Choose the decision supported by the failure evidence. The questions separate controller status, durable effects, and user-visible recovery.
Review these lessons
- Replica lag: define when a read is allowed to be stale
- Consumer rebalances: preserve ordering and effect ownership
- Metric cardinality: keep observability usable during a surge
- Trace sampling: retain useful failures without flooding storage
- Stuck finalizers: finish cleanup before removing the guard
- Runbook automation: put a stop gate before the irreversible step
Other checks
- DevOps checks
- DevOps system boundaries decisions
- DevOps stateful operations decisions
- DevOps operations decisions
Common Mistakes
- Do not infer a complete trace from one retained span.
- Do not infer a durable business effect from a committed broker offset.