Skip to main content

Subsystem: Checkpoints & Crash Recovery

Two-phase commit checkpoints and startup recovery (src/checkpoint/).

Crash Recovery Flow​

  1. On startup, queries checkpoints and execution_runs.
  2. Inspects unsealed spool files and seals them with [INTERRUPTED].
  3. Reconciles running jobs and orphan processes.
  4. Classifies status into SafeToResume, NeedsRepair, Ambiguous, or Corrupt.