You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
docs: document behavior when the storage is unavailable (outage, failover) #633
Found by the persona usability test of docs/wiki (docs/wiki-guardian-full-review): readers answered questions using only the wiki, without reading src/.
Gap (SRE persona, no answer in the wiki): nothing says what the dispatcher, the heartbeat and the enqueue do when the database goes away and comes back: does the host survive, does polling resume by itself, can HeartbeatTimeout requeue jobs that are still running during a long outage, what does the caller of EnqueueAsync see. Only Redis has a note, and it says failover was not tried. Postgres, SQL Server and MongoDB have nothing.
Work: read the dispatcher/heartbeat/enqueue code per provider, ideally prove it with an integration test (see the integration-tests-per-provider rule), then write a section 'When storage is unavailable' in concepts/delivery-guarantees.md and a one-line pointer in each storage page. No claim without code or test evidence.
Activity
oluciano commentedon Oct 19, 2026
Work Started (WIP)
docs/633-storage-unavailableMerge pull request #643 from oluciano/docs/633-storage-unavailable
oluciano commentedon Oct 19, 2026
Acceptance & Resolution Statement
develop.Merge pull request #649 from oluciano/test/642-storage-outage-risks