convertor: Retry on 5xx errors (#19)

**What this PR does / why we need it**: Also retry on 5xx errors.
Occasionally we get various gateway errors or just 500 errors, the retry
logic should apply to these too.

**Which issue(s) this PR fixes** *(optional, in `fixes #<issue number>(,
fixes #<issue_number>, ...)` format, will close the issue(s) when PR
gets merged)*:
Fixes #

**Please check the following list**:
- [ ] Does the affected code have corresponding tests, e.g. unit test,
E2E test?
- [ ]  Does this change require a documentation update?
- [ ] Does this introduce breaking changes that would require an
announcement or bumping the major version?
- [ ]  Do all new files have an appropriate license header?

<!-- If this is a security issue, please do not discuss on GitHub.
Please report any suspected or confirmed security issues directly to
https://github.com/containerd/accelerated-container-image/blob/main/MAINTAINERS.
-->

Author: gautam-rl

cmd/convertor/builder/builder_utils.go

@@ -35,6 +35,7 @@ import (
 	"github.com/containerd/containerd/v2/core/images"
 	"github.com/containerd/containerd/v2/core/remotes"
 	"github.com/containerd/containerd/v2/core/remotes/docker"
+	remoteerrors "github.com/containerd/containerd/v2/core/remotes/errors"
 	"github.com/containerd/containerd/v2/pkg/archive/compression"
 	"github.com/containerd/continuity"
 	"github.com/containerd/errdefs"
@@ -54,18 +55,33 @@ func isRetryableError(err error) bool {
 		return false
 	}
 
-	// Check for containerd docker error types
-	var dockerErr *docker.Error
-	if errors.As(err, &dockerErr) {
-		switch dockerErr.Code {
-		case docker.ErrorCodeTooManyRequests:
-			return true
-		case docker.ErrorCodeUnavailable:
+	// Registry may return multiple errors in one response (docker.Errors)
+	var regErrs docker.Errors
+	if errors.As(err, &regErrs) {
+		for _, e := range regErrs {
+			switch v := e.(type) {
+			case docker.ErrorCode:
+				if v == docker.ErrorCodeTooManyRequests || v == docker.ErrorCodeUnavailable || v == docker.ErrorCodeUnknown {
+					return true
+				}
+			case docker.Error:
+				if v.Code == docker.ErrorCodeTooManyRequests || v.Code == docker.ErrorCodeUnavailable || v.Code == docker.ErrorCodeUnknown {
+					return true
+				}
+			}
+		}
+	}
+
+	// Unexpected HTTP status codes (e.g., 429 or 5xx)
+	var us remoteerrors.ErrUnexpectedStatus
+	if errors.As(err, &us) {
+		if us.StatusCode == 429 || (us.StatusCode >= 500 && us.StatusCode <= 599) {
 			return true
-		default:
-			return false
 		}
-	} else if err, ok := err.(net.Error); ok && err.Timeout() {
+	}
+
+	// Network timeouts
+	if ne, ok := err.(net.Error); ok && ne.Timeout() {
 		return true
 	}
 

cmd/convertor/main.go

@@ -396,7 +396,7 @@ func init() {
 	rootCmd.Flags().IntVar(&concurrencyLimit, "concurrency-limit", 4, "the number of manifests that can be built at the same time, used for multi-arch images, 0 means no limit")
 	rootCmd.Flags().BoolVar(&disableSparse, "disable-sparse", false, "disable sparse file for overlaybd")
 	rootCmd.Flags().BoolVar(&referrer, "referrer", false, "push converted manifests with subject, note '--oci' will be enabled automatically if '--referrer' is set, cause the referrer must be in OCI format.")
-	rootCmd.Flags().IntVar(&retryCount, "retry-count", 5, "number of retries for registry upload operations when encountering 429 rate limiting")
+	rootCmd.Flags().IntVar(&retryCount, "retry-count", 5, "number of retries for registry upload operations when encountering 429 rate limiting or 5xx errors")
 
 	// tar import/export
 	rootCmd.Flags().StringVar(&importTar, "import-tar", "", "import image from tar file (OCI layout format)")