermark is min(latest_timestamp for each stream).
+	watermark time.Time
+	output    chan api.ServiceLogEntry
+	// lastEmitted is the timestamp of the last emitted log entry or heartbeat.
+	lastEmitted        time.Time
+	stallTimeout       time.Duration
+	stallCheckInterval time.Duration
+}
+
+// NewLogMerger creates a new LogMerger for the given input streams with the specified options.
+func NewLogMerger(streams []<-chan api.ServiceLogEntry, opts LogMergerOptions) *LogMerger {
+	mergerStreams := make([]*mergerStream, len(streams))
+	now := time.Now()
+	for i, ch := range streams {
+		mergerStreams[i] = &mergerStream{
+			stream:       ch,
+			semaphore:    make(chan struct{}, logMergerMaxInFlightPerStream),
+			lastActivity: now,
+		}
+	}
+
+	return &LogMerger{
+		streams:            mergerStreams,
+		output:             make(chan api.ServiceLogEntry),
+		stallTimeout:       opts.StallTimeout,
+		stallCheckInterval: opts.StallCheckInterval,
+	}
+}
+
+// Stream starts the merge process and returns a channel that emits log entries in chronological order.
+// The returned channel is closed when all input streams are closed.
+func (m *LogMerger) Stream() <-chan api.ServiceLogEntry {
+	if len(m.streams) == 0 {
+		close(m.output)
+		return m.output
+	}
+
+	go m.run()
+
+	return m.output
+}
+
+// mergerStream combines a stream channel with its state and flow control.
+type mergerStream struct {
+	stream    <-chan api.ServiceLogEntry
+	semaphore chan struct{}
+	// Latest timestamp seen from this stream (log or heartbeat).
+	lastSeen time.Time
+	// Wall clock time when we last received any data from this stream.
+	lastActivity time.Time
+	// Metadata associated with this stream. It's populated from the first log entry received.
+	metadata *api.ServiceLogEntryMetadata
+	// Whether the stream channel has closed.
+	closed bool
+	// Whether the stream is considered stalled (no data received within timeout).
+	stalled bool
+}
+
+// streamEvent represents an event from a stream (entry received or stream closed).
+type streamEvent struct {
+	stream *mergerStream
+	entry  api.ServiceLogEntry
+	closed bool
+}
+
+// queuedEntry wraps a log entry with its source semaphore for release tracking.
+type queuedEntry struct {
+	entry     api.ServiceLogEntry
+	semaphore chan struct{}
+}
+
+// run is the main processing loop that merges all streams.
+func (m *LogMerger) run() {
+	defer close(m.output)
+
+	// Fan-in channel for stream events.
+	events := make(chan streamEvent)
+
+	// Start a reader goroutine for each stream to send entries to the events channel with flow control.
+	var wg sync.WaitGroup
+	for _, stream := range m.streams {
+		wg.Go(func() {
+			for entry := range stream.stream {
+				// Acquire semaphore slot before sending the entry to limit in-flight unprocessed entries per stream.
+				stream.semaphore <- struct{}{}
+				events <- streamEvent{stream: stream, entry: entry}
+			}
+
+			events <- streamEvent{stream: stream, closed: true}
+		})
+	}
+
+	// Close events channel when all readers finish.
+	go func() {
+		wg.Wait()
+		close(events)
+	}()
+
+	// Set up stall detection timer if enabled.
+	var stallCh <-chan time.Time
+	if m.stallTimeout > 0 && m.stallCheckInterval > 0 {
+		stallTicker := time.NewTicker(m.stallCheckInterval)
+		stallCh = stallTicker.C
+		defer stallTicker.Stop()
+	}
+
+	// Process events and emit entries.
+	for {
+		select {
+		case e, ok := <-events:
+			if !ok {
+				// All streams closed: flush remaining entries in order.
+				for m.queue.Len() > 0 {
+					qe := heap.Pop(&m.queue).(queuedEntry)
+					m.output <- qe.entry
+					<-qe.semaphore
+				}
+				return
+			}
+
+			e.stream.lastActivity = time.Now()
+			if e.stream.stalled {
+				e.stream.stalled = false
+			}
+			if e.stream.metadata == nil {
+				e.stream.metadata = &e.entry.Metadata
+			}
+
+			if e.closed {
+				e.stream.closed = true
+
+				m.updateWatermark()
+				m.emitReadyEntries()
+				continue
+			}
+
+			// Forward errors immediately and release semaphore.
+			if e.entry.Err != nil {
+				m.output <- e.entry
+				<-e.stream.semaphore
+				continue
+			}
+
+			if e.entry.Timestamp.After(e.stream.lastSeen) {
+				e.stream.lastSeen = e.entry.Timestamp
+			}
+			if e.entry.Stream == api.LogStreamStdout || e.entry.Stream == api.LogStreamStderr {
+				heap.Push(&m.queue, queuedEntry{entry: e.entry, semaphore: e.stream.semaphore})
+			}
+
+			m.updateWatermark()
+			m.emitReadyEntries()
+
+			// When merging streams, each input emits its own heartbeats. We want to debounce them and emit our own
+			// heartbeats at the same rate as a single input stream. Note that we need to adjust the heartbeat timestamp
+			// to the current watermark to not violate ordering guarantees.
+			if e.entry.Stream == api.LogStreamHeartbeat {
+				if m.watermark.Sub(m.lastEmitted) >= logMergerHeartbeatDebounceInterval {
+					heartbeat := e.entry
+					heartbeat.Timestamp = m.watermark
+					m.output <- heartbeat
+					m.lastEmitted = m.watermark
+				}
+				// Heartbeat processed, release semaphore.
+				<-e.stream.semaphore
+			}
+
+		case <-stallCh:
+			stalled := m.checkStalledStreams()
+			if len(stalled) == 0 {
+				continue
+			}
+
+			for _, s := range stalled {
+				errEntry := api.ServiceLogEntry{
+					ContainerLogEntry: api.ContainerLogEntry{
+						Err: api.ErrLogStreamStalled,
+					},
+				}
+				if s.metadata != nil {
+					errEntry.Metadata = *s.metadata
+				}
+
+				m.output <- errEntry
+			}
+
+			m.updateWatermark()
+			m.emitReadyEntries()
+		}
+	}
+}
+
+// checkStalledStreams marks streams as stalled if they haven't received any data within the timeout.
+// Returns true if any stream's stalled state changed.
+func (m *LogMerger) checkStalledStreams() []*mergerStream {
+	var stalled []*mergerStream
+	now := time.Now()
+
+	for _, s := range m.streams {
+		if s.closed || s.stalled {
+			continue
+		}
+
+		if now.Sub(s.lastActivity) > m.stallTimeout {
+			s.stalled = true
+			stalled = append(stalled, s)
+		}
+	}
+
+	return stalled
+}
+
+// updateWatermark recalculates the low watermark based on the lastSeen timestamps of all active streams.
+func (m *LogMerger) updateWatermark() {
+	first := true
+
+	for _, s := range m.streams {
+		if s.closed || s.stalled {
+			// Closed and stalled streams don't affect watermark.
+			continue
+		}
+		if first || s.lastSeen.Before(m.watermark) {
+			m.watermark = s.lastSeen
+			first = false
+		}
+	}
+}
+
+// emitReadyEntries pops and emits all buffered entries from the queue with timestamp before the watermark.
+func (m *LogMerger) emitReadyEntries() {
+	if m.watermark.IsZero() {
+		// No entries received yet.
+		return
+	}
+
+	for m.queue.Len() > 0 && m.queue[0].entry.Timestamp.Compare(m.watermark) <= 0 {
+		qe := heap.Pop(&m.queue).(queuedEntry)
+		m.output <- qe.entry
+		m.lastEmitted = qe.entry.Timestamp
+		<-qe.semaphore
+	}
+}
+
+// logsHeap is a min-heap (heap.Interface) of queued entries ordered by timestamp.
+type logsHeap []queuedEntry
+
+func (h *logsHeap) Len() int {
+	return len(*h)
+}
+
+func (h *logsHeap) Less(i, j int) bool {
+	return (*h)[i].entry.Timestamp.Before((*h)[j].entry.Timestamp)
+}
+
+func (h *logsHeap) Swap(i, j int) {
+	(*h)[i], (*h)[j] = (*h)[j], (*h)[i]
+}
+
+func (h *logsHeap) Push(x any) {
+	*h = append(*h, x.(queuedEntry))
+}
+
+func (h *logsHeap) Pop() any {
+	old := *h
+	n := len(old)
+	x := old[n-1]
+	*h = old[:n-1]
+	return x
+}

