hashicorp/nomad · error

failed to create socket mount point: %w

Error message

failed to create socket mount point: %w

What it means

The same Prestart hook also creates h.socketMountPoint (0700) — the host directory where the CSI plugin's socket will be exposed and bind-mounted into the task via MountConfig.HostPath. Failure of os.MkdirAll (other than already-exists) aborts with 'failed to create socket mount point: %w', preventing plugin registration.

Source

Thrown at client/allocrunner/taskrunner/plugin_supervisor_hook.go:164

	return "csi_plugin_supervisor"
}

// Prestart is called before the task is started including after every
// restart. This requires that the mount paths for a plugin be
// idempotent, despite us not knowing the name of the plugin ahead of
// time.  Because of this, we use the allocid_taskname as the unique
// identifier for a plugin on the filesystem.
func (h *csiPluginSupervisorHook) Prestart(ctx context.Context,
	req *interfaces.TaskPrestartRequest, resp *interfaces.TaskPrestartResponse) error {

	// Create the mount directory that the container will access if it doesn't
	// already exist. Default to only nomad user access.
	if err := os.MkdirAll(h.mountPoint, 0700); err != nil && !os.IsExist(err) {
		return fmt.Errorf("failed to create mount point: %w", err)
	}

	if err := os.MkdirAll(h.socketMountPoint, 0700); err != nil && !os.IsExist(err) {
		return fmt.Errorf("failed to create socket mount point: %w", err)
	}

	// where the socket will be mounted
	configMount := &drivers.MountConfig{
		TaskPath:        h.task.CSIPluginConfig.MountDir,
		HostPath:        h.socketMountPoint,
		Readonly:        false,
		PropagationMode: "bidirectional",
	}
	// where the staging and per-alloc directories will be mounted
	volumeStagingMounts := &drivers.MountConfig{
		TaskPath:        h.task.CSIPluginConfig.StagePublishBaseDir,
		HostPath:        h.mountPoint,
		Readonly:        false,
		PropagationMode: "bidirectional",
	}
	// devices from the host
	devMount := &drivers.MountConfig{

View on GitHub (pinned to 482b49bf1a)

Solutions

1. Inspect and fix permissions on the socket mount path's parents so the nomad user can create directories
2. Delete any non-directory file occupying the path and restart the allocation
3. Check mount/SELinux status; ensure the filesystem is writable and security policy allows the nomad data dir
4. Restart the nomad client agent to recreate clean alloc hook state if the path is managed internally

Example fix

# before
$ file /var/lib/nomad/alloc/<id>/csi
 /var/lib/nomad/alloc/<id>/csi: ASCII text (stale file)
# after
$ sudo rm /var/lib/nomad/alloc/<id>/csi && nomad alloc stop <alloc>

// host-side precheck for socket mount point
sudo -u nomad test -w "$(dirname socketMountPoint)" && echo ok

// on failure, clean stale path and restart the allocation
if strings.Contains(err.Error(), "failed to create socket mount point") {
    removeStalePath(); nomadAllocStopRestart()
}

When it happens

Trigger: os.MkdirAll(h.socketMountPoint, 0700) fails non-IsExist — permission denied, read-only FS, path component is a regular file, container runtime/SELinux restrictions on the socket path

Common situations: Stale file at the socket mount path from a previous crashed run, nomad agent running as a different user than before, enforced security modules blocking mkdir under the alloc dir, full or read-only client disk

Understand the failure class

Background: mkdir permission denied (EACCES): failed to create directory errors explained — this error's family across 32 libraries.

Related errors

• failed to create mount point: %w
• plugin not executable
• failed to snapshot %s: %w
• error creating task %q dir: %w
• Couldn't open src file %v: %w

AI-assisted analysis of hashicorp/nomad@482b49bf1a (2026-09-04). Data as JSON: /api/errors/edcc9739ac19d15f. Report an issue: GitHub.