Operations Runbooks
This folder contains incident response runbooks for recommendation serving and personalization operations.
Runbook Index
- Model Rollback Runbook
- Use for canary guardrail breaches, ranker failures, or bad model promotions.
- Personalization Queue Lag Runbook
- Use when pending refresh age, queue depth, or worker failure alerts fire.
- ANN Index Staleness Runbook
- Use when ANN index age exceeds threshold or retrieval quality degrades due to stale vectors.
- Incident Drill Checklist
- Use for weekly/monthly operational readiness drills.
- Canary Dashboard Spec
- Use to build/validate required rollout guardrail dashboards.
- Serving API V2 Performance Runbook
- Use after staging deployments to validate functional API flows, latency gates, and bounded load.
- Rollbar Operations Runbook
- Use to configure Rollbar tokens, environment gating, and safe test errors.
- Canonical Ingestion Canary Dashboard Spec
- Use to validate Pegasus -> Tiger ->
hh-lionevent canary health in engineering and production.
- Use to validate Pegasus -> Tiger ->
- Retrain and Reindex Playbook
- Use for scheduled refreshes and post-incident model/index regeneration.
- Legacy ClickHouse Table Cleanup
- Use to remove decommissioned v1 feature-store tables and unused ClickHouse scaffolding tables.
Primary Alert Mappings
RecSysCanaryLatencyAutoRollback->model_rollback_runbook.mdRecSysCanaryFallbackAutoRollback->model_rollback_runbook.mdRecSysCanaryPendingTimeoutAutoRollback->personalization_queue_lag_runbook.mdRecSysCanaryFeatureAvailabilityAutoRollback->model_rollback_runbook.mdRecSysPersonalizationQueueLagHigh->personalization_queue_lag_runbook.mdRecSysPendingRefreshAgeHigh->personalization_queue_lag_runbook.mdRecSysWorkerFailureRateHigh->personalization_queue_lag_runbook.mdRecSysStaleAnnIndex->ann_index_staleness_runbook.md