pkg/client/logmerger_test.go

@@ -0,0 +1,383 @@
+package client
+
+import (
+	"testing"
+	"time"
+
+	"github.com/psviderski/uncloud/pkg/api"
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+)
+
+// testEntry creates a ServiceLogEntry for testing.
+func testEntry(stream api.LogStreamType, ts time.Time, msg string) api.ServiceLogEntry {
+	return api.ServiceLogEntry{
+		ContainerLogEntry: api.ContainerLogEntry{
+			Stream:    stream,
+			Timestamp: ts,
+			Message:   []byte(msg),
+		},
+	}
+}
+
+// collectEntries collects up to maxCount entries from the channel or until it is closed.
+func collectEntries(t *testing.T, ch <-chan api.ServiceLogEntry, maxCount int) []api.ServiceLogEntry {
+	t.Helper()
+	var entries []api.ServiceLogEntry
+
+	for i := 0; i < maxCount || maxCount <= 0; i++ {
+		select {
+		case e, ok := <-ch:
+			if !ok {
+				return entries
+			}
+			entries = append(entries, e)
+		case <-time.After(1 * time.Second):
+			require.FailNow(t, "timed out waiting for log entry")
+		}
+	}
+
+	return entries
+}
+
+func TestLogMerger_EmptyStreams(t *testing.T) {
+	t.Parallel()
+
+	merger := NewLogMerger(nil, LogMergerOptions{})
+	output := merger.Stream()
+
+	_, ok := <-output
+	assert.False(t, ok, "output channel should be closed for empty streams")
+}
+
+func TestLogMerger_SingleStream(t *testing.T) {
+	t.Parallel()
+
+	ch := make(chan api.ServiceLogEntry, 10)
+	merger := NewLogMerger([]<-chan api.ServiceLogEntry{ch}, LogMergerOptions{})
+	output := merger.Stream()
+
+	t1 := time.Now()
+	t2 := t1.Add(time.Second)
+
+	ch <- testEntry(api.LogStreamStdout, t1, "first")
+	ch <- testEntry(api.LogStreamStdout, t2, "second")
+
+	results := collectEntries(t, output, 2)
+
+	require.Len(t, results, 2)
+	assert.Equal(t, "first", string(results[0].Message))
+	assert.Equal(t, "second", string(results[1].Message))
+
+	close(ch)
+
+	results = collectEntries(t, output, 0)
+	assert.Len(t, results, 0, "no entries expected after close")
+}
+
+func TestLogMerger_PreservesData(t *testing.T) {
+	t.Parallel()
+
+	ch := make(chan api.ServiceLogEntry, 10)
+	merger := NewLogMerger([]<-chan api.ServiceLogEntry{ch}, LogMergerOptions{})
+	output := merger.Stream()
+
+	metadata := api.ServiceLogEntryMetadata{
+		ServiceID:   "svc-123",
+		ServiceName: "my-service",
+		MachineID:   "machine-456",
+		MachineName: "machine-1",
+	}
+
+	e := api.ServiceLogEntry{
+		Metadata: metadata,
+		ContainerLogEntry: api.ContainerLogEntry{
+			Stream:    api.LogStreamStdout,
+			Timestamp: time.Now(),
+			Message:   []byte("test"),
+		},
+	}
+	ch <- e
+	close(ch)
+
+	results := collectEntries(t, output, 0)
+	require.Len(t, results, 1)
+	assert.Equal(t, e, results[0])
+}
+
+func TestLogMerger_BasicMerge(t *testing.T) {
+	t.Parallel()
+
+	ch1 := make(chan api.ServiceLogEntry, 10)
+	ch2 := make(chan api.ServiceLogEntry, 10)
+
+	merger := NewLogMerger([]<-chan api.ServiceLogEntry{ch1, ch2}, LogMergerOptions{})
+	output := merger.Stream()
+
+	t1 := time.Now()
+	t2 := t1.Add(time.Second)
+	t3 := t1.Add(2 * time.Second)
+	t4 := t1.Add(3 * time.Second)
+
+	// Send interleaved entries from both streams.
+	ch1 <- testEntry(api.LogStreamStdout, t1, "ch1-first")
+	ch2 <- testEntry(api.LogStreamStdout, t2, "ch2-first")
+	ch1 <- testEntry(api.LogStreamStdout, t3, "ch1-second")
+	ch2 <- testEntry(api.LogStreamStdout, t4, "ch2-second")
+
+	// Send heartbeats to advance watermark past all entries.
+	t5 := t1.Add(4 * time.Second)
+	ch1 <- testEntry(api.LogStreamHeartbeat, t5, "")
+	ch2 <- testEntry(api.LogStreamHeartbeat, t5, "")
+
+	results := collectEntries(t, output, 4)
+	require.Len(t, results, 4)
+	assert.Equal(t, "ch1-first", string(results[0].Message))
+	assert.Equal(t, "ch2-first", string(results[1].Message))
+	assert.Equal(t, "ch1-second", string(results[2].Message))
+	assert.Equal(t, "ch2-second", string(results[3].Message))
+
+	close(ch1)
+	close(ch2)
+
+	results = collectEntries(t, output, 0)
+	assert.Len(t, results, 1, "one debounced heartbeat expected after close")
+	assert.Equal(t, api.LogStreamHeartbeat, results[0].Stream)
+	assert.Equal(t, t5, results[0].Timestamp)
+}
+
+func TestLogMerger_HeartbeatAdvancesWatermark(t *testing.T) {
+	t.Parallel()
+
+	ch1 := make(chan api.ServiceLogEntry, 10)
+	ch2 := make(chan api.ServiceLogEntry, 10)
+
+	merger := NewLogMerger([]<-chan api.ServiceLogEntry{ch1, ch2}, LogMergerOptions{})
+	output := merger.Stream()
+
+	t1 := time.Now()
+	t2 := t1.Add(time.Second)
+	t3 := t1.Add(2 * time.Second)
+
+	// Stream 1 sends a log entry.
+	ch1 <- testEntry(api.LogStreamStdout, t1, "ch1-first")
+	ch1 <- testEntry(api.LogStreamStdout, t3, "ch1-second")
+	// Stream 2 is quiet but sends a heartbeat.
+	ch2 <- testEntry(api.LogStreamHeartbeat, t2, "")
+
+	// Now stream 1's entry should be emitted because watermark is t1 (min of t1, t2).
+	results := collectEntries(t, output, 1)
+	require.Len(t, results, 1)
+	assert.Equal(t, "ch1-first", string(results[0].Message))
+
+	close(ch1)
+	close(ch2)
+
+	results = collectEntries(t, output, 0)
+	require.NotEmpty(t, results)
+
+	// Depending on timing, we may get a heartbeat before the second log entry.
+	require.LessOrEqual(t, len(results), 2, "one or two entries expected after close")
+	if len(results) == 2 {
+		assert.Equal(t, api.LogStreamHeartbeat, results[0].Stream)
+		assert.Equal(t, t2, results[0].Timestamp)
+		results = results[1:]
+	}
+	assert.Equal(t, "ch1-second", string(results[0].Message))
+}
+
+func TestLogMerger_ErrorForwarding(t *testing.T) {
+	t.Parallel()
+
+	ch1 := make(chan api.ServiceLogEntry, 10)
+	ch2 := make(chan api.ServiceLogEntry, 10)
+	merger := NewLogMerger([]<-chan api.ServiceLogEntry{ch1, ch2}, LogMergerOptions{})
+	output := merger.Stream()
+
+	t1 := time.Now()
+
+	ch1 <- testEntry(api.LogStreamStdout, t1, "ch1-first")
+	// Send an error entry.
+	ch1 <- api.ServiceLogEntry{
+		ContainerLogEntry: api.ContainerLogEntry{
+			Err: assert.AnError,
+		},
+	}
+
+	results := collectEntries(t, output, 1)
+	require.Len(t, results, 1, "expected error to emitted out of order")
+	assert.Equal(t, assert.AnError, results[0].Err)
+
+	close(ch1)
+	close(ch2)
+
+	results = collectEntries(t, output, 0)
+	assert.Len(t, results, 1)
+	assert.Equal(t, "ch1-first", string(results[0].Message))
+}
+
+func TestLogMerger_OutOfOrderSingleStream(t *testing.T) {
+	t.Parallel()
+
+	ch := make(chan api.ServiceLogEntry, 10)
+	merger := NewLogMerger([]<-chan api.ServiceLogEntry{ch}, LogMergerOptions{})
+	output := merger.Stream()
+
+	t1 := time.Now()
+	t2 := t1.Add(time.Second)
+	t3 := t1.Add(2 * time.Second)
+	t4 := t1.Add(3 * time.Second)
+
+	// Send entries out of order within the stream.
+	ch <- testEntry(api.LogStreamStdout, t3, "third")
+	ch <- testEntry(api.LogStreamStdout, t2, "second")
+	ch <- testEntry(api.LogStreamStdout, t4, "forth")
+	ch <- testEntry(api.LogStreamStdout, t1, "first")
+
+	results := collectEntries(t, output, 4)
+	require.Len(t, results, 4)
+	// Emitted in the same order as sent since no buffering/reordering is done within a single stream.
+	assert.Equal(t, "third", string(results[0].Message))
+	assert.Equal(t, "second", string(results[1].Message))
+	assert.Equal(t, "forth", string(results[2].Message))
+	assert.Equal(t, "first", string(results[3].Message))
+
+	close(ch)
+
+	results = collectEntries(t, output, 0)
+	assert.Len(t, results, 0, "no entries expected after close")
+}
+
+// Test that entries from streams with different rates are merged correctly in chronological order.
+func TestLogMerger_UnevenStreams(t *testing.T) {
+	t.Parallel()
+
+	numFastEntries := logMergerMaxInFlightPerStream
+
+	// Use buffered channels to allow sends without blocking on receiver.
+	ch1 := make(chan api.ServiceLogEntry, numFastEntries+5)
+	ch2 := make(chan api.ServiceLogEntry, 10)
+
+	merger := NewLogMerger([]<-chan api.ServiceLogEntry{ch1, ch2}, LogMergerOptions{})
+	output := merger.Stream()
+
+	baseTime := time.Now()
+
+	// Stream 1 sends many entries quickly (0ms - 104ms).
+	for i := 0; i < numFastEntries; i++ {
+		ch1 <- testEntry(api.LogStreamStdout, baseTime.Add(time.Duration(i)*time.Millisecond), "fast")
+	}
+
+	// Stream 2 sends entries that interleave with fast entries.
+	ch2 <- testEntry(api.LogStreamStdout, baseTime.Add(5*time.Millisecond), "slow")
+	ch2 <- testEntry(api.LogStreamStdout, baseTime.Add(50*time.Millisecond), "slow")
+
+	// Send heartbeats to advance watermark past all entries so they get emitted.
+	finalTime := baseTime.Add(time.Second)
+	ch1 <- testEntry(api.LogStreamHeartbeat, finalTime, "")
+	ch2 <- testEntry(api.LogStreamHeartbeat, finalTime.Add(10*time.Millisecond), "")
+
+	// Read exactly numFastEntries + 2 log entries.
+	expectedCount := numFastEntries + 2
+	results := collectEntries(t, output, expectedCount)
+	require.Len(t, results, expectedCount)
+
+	// Count entries from each stream.
+	fastCount := 0
+	slowCount := 0
+	for _, entry := range results {
+		switch string(entry.Message) {
+		case "fast":
+			fastCount++
+		case "slow":
+			slowCount++
+		}
+	}
+	assert.Equal(t, numFastEntries, fastCount)
+	assert.Equal(t, 2, slowCount)
+
+	// Verify entries are in chronological order.
+	for i := 1; i < len(results); i++ {
+		assert.False(t, results[i].Timestamp.Before(results[i-1].Timestamp),
+			"entry %d (ts=%v) should not be before entry %d (ts=%v)",
+			i, results[i].Timestamp, i-1, results[i-1].Timestamp)
+	}
+
+	close(ch1)
+	close(ch2)
+
+	results = collectEntries(t, output, 0)
+	require.Len(t, results, 1, "one debounced heartbeat expected after close")
+	assert.Equal(t, api.LogStreamHeartbeat, results[0].Stream)
+}
+
+func TestLogMerger_StalledStreamExcludedFromWatermark(t *testing.T) {
+	t.Parallel()
+
+	ch1 := make(chan api.ServiceLogEntry, 10)
+	ch2 := make(chan api.ServiceLogEntry, 10)
+
+	stallTimeout := 100 * time.Millisecond
+	merger := NewLogMerger([]<-chan api.ServiceLogEntry{ch1, ch2}, LogMergerOptions{
+		StallTimeout:       stallTimeout,
+		StallCheckInterval: 20 * time.Millisecond,
+	})
+	output := merger.Stream()
+
+	//// Collect all output entries in a separate goroutine.
+	//var results []api.ServiceLogEntry
+	//doneCollecting := make(chan struct{})
+	//go func() {
+	//	for entry := range output {
+	//		results = append(results, entry)
+	//	}
+	//	close(doneCollecting)
+	//}()
+
+	t1 := time.Now()
+	t2 := t1.Add(time.Second)
+	t3 := t1.Add(2 * time.Second)
+	t4 := t1.Add(3 * time.Second)
+
+	e1 := testEntry(api.LogStreamStdout, t1, "ch1-entry")
+	e1.Metadata = api.ServiceLogEntryMetadata{
+		ServiceID:   "ch1-svc1",
+		ServiceName: "ch1-svcName",
+		ContainerID: "ch1-ctr1",
+		MachineID:   "ch1-machine1",
+		MachineName: "ch1-machineName",
+	}
+
+	ch1 <- e1
+	ch2 <- testEntry(api.LogStreamStdout, t2, "ch2-first")
+
+	// Watermark is t1 now so ch1's entry could be collected.
+	results := collectEntries(t, output, 1)
+	require.Len(t, results, 1)
+	assert.Equal(t, "ch1-entry", string(results[0].Message))
+
+	// Keep pushing to stream 2 so that it does not stall. But it can't emit anything because watermark is stuck at t1.
+	time.Sleep(stallTimeout / 2)
+	ch2 <- testEntry(api.LogStreamStdout, t3, "ch2-second")
+
+	// 1/2 + 2/3 = 7/6 > 1, so stream 1 should be considered stalled now.
+	time.Sleep(stallTimeout / 3 * 2)
+	ch2 <- testEntry(api.LogStreamStdout, t4, "ch2-third")
+
+	results = collectEntries(t, output, 1)
+	require.ErrorIs(t, results[0].Err, api.ErrLogStreamStalled)
+	assert.Equal(t, e1.Metadata, results[0].Metadata, "stalled entry should have stream metadata")
+
+	// Now that stream 1 is marked as stalled and ignored, all in-flight entries from stream 2 are emitted.
+	results = collectEntries(t, output, 3)
+	require.Len(t, results, 3)
+	assert.Equal(t, "ch2-first", string(results[0].Message))
+	assert.Equal(t, "ch2-second", string(results[1].Message))
+	assert.Equal(t, "ch2-third", string(results[2].Message))
+
+	close(ch1)
+	close(ch2)
+
+	results = collectEntries(t, output, 0)
+	assert.Len(t, results, 0)
+}

