Skip to content

Two reconcile runs stuck in RUNNING since 2026-07-15 with nothing reaping them #34

Description

@raisedadead

Summary

Two reconcile workflow runs have been in RUNNING state since 2026-07-15 — three weeks — in the production Hatchet run history. Nothing reaps them.

Detail

From v1_statuses_olap: reconcile shows 985 COMPLETED, 3 FAILED, and 2 RUNNING with min(inserted_at) = max(inserted_at) = 2026-07-15.

Needs triage to establish whether this is:

  • an artemis-side handler that can block indefinitely (ReconcileSite has no explicit timeout of its own — it inherits whatever ctx Hatchet supplies), or
  • upstream Hatchet behaviour where an orphaned run is never transitioned after a worker disappears.

Why it matters

Orphaned RUNNING rows distort any "is the fleet healthy" read and, if they hold a concurrency slot keyed on site, could starve later runs for the same site.

Fix direction

Establish the cause first. If artemis-side, bound ReconcileSite with an explicit deadline. If upstream, confirm whether the deployed engine version reaps stale runs and track the upstream issue.

Metadata

Metadata

Assignees

No one assigned

    Labels

    bugSomething isn't working

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions