What causes "failed to attach list exec for '%s': %w" in vxcontrol/pentagi?

Retry the whole ListContainerDir call once — the exec may have raced a container shutdown

failed to attach list exec for '%s': %w

Error message

failed to attach list exec for '%s': %w

What it means

After the list exec is created, ListContainerDir attaches to its stdio stream via ContainerExecAttach. This error wraps a failure of that attach call — the exec's output stream could not be connected. The exec may have been created but already finished, been removed, or the daemon connection dropped.

Source

Thrown at backend/pkg/docker/client.go:998

	// TTY GNU coreutils shell-quotes names and busybox wraps them in ANSI escapes,
	// so a readable file with a space / quote / non-ASCII byte would be mis-stat'd
	// and reported unreadable. `find -print0` emits literal bytes and is portable
	// (GNU + busybox); the NUL delimiter also survives names containing newlines.
	// No TTY: a TTY's onlcr would rewrite every \n in the stream to \r\n —
	// including a \n that is part of a filename — corrupting the name. Without a
	// TTY the exec stream is multiplexed and demuxed below.
	createResp, err := dc.ContainerExecCreate(ctx, containerID, client.ExecCreateOptions{
		Cmd:          []string{"find", dirPath, "-maxdepth", "1", "-mindepth", "1", "!", "-name", ".*", "-print0"},
		AttachStdout: true,
		AttachStderr: true,
	})
	if err != nil {
		return ContainerDirListing{}, fmt.Errorf("failed to create list exec for '%s': %w", dirPath, err)
	}

	resp, err := dc.ContainerExecAttach(ctx, createResp.ID, client.ExecAttachOptions{})
	if err != nil {
		return ContainerDirListing{}, fmt.Errorf("failed to attach list exec for '%s': %w", dirPath, err)
	}
	output, readErr := demuxExecStdout(resp.Reader, maxListStdoutBytes)
	resp.Close()
	if readErr != nil {
		return ContainerDirListing{}, fmt.Errorf("failed to read list output for '%s': %w", dirPath, readErr)
	}

	inspect, err := dc.ContainerExecInspect(ctx, createResp.ID)
	if err != nil {
		return ContainerDirListing{}, fmt.Errorf("failed to inspect list exec for '%s': %w", dirPath, err)
	}
	if inspect.ExitCode != 0 {
		return ContainerDirListing{}, fmt.Errorf("list command failed for '%s' with exit code %d: %s", dirPath, inspect.ExitCode, string(output))
	}

	entryPaths, truncated := parseFindEntries(output)

	stats, failures := statContainerEntries(ctx, entryPaths, containerListWorkers, func(ctx context.Context, entryPath string) (container.PathStat, error) {

View on GitHub (pinned to ea665308ba)

Solutions

Retry the whole ListContainerDir call once — the exec may have raced a container shutdown
Confirm the container was still running at the time of the call and re-run on a live container
Check Docker daemon health/socket connectivity (`docker ps`, DOCKER_HOST config)
If using a remote daemon, verify network stability and keep-alive settings

Defensive patterns

Strategy: retry

Validate before calling

state, err := dc.ContainerInspect(ctx, containerID)
if err != nil || !state.State.Running {
    return fmt.Errorf("container gone before attach")
}

Try / catch

listing, err := client.ListContainerDir(ctx, containerID, dir)
if err != nil && strings.Contains(err.Error(), "failed to attach list exec") {
    // transient: retry once with a fresh exec
    listing, err = client.ListContainerDir(ctx, containerID, dir)
}

Prevention

Avoid stopping/removing containers while introspection runs
Keep Docker daemon connections healthy; reconnect on idle timeouts
For remote daemons, use stable low-latency links
Wrap docker calls with bounded retry + backoff

When it happens

Trigger: ContainerExecAttach(ctx, createResp.ID, ...) fails because the exec instance no longer exists (container stopped between create and attach), the Docker daemon connection broke, or the API returned 404/500 for the exec ID.

Common situations: Race where the container is killed right after exec creation; Docker daemon restart or socket drop; network instability against a remote Docker (tcp://) endpoint; long-lived clients hitting an idle-connection timeout.

Related errors

AI-assisted analysis of vxcontrol/pentagi@ea665308ba (2026-09-01). Data as JSON: /api/errors/44746b0680d29b53. Report an issue: GitHub.