pkg/client/logs.go

@@ -0,0 +1,159 @@
+package client
+
+import (
+	"context"
+	"errors"
+	"fmt"
+	"io"
+
+	"github.com/docker/docker/pkg/stringid"
+	"github.com/psviderski/uncloud/internal/machine/api/pb"
+	"github.com/psviderski/uncloud/pkg/api"
+)
+
+// ServiceLogs streams log entries from all service containers in chronological order based on timestamps.
+// Keep in mind that perfect ordering of log events across multiple machines can't be guaranteed due to the
+// imperfection of physical clocks or potential clock skew between machines.
+// It uses a low watermark algorithm to ensure proper ordering across multiple machines.
+// Heartbeat entries from the server advance the watermark to enable timely emission of buffered logs.
+func (cli *Client) ServiceLogs(
+	ctx context.Context, serviceNameOrID string, opts api.ServiceLogsOptions,
+) (api.Service, <-chan api.ServiceLogEntry, error) {
+	svc, err := cli.InspectService(ctx, serviceNameOrID)
+	if err != nil {
+		return svc, nil, fmt.Errorf("inspect service: %w", err)
+	}
+
+	if len(svc.Containers) == 0 {
+		return svc, nil, fmt.Errorf("no containers found for service: %s", serviceNameOrID)
+	}
+
+	machines, err := cli.ListMachines(ctx, &api.MachineFilter{
+		NamesOrIDs: opts.Machines,
+	})
+	if err != nil {
+		return svc, nil, fmt.Errorf("list machines: %w", err)
+	}
+
+	ctrStreams := make([]<-chan api.ServiceLogEntry, 0, len(svc.Containers))
+	for _, ctr := range svc.Containers {
+		// Skip containers not running on the specified machines.
+		m := machines.FindByNameOrID(ctr.MachineID)
+		if len(opts.Machines) > 0 && m == nil {
+			continue
+		}
+
+		// Machine name for ServiceLogEntry metadata and friendlier error message.
+		machineName := ctr.MachineID
+		if m != nil {
+			machineName = m.Machine.Name
+		}
+
+		stream, err := cli.ContainerLogs(ctx, ctr.MachineID, ctr.Container.ID, opts)
+		if err != nil {
+			return svc, nil, fmt.Errorf("stream logs from service container '%s' on machine '%s': %w",
+				stringid.TruncateID(ctr.Container.ID), machineName, err)
+		}
+
+		// Enrich log entries from the container with service metadata.
+		metadata := api.ServiceLogEntryMetadata{
+			ServiceID:   svc.ID,
+			ServiceName: svc.Name,
+			ContainerID: ctr.Container.ID,
+			MachineID:   ctr.MachineID,
+			MachineName: machineName,
+		}
+		enrichedStream := logsStreamWithServiceMetadata(stream, metadata)
+		ctrStreams = append(ctrStreams, enrichedStream)
+	}
+
+	if len(ctrStreams) == 0 {
+		return svc, nil, errors.New("no service containers found on the specified machine(s)")
+	}
+
+	// Use the log merger to combine streams from all containers in chronological order.
+	merger := NewLogMerger(ctrStreams, DefaultLogMergerOptions)
+	mergedStream := merger.Stream()
+
+	return svc, mergedStream, nil
+}
+
+// ContainerLogs streams log entries from a single container on a specified machine.
+func (cli *Client) ContainerLogs(
+	ctx context.Context, machineNameOrID string, containerID string, opts api.ServiceLogsOptions,
+) (<-chan api.ContainerLogEntry, error) {
+	proxyCtx, _, err := cli.ProxyMachinesContext(ctx, []string{machineNameOrID})
+	if err != nil {
+		return nil, fmt.Errorf("create request context to proxy to machine '%s': %w", machineNameOrID, err)
+	}
+
+	req := &pb.ContainerLogsRequest{
+		ContainerId: containerID,
+		Follow:      opts.Follow,
+		Tail:        int32(opts.Tail),
+		Since:       opts.Since,
+		Until:       opts.Until,
+	}
+	if !opts.Follow && opts.Tail == 0 {
+		// If not following and tail is 0, set tail to -1 to return all logs.
+		// Otherwise, no logs will be returned at all.
+		req.Tail = -1
+	}
+
+	stream, err := cli.Docker.GRPCClient.ContainerLogs(proxyCtx, req)
+	if err != nil {
+		return nil, err
+	}
+
+	ch := make(chan api.ContainerLogEntry)
+
+	go func() {
+		defer close(ch)
+
+		for {
+			pbEntry, err := stream.Recv()
+			if err == io.EOF {
+				return
+			}
+			if err != nil {
+				ch <- api.ContainerLogEntry{
+					Err: err,
+				}
+				return
+			}
+
+			entry := api.ContainerLogEntry{
+				Stream:    api.LogStreamTypeFromProto(pbEntry.Stream),
+				Message:   pbEntry.Message,
+				Timestamp: pbEntry.Timestamp.AsTime(),
+			}
+
+			select {
+			case ch <- entry:
+			case <-ctx.Done():
+				return
+			}
+		}
+	}()
+
+	return ch, nil
+}
+
+// logsStreamWithServiceMetadata wraps a container logs stream and enriches each log entry with service metadata.
+func logsStreamWithServiceMetadata(
+	stream <-chan api.ContainerLogEntry, metadata api.ServiceLogEntryMetadata,
+) <-chan api.ServiceLogEntry {
+	out := make(chan api.ServiceLogEntry)
+
+	go func() {
+		for entry := range stream {
+			out <- api.ServiceLogEntry{
+				Metadata:          metadata,
+				ContainerLogEntry: entry,
+			}
+		}
+		close(out)
+	}()
+
+	return out
+}

