Sangeetha-Grantha

Metadata Value
Status Active
Version 1.2.0
Last Updated 2026-09-10
Author Sangeetha Grantha Team
Document Type Current guide

Investigate an application incident


Start by identifying the affected environment, user journey, time range, and scope. Preserve the source payload, request/task identifiers, and relevant logs before retrying or changing data.

Triage

  1. Check service state and logs; basic /health is not a full readiness probe.
  2. Identify the boundary: client transport, authorization, API/service, database, extraction, provider, or indexing.
  3. Reproduce with the smallest affected request/source scope.
  4. Compare the observed result with the current API/pipeline contract.

Choose the recovery path

Failure Recovery direction
Wrong local endpoint or credentials Correct component-specific configuration and restart
Migration/history failure Preserve backup/history and rehearse a targeted Flyway reconciliation in isolation
Failed extraction Inspect source/path/format/provider error, then retry the intended request
Incorrect accepted text Fix parser/mapping and use targeted reingestion/curation with provenance
Search index mismatch Inspect active profile, content hashes and query embedder before reindexing/activation
Native reader issue Capture device/runtime, selected variant and request state; exercise the affected journey

Avoid broad resets or queue-wide retries before understanding which records are affected. A retry may create provider usage or mutate accepted content.

Close with evidence

Record cause, affected scope, corrective action, checks actually run, and unresolved follow-up. Verify the affected user journey and relevant database/API state after recovery. Historical “zero issues” reports and empty placeholder dashboards are not evidence for this incident.

Monitoring · Configuration · Database runbook · Post-import checks


Section index · Documentation home · Feature status