Capacity benchmark results
Evidence-derived load measurements and certification status for supported deployment topologies.
These results are generated from the PR #777 certification artifact. They are measurements from one test profile, not universal production guarantees.
Capacity status: Every tested topology in this artifact is NOT CERTIFIED. Two supported variants were NOT TESTED. Some individual scenarios or correctness invariants passed, but no topology passed the complete suite. OpsKnight therefore publishes no certified alert, notification, user, SSE, or status-fanout envelope from this run.
Current planning fixture shapes
These dimensions come from the current load-fixture source. Use them as shapes for planning and pre-production tests. They are not the historical PR #777 input and are not certified capacity limits.
| Profile | Users | Services | Integrations/service | SSE sessions | Status subscribers |
|---|---|---|---|---|---|
| Small | 40 | 12 | 2 | 100 | 1,000 |
| Medium | 120 | 32 | 4 | 500 | 10,000 |
| Large | 400 | 80 | 5 | 2,500 | 100,000 |
| Storm | 1,000 | 200 | 5 | 5,000 | 100,000 |
Selected observed peaks
These are the highest measured scenario rates in the historical artifact. A peak is an observation, not a supported-rate statement; every listed topology failed the complete certification contract.
| Topology | Observed peak RPS | Certification |
|---|---|---|
| Compose Split + PgBouncer | 230.5 | NOT CERTIFIED |
| Swarm HA Split + PgBouncer | 104.6 | NOT CERTIFIED |
| Helm Split + PgBouncer | 95.5 | NOT CERTIFIED |
| Kustomize Split + PgBouncer | 66.1 | NOT CERTIFIED |
Certification summary
| Topology | Load levels | Scenarios passing | Invariants | Capacity status |
|---|---|---|---|---|
compose_integrated_bundled_db |
L0 | 2/6 | Failed | NOT CERTIFIED |
compose_integrated_external_db |
— | — | — | NOT TESTED |
compose_split_bundled_db |
L0 | 0/7 | Failed | NOT CERTIFIED |
compose_split_pgbouncer |
L0 | 1/8 | Failed | NOT CERTIFIED |
compose_split_pgbouncer_external_db_ca |
— | — | — | NOT TESTED |
swarm_single_node_split |
L0 | 2/7 | Failed | NOT CERTIFIED |
swarm_ha_split |
L0 | 4/8 | Failed | NOT CERTIFIED |
kind_helm_split_pgbouncer |
L0 | 0/8 | Passed | NOT CERTIFIED |
kind_kustomize_split_pgbouncer |
L0 | 0/8 | Passed | NOT CERTIFIED |
phase6_compose_integrated |
L0 | 0/10 | Failed | NOT CERTIFIED |
phase6_compose_split |
L0 | 0/11 | Failed | NOT CERTIFIED |
phase6_compose_split_pgbouncer |
L1, L2, L3, L4, L5, L6, L7, L8, L9 | 0/99 | Failed | NOT CERTIFIED |
phase6_swarm_integrated |
L0 | 0/8 | Failed | NOT CERTIFIED |
phase6_swarm_split |
L0 | 0/9 | Failed | NOT CERTIFIED |
phase6_swarm_split_pgbouncer |
L0 | 0/10 | Failed | NOT CERTIFIED |
phase6_swarm_ha_split_pgbouncer |
L1, L2, L3, L4, L5, L6, L7, L8, L9 | 0/99 | Failed | NOT CERTIFIED |
phase6_helm_integrated |
L0 | 0/6 | Failed | NOT CERTIFIED |
phase6_helm_split |
L0 | 0/7 | Failed | NOT CERTIFIED |
phase6_helm_split_pgbouncer |
L1, L2, L3, L4, L5, L6, L7, L8, L9 | 0/99 | Passed | NOT CERTIFIED |
phase6_kustomize_integrated |
L0 | 0/6 | Failed | NOT CERTIFIED |
phase6_kustomize_split |
L0 | 0/7 | Failed | NOT CERTIFIED |
phase6_kustomize_split_pgbouncer |
L1, L9 | 0/20 | Failed | NOT CERTIFIED |
Provenance
- Artifact:
artifacts/load-certification/certification-summary.json - Pull request: #777
- Artifact generated: 2026-09-28T12:20:49.770Z
- Product source revision tested:
52d6a3c5c791e7ad9ad21cd76f11cbdf11d3176f - Test harness revision recorded:
e2cd6e28092f544eea2342ebe455cac680fd157b - Test window: 2026-09-27T16:33:18.158Z through 2026-09-29T18:27:12.781Z
- Host profile: 10-core CPU, 16 GB RAM; Docker Engine 28.x; Kind v0.31.0
- Tested level: L0, L1, L2, L3, L4, L5, L6, L7, L8, L9
- Database and runtime profiles: recorded in the source certification report and topology definitions; do not transpose these measurements to a different pool, replica, host, or provider configuration.
The benchmark artifact was produced from product revision
52d6a3c5c791e7ad9ad21cd76f11cbdf11d3176f. The harness was later revised ate2cd6e28092f544eea2342ebe455cac680fd157b, and the benchmark was not rerun after that harness change. The results remain historical evidence, not a certification of the later harness revision.
How to use these results
Use the failures to choose what to observe and scale, not as sizing promises. See Scaling signals and Certification methodology. Numeric production guidance will be published only after a topology reaches CERTIFIED status.
Last updated for v2.0.0
Edit this page on GitHub