kubernetes/kops · warning

error listing %q: %v

Error message

error listing %q: %v

What it means

Returned by logDumperNode.findFiles (pkg/dump/dumper.go:529) when ExecPiped of `sudo find <dir> -type f -print0` over the SSH client returns an error. The message names the directory that could not be listed. Callers (dump) translate this into the 'error reading /var/log' or 'error listing /etc/containerd' variants, and it is the root cause behind most per-collection listing failures.

Source

Thrown at pkg/dump/dumper.go:529

		rel := strings.TrimPrefix(f, "/etc/containerd/")
		dest := filepath.Join(n.dir, "containerd", rel)
		cmd := "sudo cat '" + strings.ReplaceAll(f, "'", `'\''`) + "'"
		if err := n.shellToFile(ctx, cmd, dest); err != nil {
			errors = append(errors, err)
		}
	}

	return errors
}

// findFiles lists files under the specified directory (recursively)
func (n *logDumperNode) findFiles(ctx context.Context, dir string) ([]string, error) {
	var stdout bytes.Buffer
	var stderr bytes.Buffer

	err := n.client.ExecPiped(ctx, "sudo find "+dir+" -type f -print0", &stdout, &stderr)
	if err != nil {
		return nil, fmt.Errorf("error listing %q: %v", dir, err)
	}

	paths := []string{}
	for _, b := range bytes.Split(stdout.Bytes(), []byte{0}) {
		if len(b) == 0 {
			// Likely the last value
			continue
		}
		paths = append(paths, string(b))
	}
	return paths, nil
}

// listSystemdUnits returns the list of systemd units on the node
func (n *logDumperNode) listSystemdUnits(ctx context.Context) ([]string, error) {
	var stdout bytes.Buffer
	var stderr bytes.Buffer

View on GitHub (pinned to 4c8573c808)

Solutions

1. Reproduce locally: `ssh <node> sudo find <dir> -type f -print0`; the remote exit status/stderr shows the true cause.
2. Check the named directory exists on the node image; skip the listing when the runtime (docker vs containerd) doesn't use it.
3. Verify passwordless sudo (NOPASSWD) for the SSH user on the node.
4. Raise --node-dump-timeout if the wrapped error is context deadline exceeded.
5. Retry on transient SSH errors; check node stability (no concurrent replacement/reboot).

Example fix

// before
err := n.client.ExecPiped(ctx, "sudo find "+dir+" -type f -print0", &stdout, &stderr)
if err != nil {
    return nil, fmt.Errorf("error listing %q: %v", dir, err)
}
// after
cmd := "if [ -d " + quoteShell(dir) + " ]; then sudo find " + dir + " -type f -print0; fi"
err := n.client.ExecPiped(ctx, cmd, &stdout, &stderr)
if err != nil {
    return nil, fmt.Errorf("error listing %q: %w", dir, err)
}

// guard the remote find with an existence test so missing dirs don't error
cmd := fmt.Sprintf("test -d %q && sudo find %q -type f -print0 || true", dir, dir)
err := n.client.ExecPiped(ctx, cmd, &stdout, &stderr)

paths, err := n.findFiles(ctx, dir)
if err != nil {
    if errors.Is(err, context.DeadlineExceeded) {
        return nil, fmt.Errorf("find on %q timed out; increase node timeout", dir)
    }
    return nil, err
}

When it happens

Trigger: n.client.ExecPiped(ctx, "sudo find "+dir+" -type f -print0", &stdout, &stderr) fails: non-zero exit of find (missing directory, permission denied under sudo), dead SSH session, or ctx.Err() (nodeDumpTimeout expired) checked at ExecPiped entry.

Common situations: find run on a directory that doesn't exist on that node image (e.g. /etc/containerd on docker nodes); sudo requiring a password; node overloaded so the 1-minute dump timeout expires; SSH connection reset during long find on huge directory trees.

Related errors

• error reading /var/log: %v
• error listing systemd services: %v
• error listing /etc/containerd: %v
• error listing systemd units: %v
• error creating file %q: %v

AI-assisted analysis of kubernetes/kops@4c8573c808 (2026-09-05). Data as JSON: /api/errors/634527059ca0d9bc. Report an issue: GitHub.