Capacity certification methodology
Understand OpsKnight load levels, correctness invariants, pass criteria, and benchmark provenance.
A topology is CERTIFIED only when every required scenario threshold and correctness invariant passes for a declared resource profile. A passing scenario inside a failed suite is a measurement, not a certified envelope.
| Level | Workload intent |
|---|---|
| L0 | Smoke and harness validation |
| L1 | Steady baseline |
| L2 | Normal production peak |
| L3 | Alert storm |
| L4 | Deduplication storm |
| L5 | Major outage and notification fanout |
| L6 | Provider degradation |
| L7 | Realtime and SSE scale |
| L8 | Recovery and chaos |
| L9 | Breaking-point ramp |
Required invariants
Certification checks for no duplicate open incidents, no lost accepted alerts, no false escalation after acknowledgement or resolution, no corrupted incident state, no critical-notification starvation, and provider idempotency.
Required provenance
Every result must record the test date, source commit, host and resource profile, runtime topology, database pools, load level, scenario thresholds, invariants, and overall status. Missing evidence produces NOT TESTED; measurements without a declared result produce MEASURED; a completed failing suite produces NOT CERTIFIED.
The current evidence and exact profile are summarized in Capacity benchmark results.
Last updated for v2.0.0
Edit this page on GitHub