Operations Proof

This page pulls together the measured benchmark snapshot, release checklist, rollback path, and operator-facing UX controls used to run OffGridFlow. Benchmark figures below come from the latest documented benchmark run on December 5, 2025 in a controlled Docker Compose environment.

18 ms
Health endpoint p95
Target < 50 ms
78 ms
Auth p95
Target < 100 ms
156 ms
Emissions calc p95
Target < 200 ms
780 ms
Report generation p95
Target < 1000 ms
58 ms
Database query p95
Target < 100 ms
99.8%
Overall success rate
14,275 / 14,300 requests

1. Benchmark Snapshot

Latest documented test configuration: 10 workers, 60-second duration, 100 RPS target for the main workload.

WorkloadTotal requestsSuccessAvg latencyp95Throughput
Health endpoint500100%12 ms18 ms50.2 RPS
Authentication API3,00099.9%45 ms78 ms99.9 RPS
Emissions calculation API6,00099.8%85 ms156 ms99.8 RPS
Report generation30099.3%450 ms780 ms9.93 RPS
Database query load4,50099.8%32 ms58 ms149.7 RPS
Capacity model from the same benchmark set: API target 1,000 RPS, web target 500 to 1,000 concurrent users, and worker target 50 batch jobs per hour per baseline replica set.

2. Release Discipline

Major releases follow a written pre-deployment, deployment, and post-deployment checklist.

Run backend, frontend, integration, and security checks before major release.
Verify secrets, database connectivity, backups, indexes, HPA, PDB, and ingress controls.
Execute migrations first, then roll out API, worker, and web services with rollout status checks.
Run smoke tests and live health checks after deploy, then monitor error rate and p95 latency for two hours.
Monitoring window after deployment: error rate < 1%, p95 latency < 500 ms, database connections < 80% of max, no unexpected alerts firing.

3. Rollback and Recovery

Recovery targets are documented and rollback steps are explicit rather than ad hoc.

Roll back API, worker, and web deployments with rollout undo.
Verify rollback health before considering database rollback.
If schema change is implicated, roll back one migration step and re-check service health.
Continue monitoring for 30 minutes and run a post-incident review.
RTO
4 hours
RPO
1 hour

4. Accessibility and Dense Workflow Evidence

The operator-facing application is designed for keyboard, assistive technology, and high-density data work.

Dashboard keyboard navigation coverage exists in Playwright workflow tests.
Dynamic chart components expose aria-label, aria-labelledby, and aria-describedby metadata.
Dialogs use role="dialog" and aria-modal="true"; error states use aria-live alert regions.
Loading states, focusable controls, and dense tables are exercised in end-to-end flows.

5. Operational Caveat

The benchmark figures on this page are the latest documented internal measurements, not a public third-party attestation. They are published to show actual performance envelopes, targets, and release controls without overstating certification or uptime claims.