feat: add backoff retry to docker version to avoid transient failure (#…

…19)

.golangci.yml

@@ -95,3 +95,7 @@ linters-settings:
   gofumpt:
     extra-rules: true
     lang-version: "1.21"
+  ireturn:
+    allow:
+      - error
+      - backoff.BackOff

go.mod

@@ -3,6 +3,7 @@ module github.com/thegeeklab/wp-docker-buildx
 go 1.21
 
 require (
+	github.com/cenkalti/backoff v2.2.1+incompatible
 	github.com/coreos/go-semver v0.3.1
 	github.com/rs/zerolog v1.30.0
 	github.com/thegeeklab/wp-plugin-go v1.0.1

go.sum

@@ -1,3 +1,5 @@
+github.com/cenkalti/backoff v2.2.1+incompatible h1:tNowT99t7UNflLxfYYSlKYsBpXdEet03Pg2g16Swow4=
+github.com/cenkalti/backoff v2.2.1+incompatible/go.mod h1:90ReRw6GdpyfrHakVjL/QHaoyV4aDUVVkXQJJJ3NXXM=
 github.com/coreos/go-semver v0.3.1 h1:yi21YpKnrx1gt5R+la8n5WgS0kCrsPp33dmEyHReZr4=
 github.com/coreos/go-semver v0.3.1/go.mod h1:irMmmIw/7yzSRPWryHsK7EYSg09caPQL03VsM8rvUec=
 github.com/coreos/go-systemd/v22 v22.5.0/go.mod h1:Y58oyj3AT4RCenI/lSvhwexgC+NSVTIJ3seZv2GcEnc=

plugin/impl.go

@@ -8,6 +8,7 @@ import (
 	"path/filepath"
 	"time"
 
+	"github.com/cenkalti/backoff"
 	"github.com/rs/zerolog/log"
 	"github.com/thegeeklab/wp-plugin-go/types"
 	"github.com/urfave/cli/v2"
@@ -16,7 +17,12 @@ import (
 
 var ErrTypeAssertionFailed = errors.New("type assertion failed")
 
-const strictFilePerm = 0o600
+const (
+	strictFilePerm               = 0o600
+	daemonBackoffMaxRetries      = 3
+	daemonBackoffInitialInterval = 2 * time.Second
+	daemonBackoffMultiplier      = 3.5
+)
 
 //nolint:revive
 func (p *Plugin) run(ctx context.Context) error {
@@ -149,16 +155,31 @@ func (p *Plugin) Execute() error {
 	// add proxy build args
 	addProxyBuildArgs(&p.Settings.Build)
 
-	var cmds []*execabs.Cmd
-	cmds = append(cmds, commandVersion()) // docker version
-	cmds = append(cmds, commandInfo())    // docker info
-	cmds = append(cmds, commandBuilder(p.Settings.Daemon))
-	cmds = append(cmds, commandBuildx())
+	versionCmd := commandVersion() // docker version
+
+	versionCmd.Stdout = os.Stdout
+	versionCmd.Stderr = os.Stderr
+	trace(versionCmd)
+
+	backoffOps := func() error {
+		return versionCmd.Run()
+	}
+	backoffLog := func(err error, delay time.Duration) {
+		log.Error().Msgf("failed to exec docker version: %v: retry in %s", err, delay.Truncate(time.Second))
+	}
+
+	if err := backoff.RetryNotify(backoffOps, newBackoff(daemonBackoffMaxRetries), backoffLog); err != nil {
+		return err
+	}
 
-	cmds = append(cmds, commandBuild(p.Settings.Build, p.Settings.Dryrun)) // docker build
+	var batchCmd []*execabs.Cmd
+	batchCmd = append(batchCmd, commandInfo()) // docker info
+	batchCmd = append(batchCmd, commandBuilder(p.Settings.Daemon))
+	batchCmd = append(batchCmd, commandBuildx())
+	batchCmd = append(batchCmd, commandBuild(p.Settings.Build, p.Settings.Dryrun)) // docker build
 
 	// execute all commands in batch mode.
-	for _, cmd := range cmds {
+	for _, cmd := range batchCmd {
 		cmd.Stdout = os.Stdout
 		cmd.Stderr = os.Stderr
 		trace(cmd)
@@ -189,3 +210,11 @@ func (p *Plugin) FlagsFromContext() error {
 
 	return nil
 }
+
+func newBackoff(maxRetries uint64) backoff.BackOff {
+	b := backoff.NewExponentialBackOff()
+	b.InitialInterval = daemonBackoffInitialInterval
+	b.Multiplier = daemonBackoffMultiplier
+
+	return backoff.WithMaxRetries(b, maxRetries)
+}