add ListServices method and ls CLI command

cmd/uncloud/machine/list.go

@@ -32,13 +32,13 @@ func NewListCommand() *cobra.Command {
 }
 
 func runList(ctx context.Context, uncli *cli.CLI, clusterName string) error {
-	c, err := uncli.ConnectCluster(ctx, clusterName)
+	client, err := uncli.ConnectCluster(ctx, clusterName)
 	if err != nil {
 		return fmt.Errorf("connect to cluster: %w", err)
 	}
-	defer c.Close()
+	defer client.Close()
 
-	machines, err := c.ListMachines(ctx)
+	machines, err := client.ListMachines(ctx)
 	if err != nil {
 		return fmt.Errorf("list machines: %w", err)
 	}

cmd/uncloud/main.go

@@ -44,6 +44,7 @@ func main() {
 		machine.NewRootCommand(),
 		service.NewRootCommand(),
 		service.NewInspectCommand(),
+		service.NewListCommand(),
 		service.NewRunCommand(),
 	)
 	cobra.CheckErr(cmd.Execute())

cmd/uncloud/service/inspect.go

@@ -10,7 +10,7 @@ import (
 	"time"
 
 	"github.com/spf13/cobra"
-	client "uncloud/internal/cli"
+	"uncloud/internal/cli"
 )
 
 type inspectOptions struct {
@@ -25,7 +25,7 @@ func NewInspectCommand() *cobra.Command {
 		Short: "Display detailed information on a service.",
 		Args:  cobra.ExactArgs(1),
 		RunE: func(cmd *cobra.Command, args []string) error {
-			uncli := cmd.Context().Value("cli").(*client.CLI)
+			uncli := cmd.Context().Value("cli").(*cli.CLI)
 			opts.service = args[0]
 			return inspect(cmd.Context(), uncli, &opts)
 		},
@@ -37,19 +37,19 @@ func NewInspectCommand() *cobra.Command {
 	return cmd
 }
 
-func inspect(ctx context.Context, uncli *client.CLI, opts *inspectOptions) error {
+func inspect(ctx context.Context, uncli *cli.CLI, opts *inspectOptions) error {
-	cli, err := uncli.ConnectCluster(ctx, opts.cluster)
+	client, err := uncli.ConnectCluster(ctx, opts.cluster)
 	if err != nil {
 		return fmt.Errorf("connect to cluster: %w", err)
 	}
-	defer cli.Close()
+	defer client.Close()
 
-	svc, err := cli.InspectService(ctx, opts.service)
+	svc, err := client.InspectService(ctx, opts.service)
 	if err != nil {
 		return fmt.Errorf("inspect service: %w", err)
 	}
 
-	machines, err := cli.ListMachines(ctx)
+	machines, err := client.ListMachines(ctx)
 	if err != nil {
 		return fmt.Errorf("list machines: %w", err)
 	}

cmd/uncloud/service/list.go

@@ -0,0 +1,53 @@
+package service
+
+import (
+	"context"
+	"fmt"
+	"github.com/spf13/cobra"
+	"os"
+	"text/tabwriter"
+	"uncloud/internal/cli"
+)
+
+func NewListCommand() *cobra.Command {
+	var cluster string
+	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 runList(cmd.Context(), uncli, cluster)
+		},
+	}
+	cmd.Flags().StringVarP(
+		&cluster, "cluster", "c", "",
+		"Name of the cluster. (default is the current cluster)",
+	)
+	return cmd
+}
+
+func runList(ctx context.Context, uncli *cli.CLI, clusterName string) error {
+	client, err := uncli.ConnectCluster(ctx, clusterName)
+	if err != nil {
+		return fmt.Errorf("connect to cluster: %w", err)
+	}
+	defer client.Close()
+
+	services, err := client.ListServices(ctx)
+	if err != nil {
+		return fmt.Errorf("list services: %w", err)
+	}
+
+	// Print the list of services in a table format.
+	tw := tabwriter.NewWriter(os.Stdout, 0, 0, 3, ' ', 0)
+	if _, err = fmt.Fprintln(tw, "SERVICE ID\tNAME\tMODE\tREPLICAS"); err != nil {
+		return fmt.Errorf("write header: %w", err)
+	}
+	for _, s := range services {
+		if _, err = fmt.Fprintf(tw, "%s\t%s\t%s\t%d\n", s.ID, s.Name, s.Mode, len(s.Containers)); err != nil {
+			return fmt.Errorf("write row: %w", err)
+		}
+	}
+	return tw.Flush()
+}

cmd/uncloud/service/root.go

@@ -10,6 +10,7 @@ func NewRootCommand() *cobra.Command {
 		Short: "Manage services in an Uncloud cluster.",
 	}
 	cmd.AddCommand(
+		NewListCommand(),
 		NewRunCommand(),
 	)
 	return cmd

cmd/uncloud/service/run.go

@@ -68,11 +68,11 @@ func runRun(ctx context.Context, uncli *cli.CLI, opts runOptions) error {
 		return fmt.Errorf("invalid replication mode: %q", opts.mode)
 	}
 
