hashicorp/nomad · error · ErrCgroupMustBeSet

configureCgroups

Error message

configureCgroups: %w

What it means

configureCgroups (called from newLibcontainerConfig) returns early if ResourceLimits is off; otherwise it requires a stats cgroup name from command.StatsCgroup(). When the cgroup is empty it wraps the sentinel ErrCgroupMustBeSet with this "configureCgroups: " prefix. It means the executor cannot place the task into a cgroup because none was assigned by the driver/client.

Solutions

  1. Ensure the driver allocates and sets the stats cgroup (SetStatsCgroup) before enabling ResourceLimits and calling Launch
  2. Check client cgroup configuration (cgroups v1/v2 setup, nomad cgroup parent) so cgroup allocation succeeds
  3. If cgroups are not wanted, disable ResourceLimits in the ExecCommand instead of leaving the cgroup unset
  4. Report/fix driver ordering: the cgroup must exist before newLibcontainerConfig runs
Defensive patterns

Strategy: validation

Validate before calling

if cmd.ResourceLimits && cmd.StatsCgroup() == "" {
    return fmt.Errorf("stats cgroup must be set before Launch when ResourceLimits is enabled")
}

Try / catch

if _, err := executor.Launch(cmd); err != nil && strings.Contains(err.Error(), "configureCgroups") && errors.Is(err, executor.ErrCgroupMustBeSet) {
    log.Printf("stats cgroup missing: %v — set cgroup before Launch", err)
}

Prevention

When it happens

Trigger: command.ResourceLimits is true but command.StopsCgroup()/StatsCgroup() returns "" — the driver never called SetStatsCgroup / the cgroup was not allocated for the task before Launch.

Common situations: Driver invoked the executor with ResourceLimits enabled but forgot to assign the stats cgroup (driver bug or ordering issue); cgroup allocation failed silently upstream (cgroups v1/v2 misconfiguration on the client); custom code constructing ExecCommand manually without StatsCgroup.

Related errors


AI-assisted analysis of hashicorp/nomad@482b49bf1a (2026-09-04). Data as JSON: /api/errors/144fb1dc1d85afc9. Report an issue: GitHub.

Appendix: source

Thrown at drivers/shared/executor/executor_linux_cgo.go:782

	}

	if len(command.Mounts) > 0 {
		cfg.Mounts = append(cfg.Mounts, cmdMounts(command.Mounts)...)
	}

	return nil
}

func (l *LibcontainerExecutor) configureCgroups(cfg *runc.Config, command *ExecCommand) error {
	// note: an alloc TR hook pre-creates the cgroup(s) in both v1 and v2

	if !command.ResourceLimits {
		return nil
	}

	cg := command.StatsCgroup()
	if cg == "" {
		return fmt.Errorf("configureCgroups: %w", ErrCgroupMustBeSet)
	}

	// // set the libcontainer hook for writing the PID to cgroup.procs file
	// TODO: this can be cg1 only, right?
	// l.configureCgroupHook(cfg, command)

	// set the libcontainer memory limits
	l.configureCgroupMemory(cfg, command)

	// set cgroup v1/v2 specific attributes (cpu, path)
	switch cgroupslib.GetMode() {
	case cgroupslib.CG1:
		return l.configureCG1(cfg, command, cg)
	default:
		return l.configureCG2(cfg, command, cg)
	}
}

View on GitHub (pinned to 482b49bf1a)