feat: service logs command to stream service logs (#196)

* chore: fix landing logo shadow

* feat: implement service logs command with colored output and strict ordering

* revert Makefile

* fix after rebase

* move logs command under services with root shortcut

* simplify server options for streaming logs and CLI flags

* refactor ContainerLogs grpc server

* update --tail flag

* minor docker.proto

* refactor client ServiceLogs and ContainerLogs

* move ProxyMachinesContext from api to client pkg

* minor refactor proto Stream

* add api/logs

* implement LogMerger

* fix LogMerger to correctly use semaphore

* increate inflish entries to 100 per stream

* send heartbeats

* refactor ContainerLogs to synchronise Send of entries and heartbeats to the steram

* minor logmerege

* remove ContainerName form ServiceLogEntryMetadata

* refactor ContainerLogs into docker.Service

* detect stalled container logs streams

* update logmerger tests

* fix comment in test

* refactor LogMerger with options

* uc logs: format one or multiple services

* make LogMerger emit heartbeats, emit entries <= watermark, rewrite tests

* update uc logs with new LogMerger

* go mod tidy

* fix after merge

---------

Co-authored-by: Evgenii Orlov <evgenii.orlov@semrush.com>
This commit is contained in:
Pasha Sviderski
2025-12-02 19:50:48 +10:00
committed by GitHub
co-authored by Evgenii Orlov
parent 234985b57d
commit 79dc05cb66
25 changed files with 3657 additions and 1925 deletions
+1 -2
View File
@@ -7,7 +7,6 @@ import (
"github.com/alecthomas/chroma/v2/quick"
"github.com/psviderski/uncloud/internal/cli"
"github.com/psviderski/uncloud/pkg/api"
"github.com/spf13/cobra"
)
@@ -46,7 +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
}
+1 -1
View File
@@ -55,7 +55,7 @@ func remove(ctx context.Context, uncli *cli.CLI, nameOrID string, opts removeOpt
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
}
+1 -1
View File
@@ -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(),
+2 -1
View File
@@ -108,9 +108,9 @@ func main() {
})
cmd.AddCommand(
NewBuildCommand(),
NewDeployCommand(),
NewDocsCommand(),
NewBuildCommand(),
NewImagesCommand(),
NewPsCommand(),
caddy.NewRootCommand(),
@@ -122,6 +122,7 @@ func main() {
service.NewExecCommand(),
service.NewInspectCommand(),
service.NewListCommand(),
service.NewLogsCommand(),
service.NewRmCommand(),
service.NewRunCommand(),
service.NewScaleCommand(),
+7 -5
View File
@@ -12,7 +12,6 @@ import (
"github.com/spf13/cobra"
"github.com/psviderski/uncloud/internal/cli"
"github.com/psviderski/uncloud/pkg/api"
"github.com/psviderski/uncloud/pkg/client"
)
@@ -46,12 +45,14 @@ This command provides a comprehensive overview of all running containers that ar
making it easy to see the distribution and status of containers across the cluster.`,
RunE: func(cmd *cobra.Command, args []string) error {
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 fmt.Errorf("invalid value for --sort: %q, must be one of '%s', '%s' or '%s'", opts.sortBy,
sortByService, sortByMachine, sortByHealth)
}
return runPs(cmd, opts)
},
}
cmd.Flags().StringVarP(&opts.sortBy, "sort", "s", sortByService, "Sort containers by 'service', 'machine' or 'health'")
cmd.Flags().StringVarP(&opts.sortBy, "sort", "s", sortByService,
"Sort containers by 'service', 'machine' or 'health'")
return cmd
}
@@ -176,7 +177,7 @@ func printContainers(containers []containerInfo) error {
}
func collectContainers(ctx context.Context, cli *client.Client) ([]containerInfo, error) {
listCtx, machines, err := api.ProxyMachinesContext(ctx, cli, nil)
listCtx, machines, err := cli.ProxyMachinesContext(ctx, nil)
if err != nil {
return nil, fmt.Errorf("proxy machines context: %w", err)
}
@@ -206,7 +207,8 @@ func collectContainers(ctx context.Context, cli *client.Client) ([]containerInfo
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))
client.PrintWarning(fmt.Sprintf("failed to list containers on machine %s: %s", machineName,
msc.Metadata.Error))
continue
}
+279
View File
@@ -0,0 +1,279 @@
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/spf13/cobra"
)
type logsOptions struct {
follow bool
tail string
since string
until string
}
func NewLogsCommand() *cobra.Command {
var options logsOptions
cmd := &cobra.Command{
Use: "logs SERVICE [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.",
Args: cobra.MinimumNArgs(1),
RunE: func(cmd *cobra.Command, args []string) error {
uncli := cmd.Context().Value("cli").(*cli.CLI)
return streamLogs(cmd.Context(), uncli, args, options)
},
}
cmd.Flags().BoolVarP(&options.follow, "follow", "f", false,
"Continually stream new logs.")
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.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().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.")
return cmd
}
func streamLogs(ctx context.Context, uncli *cli.CLI, serviceNames []string, opts logsOptions) error {
// 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,
}
// Collect log streams from all services.
machineIDsSet := mapset.NewSet[string]()
svcStreams := make([]<-chan api.ServiceLogEntry, 0, len(serviceNames))
for _, serviceName := range serviceNames {
// TODO: set Heartbeats in the opts.
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)
// 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
}
func newLogFormatter(machineNames, serviceNames []string) *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,
}
}
// formatTimestamp formats timestamp using local timezone.
func (f *logFormatter) formatTimestamp(t time.Time) string {
dimStyle := lipgloss.NewStyle().Faint(true)
t = t.In(time.Local)
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))
}
}
+3 -2
View File
@@ -8,15 +8,16 @@ 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(
NewExecCommand(),
NewInspectCommand(),
NewListCommand(),
NewLogsCommand(),
NewRmCommand(),
NewRunCommand(),
NewScaleCommand(),
NewExecCommand(),
)
return cmd
}
+1 -1
View File
@@ -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(),
+1 -1
View File
@@ -150,7 +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
+2 -2
View File
@@ -346,8 +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=
File diff suppressed because it is too large Load Diff
+62 -38
View File
@@ -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;
}
}
+73 -32
View File
@@ -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",
}
+92
View File
@@ -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()
+105
View File
@@ -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()
}
}
-41
View File
@@ -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 {
@@ -58,41 +55,3 @@ 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
}
+84
View File
@@ -0,0 +1,84 @@
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
}
// 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")
+11
View File
@@ -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{})
+39
View File
@@ -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
}
+1 -1
View File
@@ -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)
}
+310
View File
@@ -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
}
+383
View File
@@ -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)
}
+141
View File
@@ -0,0 +1,141 @@
package client
import (
"context"
"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, nil)
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 {
// Try to get machine name for ServiceLogEntry metadata and friendlier error message.
machineName := ctr.MachineID
m := machines.FindByNameOrID(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)
}
// 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,
}
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
}
+20 -20
View File
@@ -48,6 +48,26 @@ func (cli *Client) ListMachines(ctx context.Context, filter *api.MachineFilter)
return machines, nil
}
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
}
// UpdateMachine updates machine configuration in the cluster.
func (cli *Client) UpdateMachine(ctx context.Context, req *pb.UpdateMachineRequest) (*pb.MachineInfo, error) {
resp, err := cli.ClusterClient.UpdateMachine(ctx, req)
@@ -76,23 +96,3 @@ 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
}
+1 -1
View File
@@ -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)
}