pkg/client/machine.go

@@ -2,7 +2,8 @@ package client
 
 import (
 	"context"
-	"slices"
+	"fmt"
+	"strings"
 
 	"github.com/psviderski/uncloud/internal/machine/api/pb"
 	"github.com/psviderski/uncloud/pkg/api"
@@ -32,17 +33,39 @@ func (cli *Client) ListMachines(ctx context.Context, filter *api.MachineFilter)
 	if err != nil {
 		return nil, err
 	}
+	machines := api.MachineMembersList(resp.Machines)
 
-	machines := resp.Machines
+	if filter == nil {
+		return machines, nil
+	}
 
-	if filter != nil {
-		var matchedMachines api.MachineMembersList
-		for _, m := range machines {
-			if MachineMatchesFilter(m, filter) {
-				matchedMachines = append(matchedMachines, m)
+	// Apply the filter.
+	if len(filter.NamesOrIDs) > 0 {
+		var matched api.MachineMembersList
+		var notFound []string
+
+		for _, nameOrID := range filter.NamesOrIDs {
+			if m := machines.FindByNameOrID(nameOrID); m != nil {
+				matched = append(matched, m)
+			} else {
+				notFound = append(notFound, nameOrID)
 			}
 		}
-		machines = matchedMachines
+		machines = matched
+
+		if len(notFound) > 0 {
+			return nil, fmt.Errorf("machines not found: %s", strings.Join(notFound, ", "))
+		}
+	}
+
+	if filter.Available {
+		var available api.MachineMembersList
+		for _, m := range machines {
+			if m.State != pb.MachineMember_DOWN {
+				available = append(available, m)
+			}
+		}
+		machines = available
 	}
 
 	return machines, nil
@@ -76,23 +99,3 @@ func (cli *Client) RenameMachine(ctx context.Context, nameOrID, newName string)
 
 	return cli.UpdateMachine(ctx, req)
 }
-
-func MachineMatchesFilter(machine *pb.MachineMember, filter *api.MachineFilter) bool {
-	if filter == nil {
-		return true
-	}
-
-	if filter.Available && machine.State == pb.MachineMember_DOWN {
-		return false
-	}
-
-	if len(filter.NamesOrIDs) > 0 {
-		if !slices.ContainsFunc(filter.NamesOrIDs, func(nameOrID string) bool {
-			return machine.Machine.Id == nameOrID || machine.Machine.Name == nameOrID
-		}) {
-			return false
-		}
-	}
-
-	return true
-}

pkg/client/service.go

@@ -273,6 +273,80 @@ func (cli *Client) RemoveService(ctx context.Context, id string) error {
 	return err
 }
 
+// StopService stops all containers on all machines that belong to the specified service.
+// The id parameter can be either a service ID or name.
+func (cli *Client) StopService(ctx context.Context, id string, opts container.StopOptions) error {
+	svc, err := cli.InspectService(ctx, id)
+	if err != nil {
+		return err
+	}
+
+	wg := sync.WaitGroup{}
+	errCh := make(chan error)
+
+	// Stop all containers on all machines that belong to the service.
+	for _, mc := range svc.Containers {
+		wg.Add(1)
+
+		go func() {
+			defer wg.Done()
+
+			err := cli.StopContainer(ctx, svc.ID, mc.Container.ID, opts)
+			if err != nil {
+				errCh <- fmt.Errorf("stop container '%s': %w", mc.Container.ID, err)
+			}
+		}()
+	}
+
+	go func() {
+		wg.Wait()
+		close(errCh)
+	}()
+
+	err = nil
+	for e := range errCh {
+		err = errors.Join(err, e)
+	}
+	return err
+}
+
+// StartService starts all containers on all machines that belong to the specified service.
+// The id parameter can be either a service ID or name.
+func (cli *Client) StartService(ctx context.Context, id string) error {
+	svc, err := cli.InspectService(ctx, id)
+	if err != nil {
+		return err
+	}
+
+	wg := sync.WaitGroup{}
+	errCh := make(chan error)
+
+	// Start all containers on all machines that belong to the service.
+	for _, mc := range svc.Containers {
+		wg.Add(1)
+
+		go func() {
+			defer wg.Done()
+
+			err := cli.StartContainer(ctx, svc.ID, mc.Container.ID)
+			if err != nil {
+				errCh <- fmt.Errorf("start container '%s': %w", mc.Container.ID, err)
+			}
+		}()
+	}
+
+	go func() {
+		wg.Wait()
+		close(errCh)
+	}()
+
+	err = nil
+	for e := range errCh {
+		err = errors.Join(err, e)
+	}
+	return err
+}
+
 // ListServices returns a list of all services and their containers.
 func (cli *Client) ListServices(ctx context.Context) ([]api.Service, error) {
 	machines, err := cli.ListMachines(ctx, nil)

pkg/client/volume.go

@@ -55,7 +55,7 @@ func (cli *Client) ListVolumes(ctx context.Context, filter *api.VolumeFilter) ([
 		proxyMachines = filter.Machines
 	}
 
-	listCtx, machines, err := api.ProxyMachinesContext(ctx, cli, proxyMachines)
+	listCtx, machines, err := cli.ProxyMachinesContext(ctx, proxyMachines)
 	if err != nil {
 		return nil, fmt.Errorf("create request context to broadcast to all machines: %w", err)
 	}

scripts/install.sh

@@ -2,6 +2,10 @@
 
 set -euo pipefail
 
+# If set to 'true', only install the packages and dependencies, without running, reloading, or
+# restarting services or systemd.
+INSTALL_ONLY=${INSTALL_ONLY:-false}
+
 INSTALL_BIN_DIR=${INSTALL_BIN_DIR:-/usr/local/bin}
 INSTALL_SYSTEMD_DIR=${INSTALL_SYSTEMD_DIR:-/etc/systemd/system}
 
@@ -57,7 +61,7 @@ verify_system() {
 Your system architecture ($arch) is not supported."
   fi
 
-  if [[ ! -d /run/systemd/system ]]; then
+  if [[ ! -d /run/systemd/system && "${INSTALL_ONLY}" != "true" ]]; then
       error "Cannot find systemd to use as a service manager for the Uncloud machine daemon. \
 Uncloud supports only systemd-based Linux systems for now."
   fi
@@ -66,9 +70,14 @@ Uncloud supports only systemd-based Linux systems for now."
 install_docker() {
     if command_exists dockerd; then
         log "✓ Docker is already installed."
+        DOCKER_ALREADY_INSTALLED=true
+
+        if [[ "${INSTALL_ONLY}" == "true" ]]; then
+            return
+        fi
+
         docker version
 
