Search

Runbooks for maintaining the search index. These procedures cover reindexing repositories, draining the index outbox, and detecting and recovering from drift between the database and the search backend.

Pages

• OpenSearch admin recovery — Map each OpenSearch admin-probe cause code to its diagnostic path and wire Prometheus alerting for on-call paging on probe disconnects.
• Code search-index bulk-backfill — Backfill code search from on-disk repos when the code index or the OpenSearch code/symbols indexes return zero hits.
• Code-search incremental indexing — How the post-receive hook keeps code search current on branch pushes, with latency targets, skip cases, and stale-index diagnosis.
• Pages search-index bulk-backfill — Run and monitor the admin pages-only search-index backfill endpoint after pages schema migrations, with failure modes and audit trail.
• Reconcile dead-lettered ACL-sync / delete outbox rows — Re-arm dead-lettered search-index outbox rows via the admin reset endpoint or backfill, restoring fail-closed search coverage.
• Reset attempts-exceeded rows — Re-arm search-index outbox rows stranded past max_attempts after an OpenSearch outage via the admin reset endpoint.