Triage and recover disaster recovery

Map failure classes for region-b https://b.rag.example: retryable vs fatal vs needs-human. Retries need budgets, jitter, and idempotency rules aligned to restore validates citation gold set before traffic; corrupt snapshot rejected. The chapter’s signature failure — traffic cut before vector rebuild finishes — fluent ungrounded answers — must take a deliberate branch, not a generic catch-all.

Implement the degrade/rollback/refuse behavior DR lead failing over region A → B within RTO 2h / RPO 15m needs when DR-2026-07-30 repeats. Prefer scoped controls (one flag, one weight, one tenant, one secret version) over fleet-wide restarts. Preserve evidence; do not delete logs to “clean the demo.”

After containment, check rto_minutes, rpo_minutes, citation_precision_post_restore moves in the safe direction and watch for retry amplification. Write the stop condition that ends the incident response for this lab.

Chaos note: inject only one fault class at a time and restore fixtures after. Watch for dual failures — dependency down and retry amplifier — which is how traffic cut before vector rebuild finishes — fluent ungrounded answers becomes an outage. Customer communication templates (even if only for the drill) beat silence. If you queue deferred work, define poison-message handling. Budget documents should state the maximum extra spend allowed during retries. Close the loop by linking the containment action to a dashboard panel for rto_minutes, rpo_minutes, citation_precision_post_restore.

Draw a state diagram for degrade modes and put it in the repo as ASCII if needed. Cap concurrent retries across the process, not only per request. Ensure cancellation propagates to downstream HTTP clients. When failing closed, choose a user-visible message that does not leak internals. Practice the single command that flips the kill switch or weight to zero. After recovery, drain or inspect deferred work before declaring green. In this chapter the product is regional RAG platform restore with PostgreSQL + object storage + vector indexes, the human stakeholder is DR lead failing over region A → B within RTO 2h / RPO 15m, and the incident id you design against is DR-2026-07-30. Re-state the oracle in your notes — drill DR-2026-07-30: restore completes 97m; gold citation_precision ≥ 0.95 — and keep the invariant visible: restore validates citation gold set before traffic; corrupt snapshot rejected. Track rto_minutes, rpo_minutes, citation_precision_post_restore as the scoreboard. Surface under change control: region-b https://b.rag.example. If you only have forty minutes, finish the fixture for traffic cut before vector rebuild finishes — fluent ungrounded answers before polishing UI. Promotion language stays ternary: promote, hold, or roll back based on evidence, not hope.