Triage and recover rate limiting

Map failure classes for POST /v1/rag/answer: retryable vs fatal vs needs-human. Retries need budgets, jitter, and idempotency rules aligned to tenant RPM soft=60 hard=100; global 5k RPM; 429 includes Retry-After. The chapter’s signature failure — Redis down → fail-open floods provider → shared 429s — must take a deliberate branch, not a generic catch-all.

Implement the degrade/rollback/refuse behavior platform SRE stopping noisy neighbor tenant StormCo needs when RL-FAILOPEN-61 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 429_ratio_by_tenant and provider_429_ratio ≤ 0.01 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 Redis down → fail-open floods provider → shared 429s 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 429_ratio_by_tenant and provider_429_ratio ≤ 0.01.

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 layered token-bucket limiter for multi-tenant RAG API, the human stakeholder is platform SRE stopping noisy neighbor tenant StormCo, and the incident id you design against is RL-FAILOPEN-61. Re-state the oracle in your notes — StormCo at 200 RPM gets 429 after burst; Globex at 40 RPM unaffected — and keep the invariant visible: tenant RPM soft=60 hard=100; global 5k RPM; 429 includes Retry-After. Track 429_ratio_by_tenant and provider_429_ratio ≤ 0.01 as the scoreboard. Surface under change control: POST /v1/rag/answer. If you only have forty minutes, finish the fixture for Redis down → fail-open floods provider → shared 429s before polishing UI. Promotion language stays ternary: promote, hold, or roll back based on evidence, not hope.