-	c, err := uncli.ConnectCluster(ctx, opts.cluster)
+	client, err := uncli.ConnectCluster(ctx, opts.cluster)
 	if err != nil {
 		return fmt.Errorf("connect to cluster: %w", err)
 	}
-	defer c.Close()
+	defer client.Close()
 
 	spec := api.ServiceSpec{
 		Container: api.ContainerSpec{
@@ -82,7 +82,7 @@ func runRun(ctx context.Context, uncli *cli.CLI, opts runOptions) error {
 		Mode: opts.mode,
 		Name: opts.name,
 	}
-	resp, err := c.RunService(ctx, spec)
+	resp, err := client.RunService(ctx, spec)
 	if err != nil {
 		return fmt.Errorf("run service: %w", err)
 	}

internal/cli/client/service.go

@@ -437,3 +437,70 @@ func (cli *Client) RemoveService(ctx context.Context, id string) error {
 	}
 	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)
+	if err != nil {
+		return nil, fmt.Errorf("list machines: %w", err)
+	}
+
+	// Broadcast the container list request to all available machines.
+	md := metadata.New(nil)
+	for _, m := range machines {
+		if m.State == pb.MachineMember_UP || m.State == pb.MachineMember_SUSPECT {
+			machineIP, _ := m.Machine.Network.ManagementIp.ToAddr()
+			md.Append("machines", machineIP.String())
+		}
+		// TODO: warning about machines that are DOWN.
+	}
+	listCtx := metadata.NewOutgoingContext(ctx, md)
+
+	// List only uncloud-managed containers that belong to some service.
+	opts := container.ListOptions{
+		All: true,
+		Filters: filters.NewArgs(
+			filters.Arg("label", api.LabelServiceID),
+			filters.Arg("label", api.LabelManaged),
+		),
+	}
+	machineContainers, err := cli.ListContainers(listCtx, opts)
+	if err != nil {
+		return nil, fmt.Errorf("list containers: %w", err)
+	}
+
+	// TODO: optimise by extracting services from the list of all containers instead of inspecting each service.
+	//  Most of the code can be reused in both InspectService and ListServices.
+	servicesByID := make(map[string]api.Service)
+	for _, mc := range machineContainers {
+		if mc.Metadata != nil && mc.Metadata.Error != "" {
+			// TODO: return failed machines in the response.
+			fmt.Printf("WARNING: failed to list containers on machine '%s': %s\n",
+				mc.Metadata.Machine, mc.Metadata.Error)
+			continue
+		}
+
+		for _, c := range mc.Containers {
+			ctr := api.Container{Container: c}
+			if _, ok := servicesByID[ctr.ServiceID()]; ok {
+				continue
+			}
+
+			svc, err := cli.InspectService(ctx, ctr.ServiceID())
+			if err != nil {
+				if errors.Is(err, ErrNotFound) {
+					continue
+				}
+				return nil, fmt.Errorf("inspect service: %w", err)
+			}
+
+			servicesByID[ctr.ServiceID()] = svc
+		}
+	}
+
+	services := make([]api.Service, 0, len(servicesByID))
+	for _, svc := range servicesByID {
+		services = append(services, svc)
+	}
+	return services, nil
+}

test/e2e/service_test.go

@@ -56,6 +56,21 @@ func TestRunService(t *testing.T) {
 		assert.Equal(t, name, svc.Name)
 		assert.Equal(t, api.ServiceModeReplicated, svc.Mode)
 		assert.Len(t, svc.Containers, 1)
+
+		services, err := cli.ListServices(ctx)
+		require.NoError(t, err)
+
+		assert.GreaterOrEqual(t, len(services), 1)
+		found := false
+		for _, s := range services {
+			if s.ID == svc.ID {
+				assert.Equal(t, name, s.Name)
+				assert.Equal(t, api.ServiceModeReplicated, s.Mode)
+				assert.Len(t, s.Containers, 1)
+				found = true
+			}
+		}
+		assert.True(t, found)
 	})
 
 	t.Run("global mode", func(t *testing.T) {