Yeachan-Heo/oh-my-codex · warning · Error

canonical_scale_down_membership_changed

Error message

canonical_scale_down_membership_changed

What it means

The canonical team config read during scale-down no longer contains one or more of the workers requested for removal, meaning membership changed between caller capture and the barrier-held read. Scale-down aborts rather than tearing down stale targets.

Source

Thrown at src/team/scaling.ts:2070

        if (allDrained.every(Boolean)) break;
        await new Promise(r => setTimeout(r, 2_000));
      }
    }

    // Phase 3: acquire the membership barrier and task claim locks before any
    // pane effect. The barrier and locks remain held through the canonical
    // snapshot, exact pane teardown, and forward-recoverable commit.
    let removableWorkers = targetWorkers;
    let removableWorkerNames = new Set(removableWorkers.map((worker) => worker.name));
    let teardownFailure: ScaleError | null = null;
    try {
      await withTaskMembershipBarrier(sanitized, leaderCwd, async () => {
        await recoverTeamMembershipTaskTransaction(sanitized, leaderCwd);
        const authoritativeConfig = await readTeamConfig(sanitized, leaderCwd);
        if (!authoritativeConfig) throw new Error('canonical_scale_down_config_missing');
        const authoritativeWorkers = removableWorkers.map((worker) => authoritativeConfig.workers.find((candidate) => candidate.name === worker.name));
        if (authoritativeWorkers.some((worker): worker is undefined => !worker)) {
          throw new Error('canonical_scale_down_membership_changed');
        }
        // Capture targets from the canonical generation while the membership
        // authority is held; no caller-owned worker record authorizes a pane effect.
        removableWorkers = authoritativeWorkers as WorkerInfo[];
        removableWorkerNames = new Set(removableWorkers.map((worker) => worker.name));
        teamStateRoot = authoritativeConfig.team_state_root ?? resolveCanonicalTeamStateRoot(leaderCwd);
        Object.assign(config, authoritativeConfig);
        const candidateTaskIds = (await listTasks(sanitized, leaderCwd))
          .filter((task) => task.status !== 'completed' && task.status !== 'failed')
          .map((task) => task.id);
        await withTaskClaimLocks(sanitized, candidateTaskIds, leaderCwd, async () => {
          const lockedTasks = await listTasks(sanitized, leaderCwd);
          const configPath = join(teamStateRoot, 'team', sanitized, 'config.json');
          const configSnapshot = await readFile(configPath);
          const manifestPath = join(teamStateRoot, 'team', sanitized, 'manifest.v2.json');
          const manifestSnapshot = existsSync(manifestPath) ? await readFile(manifestPath) : null;
          const reconciledTasks = lockedTasks.filter((task) => task.status !== 'completed' && task.status !== 'failed'
            && (removableWorkerNames.has(task.owner ?? '') || removableWorkerNames.has(task.claim?.owner ?? '')));

View on GitHub (pinned to 3ad79a8a6f)

Solutions

1. Re-read the current team membership (team status) and retry scaleDown with workers that still exist
2. Serialize team operations so only one scaler runs at a time (lock or CLI-level queue)
3. Treat this error as success if the goal (worker removal) is already achieved

const config = await readTeamConfig(team, cwd);
const targets = workers.filter((w) => config?.workers.some((c) => c.name === w.name));
if (targets.length === 0) return; // already removed

catch (e) {
  if ((e as Error).message === 'canonical_scale_down_membership_changed') {
    return; // someone else already removed them; idempotent success
  }
  throw e;
}

When it happens

Trigger: Calling scaleDown with worker records captured earlier while another process already removed some of those workers from the config.

Common situations: Concurrent scale-down operations racing; UI showing stale worker list; retrying a scaleDown whose first attempt actually succeeded.

Related errors

• canonical_scale_up_rollback_resolved_membership_verification
• scale_down_cleanup_debt_path_missing_parent:${path}
• scale_down_cleanup_debt_path_escape:${path}
• scale_down_cleanup_debt_invalid_team_state_root:${worker.nam
• scale_down_cleanup_debt_invalid_worktree_metadata:${worker.n

AI-assisted analysis of Yeachan-Heo/oh-my-codex@3ad79a8a6f (2026-08-27). Data as JSON: /api/errors/eff8b5d606061aeb. Report an issue: GitHub.