-        DOCKER_ALREADY_INSTALLED=true
         # Check if the installed Docker configured to use the containerd image store.
         local driver_status
         driver_status=$(docker info -f '{{ .DriverStatus }}' 2>/dev/null)
@@ -93,7 +102,9 @@ install_docker() {
     log "⏳ Configuring Docker daemon (${DOCKER_DAEMON_CONFIG_FILE}) to optimise it for Uncloud..."
     echo "${DOCKER_DAEMON_CONFIG}" > "${DOCKER_DAEMON_CONFIG_FILE}"
 
-    systemctl restart docker
+    if [[ "${INSTALL_ONLY}" != "true" ]]; then
+        systemctl restart docker
+    fi
 
     log "✓ Docker installed and configured successfully."
 }
@@ -215,8 +226,11 @@ WantedBy=multi-user.target
 EOF
     log "✓ Systemd unit file created: ${uncloud_service_path}"
 
-    # Reload systemd to recognize the new or updated unit file.
-    systemctl daemon-reload
+
+    if [[ "${INSTALL_ONLY}" != "true" ]]; then
+        # Reload systemd to recognize the new or updated unit file.
+        systemctl daemon-reload
+    fi
     systemctl enable uncloud.service
 }
 
@@ -284,11 +298,17 @@ RestrictAddressFamilies=AF_INET AF_INET6 AF_UNIX
 EOF
     log "✓ Systemd unit file created: ${corrosion_service_path}"
 
-    # Reload systemd to recognize the new unit file
-    systemctl daemon-reload
+    if [[ "${INSTALL_ONLY}" != "true" ]]; then
+        # Reload systemd to recognize the new unit file
+        systemctl daemon-reload
+    fi
 }
 
 start_uncloud() {
+    if [[ "${INSTALL_ONLY}" == "true" ]]; then
+        return
+    fi
+
     log "⏳ Starting Uncloud machine daemon (uncloud.service)..."
     systemctl restart uncloud.service
     log "✓ Uncloud machine daemon started."

scripts/install_cli.sh

@@ -20,12 +20,8 @@ print_manual_install() {
 }
 
 fetch_latest_version() {
-    api_url="https://api.github.com/repos/${GITHUB_REPO}/releases/latest"
-    auth_header=""
-    if [ -n "$GITHUB_TOKEN" ]; then
-        auth_header="Authorization: Bearer $GITHUB_TOKEN"
-    fi
-    VERSION=$(curl -fsSL -H "$auth_header" "$api_url" | grep -o '"tag_name": "[^"]*' | cut -d'"' -f4)
+    latest_url="https://github.com/${GITHUB_REPO}/releases/latest"
+    VERSION=$(curl -fsSLI -o /dev/null -w '%{url_effective}' "$latest_url" | grep -o 'tag/[^/]*$' | cut -d'/' -f2)
     if [ -z "$VERSION" ]; then
         echo "Failed to fetch the latest version from GitHub."
         print_manual_install

test/e2e/compose_deploy_test.go

@@ -533,4 +533,29 @@ func TestComposeDeployment(t *testing.T) {
 		assert.ElementsMatch(t, serviceMachines.ToSlice(), []string{c.Machines[0].ID, c.Machines[2].ID},
 			"Service containers should be on machines 1 and 3 from comma-separated list")
 	})
+
+	// Catches regression: https://github.com/psviderski/uncloud/issues/176
+	t.Run("plan new deployment with volumes and recreate strategy", func(t *testing.T) {
+		t.Parallel()
+
+		project, err := compose.LoadProjectFromContent(ctx, `
+services:
+  redis:
+    image: redis:7-alpine
+    volumes:
+      - redis-data:/data
+volumes:
+  redis-data:
+`)
+		require.NoError(t, err)
+
+		deployment, err := compose.NewDeploymentWithStrategy(ctx, cli, project,
+			&deploy.RollingStrategy{ForceRecreate: true})
+		require.NoError(t, err)
+
+		plan, err := deployment.Plan(ctx)
+		require.NoError(t, err)
+
+		assert.Len(t, plan.Operations, 2, "Expected 1 volume creation and 1 service to deploy")
+	})
 }

test/e2e/dns_test.go

@@ -2,14 +2,11 @@ package e2e
 
 import (
 	"context"
-	"fmt"
 	"regexp"
 	"strings"
 	"testing"
 	"time"
 
-	"github.com/docker/docker/api/types/container"
-	dockerclient "github.com/docker/docker/client"
 	"github.com/psviderski/uncloud/internal/ucind"
 	"github.com/psviderski/uncloud/pkg/api"
 	"github.com/stretchr/testify/assert"
@@ -68,94 +65,52 @@ func TestInternalDNS(t *testing.T) {
 		return true
 	}, 30*time.Second, 1*time.Second, "Service should be deployed and running on all machines")
 
-	// Get Docker client to access ucind machine containers (shared across all DNS tests)
-	dockerCli, err := dockerclient.NewClientWithOpts(dockerclient.FromEnv, dockerclient.WithAPIVersionNegotiation())
+	// Deploy a single "network-multitool" service to be used for all DNS queries
+	queryServiceName := "dns-query-service"
+	t.Cleanup(func() {
+		err := cli.RemoveService(ctx, queryServiceName)
+		if err != nil && !strings.Contains(err.Error(), "not found") {
+			require.NoError(t, err)
+		}
+	})
+	querySvcSpec := api.ServiceSpec{
+		Name:     queryServiceName,
+		Mode:     api.ServiceModeReplicated,
+		Replicas: 1,
+		Placement: api.Placement{
+			Machines: []string{c.Machines[0].Name},
+		},
+		Container: api.ContainerSpec{
+			Image:   "wbitt/network-multitool",
+			Command: []string{"sleep", "infinity"},
+		},
+	}
+	_, err = cli.RunService(ctx, querySvcSpec)
 	require.NoError(t, err)
 
-	// Helper function to create DNS query service spec
-	createDNSQuerySpec := func(name, dnsQuery, outputFile string) api.ServiceSpec {
-		return api.ServiceSpec{
-			Name: name,
-			Mode: api.ServiceModeReplicated,
-			Container: api.ContainerSpec{
-				Image: "wbitt/network-multitool",
-				Command: []string{"sh", "-c", fmt.Sprintf("nslookup %s > %s 2>&1 && echo 'DNS query completed' && sleep infinity",
-					dnsQuery, outputFile)},
-				VolumeMounts: []api.VolumeMount{
-					{
-						VolumeName:    "host-tmp",
-						ContainerPath: "/tmp",
-					},
-				},
-			},
-			Volumes: []api.VolumeSpec{
-				{
-					Name: "host-tmp",
-					Type: "bind",
-					BindOptions: &api.BindOptions{
-						HostPath: "/tmp",
-					},
-				},
-			},
-			Replicas: 1,
+	// Wait for the query service to be deployed
+	var querySvc api.Service
+	require.Eventually(t, func() bool {
+		querySvc, err = cli.InspectService(ctx, queryServiceName)
+		if err != nil {
+			return false
 		}
-	}
-
-	// Helper function to run DNS query and get results
-	runDNSQuery := func(t *testing.T, serviceName, dnsQuery, outputFile string) string {
-		spec := createDNSQuerySpec(serviceName, dnsQuery, outputFile)
-
-		t.Cleanup(func() {
-			cli.RemoveService(ctx, spec.Name)
-		})
-
-		// Run the DNS query service
-		resp, err := cli.RunService(ctx, spec)
-		require.NoError(t, err)
-
-		// Wait for the DNS query to complete
-		var querySvc api.Service
-		require.Eventually(t, func() bool {
-			querySvc, err = cli.InspectService(ctx, resp.ID)
-			if err != nil || len(querySvc.Containers) == 0 {
-				return false
-			}
-			return querySvc.Containers[0].Container.State.Status == "running"
-		}, 30*time.Second, 1*time.Second, "DNS query container should be running")
-
-		// Give some time for DNS query to complete and write to file
-		time.Sleep(1 * time.Second)
-
-		// Find which ucind machine the DNS query container is running on
-		ctr := querySvc.Containers[0]
-		machineID := ctr.MachineID
-
-		// Get the ucind machine container name for this machine
-		var ucindMachineName string
-		for _, machine := range c.Machines {
-			if machine.ID == machineID {
-				ucindMachineName = machine.ContainerName
-				break
-			}
+		// Should have 1 container and it should be running
+		if len(querySvc.Containers) != 1 {
+			return false
 		}
-		require.NotEmpty(t, ucindMachineName, "Should find ucind machine container name")
+		return querySvc.Containers[0].Container.State.Status == "running"
+	}, 30*time.Second, 1*time.Second, "Query service should be deployed and running")
+	queryContainer := querySvc.Containers[0]
 
