Skip to main content
One SRE week embedded inside shadow mode. Seven faults are injected against production-sized state; documented recovery times become the on-call runbook. All 7 must pass, signed off by the SRE lead — this is go-live Gate 6. For each experiment document: injection method/script · preconditions (state size!) · pass criteria · rollback · evidence captured.
Vendor failover claims (e.g., Redis ”< 5s”) must be verified against your own dedup-state size and client configuration (e.g., Lettuce) before being trusted in production. Note: the newer Kafka consumer-rebalance protocol materially helps experiment 7.