Skip to main content

Resumable Execution Runbook

Resumable execution ensures long-running ACM workflows can pause, recover, and continue without losing deterministic guarantees.

Responsibilities​

  • Persist checkpoints after every task boundary with ledger correlation IDs.
  • Store checkpoint payloads in durable storage (S3, Blob, GCS) with retention aligned to replay bundles.
  • Track resume tokens inside the decision ledger (TASK_RESUMED, TASK_WAITING).

Recovery flow​

  1. Runtime detects failure or external pause signal.
  2. Operator inspects ledger entries and selects a checkpoint via CLI (pnpm --filter @ddse/acm-runtime run resume --from <id>).
  3. Policy and verification hooks re-evaluate before resuming downstream tasks.
  4. Replay bundles merge original and resumed segments for full audit coverage.

References​

  • framework/node/docs/tdr/RUNBOOK_RESUMABLE.md
  • framework/node/docs/tdr/RUNBOOK_RESUMABLE.md#decision-ledger for event taxonomy