-		// Read the DNS results from the ucind machine's filesystem
-		exec, err := dockerCli.ContainerExecCreate(ctx, ucindMachineName, container.ExecOptions{
-			Cmd:          []string{"cat", outputFile},
-			AttachStdout: true,
-			AttachStderr: true,
-		})
+	// Run nslookup for given query
+	runNslookup := func(t *testing.T, dnsQuery string) string {
+		dnsOutput, err := execInContainerAndReadOutput(
+			t, ctx, cli, queryServiceName, queryContainer.Container.ID,
+			[]string{"nslookup", dnsQuery},
+		)
 		require.NoError(t, err)
-
-		resp2, err := dockerCli.ContainerExecAttach(ctx, exec.ID, container.ExecAttachOptions{})
-		require.NoError(t, err)
-		defer resp2.Close()
-
-		// Read the DNS output
-		buf := make([]byte, 4096)
-		n, _ := resp2.Reader.Read(buf)
-		return string(buf[:n])
+		return dnsOutput
 	}
 
 	// Helper function to verify DNS output doesn't contain errors
@@ -166,7 +121,7 @@ func TestInternalDNS(t *testing.T) {
 	}
 
 	t.Run("service name resolves to all container IPs", func(t *testing.T) {
-		dnsOutput := runDNSQuery(t, "dns-test-service-name", serviceName+".internal", "/tmp/dns_result.txt")
+		dnsOutput := runNslookup(t, serviceName+".internal")
 		t.Logf("DNS query output:\n%s", dnsOutput)
 
 		// Verify that all service container IPs are in the DNS response
@@ -181,15 +136,14 @@ func TestInternalDNS(t *testing.T) {
 
 	t.Run("machine-specific service DNS lookups", func(t *testing.T) {
 		// Test the new <machine-id>.m.<service-name>.internal DNS feature
-		for i, targetContainer := range svc.Containers {
+		for _, targetContainer := range svc.Containers {
 			targetMachineID := targetContainer.MachineID
 			targetContainerIP := targetContainer.Container.UncloudNetworkIP().String()
 
 			// Construct the machine-specific DNS name
 			machineSpecificDNS := targetMachineID + ".m." + serviceName + ".internal"
-			outputFile := fmt.Sprintf("/tmp/dns_result_machine_%d.txt", i)
 
-			dnsOutput := runDNSQuery(t, fmt.Sprintf("dns-test-machine-%d", i), machineSpecificDNS, outputFile)
+			dnsOutput := runNslookup(t, machineSpecificDNS)
 			t.Logf("Machine-specific DNS query output for %s:\n%s", machineSpecificDNS, dnsOutput)
 
 			// Verify that the specific container IP is returned
@@ -213,7 +167,7 @@ func TestInternalDNS(t *testing.T) {
 
 	t.Run("service ID DNS lookup", func(t *testing.T) {
 		// Test that service ID also resolves (existing functionality)
-		dnsOutput := runDNSQuery(t, "dns-test-service-id", svc.ID+".internal", "/tmp/dns_result_service_id.txt")
+		dnsOutput := runNslookup(t, svc.ID+".internal")
 		t.Logf("Service ID DNS query output:\n%s", dnsOutput)
 
 		// Verify that all service container IPs are in the DNS response
@@ -227,36 +181,35 @@ func TestInternalDNS(t *testing.T) {
 	})
 
 	t.Run("nearest mode prioritizes local subnet IPs", func(t *testing.T) {
-		// Test the "nearest" mode which should sort local subnet IPs first
-		dnsOutput := runDNSQuery(t, "dns-test-nearest", "nearest."+serviceName+".internal", "/tmp/dns_result_nearest.txt")
-		t.Logf("Nearest mode DNS query output:\n%s", dnsOutput)
-
-		// Find which machine ran the query
-		querySvc, err := cli.InspectService(ctx, "dns-test-nearest")
-		require.NoError(t, err)
-		require.NotEmpty(t, querySvc.Containers, "Query service should have a container")
-		queryMachineID := querySvc.Containers[0].MachineID
-
-		// Find the local container IP (on the same machine as the query)
+		// Find the service container on the same machine as the query container.
 		var localIP string
 		for _, ctr := range svc.Containers {
-			if ctr.MachineID == queryMachineID {
+			if ctr.MachineID == queryContainer.MachineID {
 				localIP = ctr.Container.UncloudNetworkIP().String()
 				break
 			}
 		}
-		require.NotEmpty(t, localIP, "Should find local container IP on query machine %s", queryMachineID)
+		require.NotEmpty(t, localIP, "Should find local container IP on query machine %s", queryContainer.MachineID)
 
-		// Extract the first IP address from the DNS output using regex
-		// Pattern matches: "Name: nearest.test-dns-service.internal" followed by "Address: X.X.X.X"
+		// We will extract the first IP address from the DNS output using a regex.
+		// Pattern matches "Name: nearest.test-dns-service.internal" followed by "Address: X.X.X.X".
 		re := regexp.MustCompile(`(?m)Name:\s+[\w\.\-]+\s+Address:\s+([\d\.]+)`)
-		matches := re.FindStringSubmatch(dnsOutput)
-		require.Len(t, matches, 2, "Should find Name's Address in DNS output")
-		firstIP := matches[1]
-		assert.Equal(t, localIP, firstIP,
-			"Nearest mode should return local subnet IP first (query machine: %s, local IP: %s, first DNS result: %s)",
-			queryMachineID, localIP, firstIP)
 
-		assertNoDNSErrors(t, dnsOutput)
+		// Test the "nearest" mode which should sort local subnet IPs first.
+		// The default behavior randomizes the order, so run it a few times
+		// to reduce the chance we're just getting lucky with the order.
+		for range 5 {
+			dnsOutput := runNslookup(t, "nearest."+serviceName+".internal")
+			t.Logf("Nearest mode DNS query output:\n%s", dnsOutput)
+
+			matches := re.FindStringSubmatch(dnsOutput)
+			require.Len(t, matches, 2, "Should find Name's Address in DNS output")
+			firstIP := matches[1]
+			assert.Equal(t, localIP, firstIP,
+				"Nearest mode should return local subnet IP first (query machine: %s, local IP: %s, first DNS result: %s)",
+				queryContainer.MachineID, localIP, firstIP)
+
+			assertNoDNSErrors(t, dnsOutput)
+		}
 	})
 }

test/e2e/exec_test.go

@@ -29,10 +29,7 @@ func deployTestService(t *testing.T, ctx context.Context, cli *client.Client, na
 	}
 
 	deployment := cli.NewDeployment(spec, nil)
-	err := deployment.Validate(ctx)
-	require.NoError(t, err)
-
-	_, err = deployment.Run(ctx)
+	_, err := deployment.Run(ctx)
 	require.NoError(t, err)
 
 	// Wait for all replicas to be running
@@ -41,9 +38,8 @@ func deployTestService(t *testing.T, ctx context.Context, cli *client.Client, na
 		if err != nil {
 			return false
 		}
-		if uint(len(service.Containers)) != replicas {
-			return false
-		}
+		assertServiceMatchesSpec(t, service, spec)
+
 		for _, ctr := range service.Containers {
 			if ctr.Container.State.Status != "running" {
 				return false

test/e2e/service_test.go

@@ -3,6 +3,7 @@ package e2e
 import (
 	"context"
 	"errors"
+	"fmt"
 	"net/netip"
 	"slices"
 	"strings"
@@ -1682,4 +1683,115 @@ func TestServiceLifecycle(t *testing.T) {
 		assert.Equal(t, api.ServiceModeGlobal, svc.Mode)
 		assert.Len(t, svc.Containers, 3, "expected 1 container on each machine")
 	})
+
+	t.Run("service logs", func(t *testing.T) {
+		t.Parallel()
+
+		name := "test-service-logs"
+		t.Cleanup(func() {
+			err := cli.RemoveService(ctx, name)
+			if !errors.Is(err, api.ErrNotFound) {
+				require.NoError(t, err)
+			}
+		})
+
+		spec := api.ServiceSpec{
+			Name: name,
+			Container: api.ContainerSpec{
+				Image: "busybox:1.37.0-musl",
+				Command: []string{
+					"sh",
+					"-c",
+					"trap 'exit 0' TERM INT; " +
+						"echo \"Hello from $HOSTNAME\"; sleep 0.5;" +
+						"echo \"Hello stderr from $HOSTNAME\" >&2; " +
+						"while true; do sleep 1; done",
+				},
+			},
+			Replicas: 3,
+		}
+		deployment := cli.NewDeployment(spec, nil)
+		_, err := deployment.Run(ctx)
+		require.NoError(t, err)
+
+		svc, err := cli.InspectService(ctx, name)
+		require.NoError(t, err)
+		assertServiceMatchesSpec(t, svc, spec)
+
+		machines := serviceMachines(svc)
+		assert.Len(t, machines.ToSlice(), 3, "should have containers on all 3 machines")
+
+		time.Sleep(1 * time.Second) // Wait for containers to produce the logs.
+
+		logsSvc, stream, err := cli.ServiceLogs(ctx, name, api.ServiceLogsOptions{})
+		require.NoError(t, err)
+		assertServiceMatchesSpec(t, logsSvc, spec)
+
+		collectLogs := func(stream <-chan api.ServiceLogEntry) []string {
+			var entries []string
+			timeout := time.After(5 * time.Second)
+
+			for {
+				select {
+				case entry, ok := <-stream:
+					if !ok {
+						return entries
+					}
+
+					require.NoError(t, entry.Err)
+					if entry.Stream == api.LogStreamStdout || entry.Stream == api.LogStreamStderr {
+						entries = append(entries, string(entry.Message))
+					}
+				case <-timeout:
+					require.FailNow(t, "timed out waiting for logs")
+				}
+			}
+		}
+
+		logs := collectLogs(stream)
+		require.Len(t, logs, 6, "should have 6 log entries (1 stdout and 1 stderr from each of 3 replicas)")
+
+		for _, ctr := range svc.Containers {
+			stdoutMsg := fmt.Sprintf("Hello from %s\n", ctr.Container.Name)
+			stderrMsg := fmt.Sprintf("Hello stderr from %s\n", ctr.Container.Name)
+
+			assert.Contains(t, logs, stdoutMsg)
+			assert.Contains(t, logs, stderrMsg)
+
+			outIdx := slices.Index(logs, stdoutMsg)
+			errIdx := slices.Index(logs, stderrMsg)
+			assert.Less(t, outIdx, errIdx, "stdout log should appear before stderr log")
+		}
+
+		// Test filter logs by machine ID.
+		machineID := c.Machines[0].ID
+		_, machineStream, err := cli.ServiceLogs(ctx, name, api.ServiceLogsOptions{
+			Machines: []string{machineID},
+		})
+		require.NoError(t, err)
+
+		machineLogs := collectLogs(machineStream)
+		require.Len(t, machineLogs, 2, "should have 2 log entries (1 stdout and 1 stderr) from the filtered machine")
+
+		ctrName := ""
+		for _, ctr := range svc.Containers {
+			if ctr.MachineID == machineID {
+				ctrName = ctr.Container.Name
+				break
+			}
+		}
+		require.NotEmpty(t, ctrName, "should have found container on the target machine")
+
+		stdoutMsg := fmt.Sprintf("Hello from %s\n", ctrName)
+		stderrMsg := fmt.Sprintf("Hello stderr from %s\n", ctrName)
+
+		assert.Equal(t, []string{stdoutMsg, stderrMsg}, machineLogs)
+
+		// Test non-existent machine filter returns error.
+		_, _, err = cli.ServiceLogs(ctx, name, api.ServiceLogsOptions{
+			Machines: []string{"non-existent-machine"},
+		})
+		require.Error(t, err)
+		assert.Contains(t, err.Error(), "machines not found")
+	})
 }

website/compose.yaml

@@ -10,6 +10,13 @@ services:
         - linux/amd64
         - linux/arm64
     user: nobody
-    x-ports:
-      - uncloud.run:8000/https
+    x-caddy: |
+      uncloud.run {
+          redir /discord https://discord.gg/eR35KQJhPu
+
+          reverse_proxy {{upstreams 8000}} {
+              import common_proxy
+          }
+          log
+      }
     scale: 2

website/docs/1-overview.md

@@ -69,7 +69,7 @@ Talos [KubeSpan](https://www.talos.dev/v1.10/talos-guides/network/kubespan/).
 
 ### Managed DNS service (optional)
 
-Uncloud can provide **managed DNS records** like `<service-name>.<cluster-id>.cluster.uncloud.run` for your public
+Uncloud can provide **managed DNS records** like `<service-name>.<cluster-id>.uncld.dev` for your public
 services through free [Uncloud DNS](https://github.com/psviderski/uncloud-dns) service. You can deploy a service and
 instantly access it from anywhere with a proper DNS name and HTTPS without any manual DNS configuration. This makes
 self-hosting much more accessible and simplifies the process of adding your own domain later.

website/docs/2-getting-started/2-deploy-demo-app.md

@@ -45,7 +45,7 @@ This command will:
 - Install the Uncloud daemon on your server
 - Create a Docker network for Uncloud-managed containers
 - Deploy [Caddy](https://caddyserver.com/) as your reverse proxy listening on host ports 80 and 443
-- Reserve a free `xxxxxx.cluster.uncloud.run` subdomain via the Uncloud managed DNS service and point it to your
+- Reserve a free `xxxxxx.uncld.dev` subdomain via the Uncloud managed DNS service and point it to your
   server's IP
 
 All in about a minute!
@@ -144,7 +144,7 @@ Cluster initialised with machine 'machine-dc3c' and saved as context 'default' i
 Current cluster context is now 'default'.
 Waiting for the machine to be ready...
 
-Reserved cluster domain: 7za6s7.cluster.uncloud.run
+Reserved cluster domain: 7za6s7.uncld.dev
 [+] Deploying service caddy 7/2
  ✔ Container caddy-d7uk on machine-dc3c     Started                                            6.1s
    ✔ Image caddy:2.10.0 on machine-dc3c       Pulled                                           3.7s
@@ -154,7 +154,7 @@ Updating cluster domain records in Uncloud DNS to point to machines running cadd
  ✔ Machine machine-dc3c (157.180.72.195)  Reachable                                            0.7s
 
 DNS records updated to use only the internet-reachable machines running caddy service:
-  *.7za6s7.cluster.uncloud.run  A → 157.180.72.195
+  *.7za6s7.uncld.dev  A → 157.180.72.195
 ```
 
 </details>
@@ -177,7 +177,7 @@ You'll see the progress of the deployment and the public URL where you can acces
    ✔ Image excalidraw/excalidraw on machine-dc3c  Pulled                    4.7s
 
 excalidraw endpoints:
- • https://excalidraw.7za6s7.cluster.uncloud.run → :80
+ • https://excalidraw.7za6s7.uncld.dev → :80
 ```
 
 ## Verify your deployment
@@ -209,7 +209,7 @@ uc ls
 ```
 NAME         MODE         REPLICAS   ENDPOINTS
 caddy        global       1
-excalidraw   replicated   1          https://excalidraw.7za6s7.cluster.uncloud.run → :80
+excalidraw   replicated   1          https://excalidraw.7za6s7.uncld.dev → :80
 ```
 
 You can see `caddy` service listed here. That's your reverse proxy, running as a regular Uncloud service.
@@ -266,15 +266,15 @@ We've successfully converted our deployment created with `uc run` to a Compose f
 
 ## Use your own domain
 
-Want to use your own domain, for example, `excalidraw.example.com` instead of `excalidraw.7za6s7.cluster.uncloud.run`?
+Want to use your own domain, for example, `excalidraw.example.com` instead of `excalidraw.7za6s7.uncld.dev`?
 
 Add a CNAME record `excalidraw.example.com` in your DNS provider (Cloudflare, Namecheap, etc.) pointing to
-`excalidraw.7za6s7.cluster.uncloud.run`. Alternatively, you can add an A record pointing to your server's IP.
+`excalidraw.7za6s7.uncld.dev`. Alternatively, you can add an A record pointing to your server's IP.
 
 :::info note
 
 These instructions set up your own domain **in addition to** the Uncloud managed DNS name
-`excalidraw.7za6s7.cluster.uncloud.run`.
+`excalidraw.7za6s7.uncld.dev`.
 
 If you want to avoid the managed service altogether, add `--no-dns` to your `uc machine init` command, and point an A
 DNS record to your server(s)'s IP(s).

website/docs/4-guides/1-deployments/1-deploy-app.md

@@ -62,7 +62,7 @@ This command looks for a Compose file in your current working directory and:
 5. **Deploys service containers** using the built images and latest configuration changes with zero-downtime rolling
    updates
 
-See the [uc deploy](../../9-cli-reference/uc_deploy.md) reference for all available options.
+See the [`uc deploy` ](../../9-cli-reference/uc_deploy.md) reference for all available options.
 
 Watch `uc deploy` building and deploying a demo app (18 seconds):
 
@@ -98,9 +98,10 @@ You can pass additional build arguments or override existing ones using the `--b
 uc deploy --build-arg BUILD_ENV=dev
 ```
 
-You can also use advanced features like build caches, secrets, and SSH access.
+You can also use advanced features like build caches or SSH access.
 See [Compose Build Specification](https://github.com/compose-spec/compose-spec/blob/main/build.md) for all supported
-attributes.
+attributes. Note that build [secrets](https://github.com/compose-spec/compose-spec/blob/main/build.md#secrets) are not
+supported though.
 
 :::info note
 
@@ -140,8 +141,8 @@ This generates tags like:
 `uc deploy` renders the image templates when it loads the Compose file and then uses the resulting names for the build
 and deploy stages.
 
-See the [Image tag template](../../8-compose-file-reference/2-image-tag-template.md) reference for all available template
-variables and functions.
+See the [Image tag template](../../8-compose-file-reference/2-image-tag-template.md) reference for all available
+template variables and functions.
 
 ### Separate build and deploy steps
 

website/docs/8-compose-file-reference/1-support-matrix.md

@@ -42,6 +42,8 @@ The following table shows the support status for main Compose features:
 | `replicas`         | ✅ Supported        | Number of container replicas                                                          |
 | `resources`        | ⚠️ Limited         | CPU, memory limits and device reservations                                            |
 | `restart_policy`   | ❌ Not supported    | Defaults to `unless-stopped`                                                          |
+| `rollback_config`  | ❌ Not supported    | See [#151](https://github.com/psviderski/uncloud/issues/151)                          |
+| `update_config`    | ❌ Not supported    | See [#151](https://github.com/psviderski/uncloud/issues/151)                          |
 | **Volumes**        |                    |                                                                                       |
 | Named volumes      | ✅ Supported        | Docker volumes                                                                        |
 | Bind mounts        | ✅ Supported        | Host path binding                                                                     |

website/docs/8-compose-file-reference/2-image-tag-template.md

@@ -7,7 +7,7 @@ Template syntax for tagging built images.
 When building service images as part of [`uc build`](../9-cli-reference/uc_build.md) or
 [`uc deploy`](../9-cli-reference/uc_deploy.md) commands, Uncloud automatically generates image tags based on the current
 Git repository state. You can customise the image name and tag format for the built images using the
-[Go template](https://pkg.go.dev/text/template) syntax environment variables.
+[Go template](https://pkg.go.dev/text/template) syntax and environment variables.
 
 ## Default template
 

website/docs/9-cli-reference/uc.md

@@ -6,7 +6,8 @@ A CLI tool for managing Uncloud resources such as machines, services, and volume
 
 ```
       --connect string          Connect to a remote cluster machine without using the Uncloud configuration file. [$UNCLOUD_CONNECT]
-                                Format: [ssh://]user@host[:port], ssh+cli://user@host[:port], or tcp://host:port
+                                Format: [ssh://]user@host[:port], ssh+cli://user@host[:port], tcp://host:port, or unix:///path/to/uncloud.sock
+  -c, --context string          Name of the cluster context to use (default is the current context). [$UNCLOUD_CONTEXT]
   -h, --help                    help for uc
       --uncloud-config string   Path to the Uncloud configuration file. [$UNCLOUD_CONFIG] (default "~/.config/uncloud/config.yaml")
 ```
@@ -22,11 +23,13 @@ A CLI tool for managing Uncloud resources such as machines, services, and volume
 * [uc image](uc_image.md)	 - Manage images on machines in the cluster.
 * [uc images](uc_images.md)	 - List images on machines in the cluster.
 * [uc inspect](uc_inspect.md)	 - Display detailed information on a service.
+* [uc logs](uc_logs.md)	 - View service logs.
 * [uc ls](uc_ls.md)	 - List services.
-* [uc machine](uc_machine.md)	 - Manage machines in an Uncloud cluster.
+* [uc machine](uc_machine.md)	 - Manage machines in the cluster.
+* [uc ps](uc_ps.md)	 - List all service containers in the cluster
 * [uc rm](uc_rm.md)	 - Remove one or more services.
 * [uc run](uc_run.md)	 - Run a service.
 * [uc scale](uc_scale.md)	 - Scale a replicated service by changing the number of replicas.
-* [uc service](uc_service.md)	 - Manage services in an Uncloud cluster.
-* [uc volume](uc_volume.md)	 - Manage volumes in an Uncloud cluster.
+* [uc service](uc_service.md)	 - Manage services in the cluster.
+* [uc volume](uc_volume.md)	 - Manage volumes in the cluster.
 

website/docs/9-cli-reference/uc_build.md

@@ -41,7 +41,6 @@ uc build [FLAGS] [SERVICE...] [flags]
       --build-arg stringArray   Set a build-time variable for services. Used in Dockerfiles that declare the variable with ARG.
                                 Can be specified multiple times. Format: --build-arg VAR=VALUE
       --check                   Check the build configuration for services without building them.
-  -c, --context string          Name of the cluster context. (default is the current context)
       --deps                    Also build services declared as dependencies of the selected services.
   -f, --file strings            One or more Compose files to build. (default compose.yaml)
   -h, --help                    help for build
@@ -59,7 +58,8 @@ uc build [FLAGS] [SERVICE...] [flags]
 
 ```
       --connect string          Connect to a remote cluster machine without using the Uncloud configuration file. [$UNCLOUD_CONNECT]
-                                Format: [ssh://]user@host[:port], ssh+cli://user@host[:port], or tcp://host:port
+                                Format: [ssh://]user@host[:port], ssh+cli://user@host[:port], tcp://host:port, or unix:///path/to/uncloud.sock
+  -c, --context string          Name of the cluster context to use (default is the current context). [$UNCLOUD_CONTEXT]
       --uncloud-config string   Path to the Uncloud configuration file. [$UNCLOUD_CONFIG] (default "~/.config/uncloud/config.yaml")
 ```
 

website/docs/9-cli-reference/uc_caddy.md

@@ -12,7 +12,8 @@ Manage Caddy reverse proxy service.
 
 ```
       --connect string          Connect to a remote cluster machine without using the Uncloud configuration file. [$UNCLOUD_CONNECT]
-                                Format: [ssh://]user@host[:port], ssh+cli://user@host[:port], or tcp://host:port
+                                Format: [ssh://]user@host[:port], ssh+cli://user@host[:port], tcp://host:port, or unix:///path/to/uncloud.sock
+  -c, --context string          Name of the cluster context to use (default is the current context). [$UNCLOUD_CONTEXT]
       --uncloud-config string   Path to the Uncloud configuration file. [$UNCLOUD_CONFIG] (default "~/.config/uncloud/config.yaml")
 ```
 

website/docs/9-cli-reference/uc_caddy_config.md

@@ -13,7 +13,6 @@ uc caddy config [flags]
 ## Options
 
 ```
-  -c, --context string   Name of the cluster context. (default is the current context)
   -h, --help             help for config
   -m, --machine string   Name or ID of the machine to get the configuration from. (default is connected machine)
       --no-color         Disable syntax highlighting for the output.
@@ -23,7 +22,8 @@ uc caddy config [flags]
 
 ```
       --connect string          Connect to a remote cluster machine without using the Uncloud configuration file. [$UNCLOUD_CONNECT]
-                                Format: [ssh://]user@host[:port], ssh+cli://user@host[:port], or tcp://host:port
+                                Format: [ssh://]user@host[:port], ssh+cli://user@host[:port], tcp://host:port, or unix:///path/to/uncloud.sock
+  -c, --context string          Name of the cluster context to use (default is the current context). [$UNCLOUD_CONTEXT]
       --uncloud-config string   Path to the Uncloud configuration file. [$UNCLOUD_CONFIG] (default "~/.config/uncloud/config.yaml")
 ```
 

website/docs/9-cli-reference/uc_caddy_deploy.md

@@ -15,7 +15,6 @@ uc caddy deploy [flags]
 
 ```
       --caddyfile string   Path to a custom global Caddy config (Caddyfile) that will be prepended to the auto-generated Caddy config.
-  -c, --context string     Name of the cluster context to deploy to. (default is the current context)
   -h, --help               help for deploy
       --image string       Caddy Docker image to deploy. (default caddy:LATEST_VERSION)
   -m, --machine strings    Machine names or IDs to deploy to. Can be specified multiple times or as a comma-separated list. (default is all machines)
@@ -25,7 +24,8 @@ uc caddy deploy [flags]
 
 ```
       --connect string          Connect to a remote cluster machine without using the Uncloud configuration file. [$UNCLOUD_CONNECT]
-                                Format: [ssh://]user@host[:port], ssh+cli://user@host[:port], or tcp://host:port
+                                Format: [ssh://]user@host[:port], ssh+cli://user@host[:port], tcp://host:port, or unix:///path/to/uncloud.sock
+  -c, --context string          Name of the cluster context to use (default is the current context). [$UNCLOUD_CONTEXT]
       --uncloud-config string   Path to the Uncloud configuration file. [$UNCLOUD_CONFIG] (default "~/.config/uncloud/config.yaml")
 ```
 

website/docs/9-cli-reference/uc_ctx.md

@@ -16,13 +16,15 @@ uc ctx [flags]
 
 ```
       --connect string          Connect to a remote cluster machine without using the Uncloud configuration file. [$UNCLOUD_CONNECT]
-                                Format: [ssh://]user@host[:port], ssh+cli://user@host[:port], or tcp://host:port
+                                Format: [ssh://]user@host[:port], ssh+cli://user@host[:port], tcp://host:port, or unix:///path/to/uncloud.sock
+  -c, --context string          Name of the cluster context to use (default is the current context). [$UNCLOUD_CONTEXT]
       --uncloud-config string   Path to the Uncloud configuration file. [$UNCLOUD_CONFIG] (default "~/.config/uncloud/config.yaml")
 ```
 
 ## See also
 
 * [uc](uc.md)	 - A CLI tool for managing Uncloud resources such as machines, services, and volumes.
+* [uc ctx connection](uc_ctx_connection.md)	 - Choose a new default connection for the current context.
 * [uc ctx ls](uc_ctx_ls.md)	 - List available cluster contexts.
 * [uc ctx use](uc_ctx_use.md)	 - Switch to a different cluster context.