Objective.
Run an App, a PostgreSQL database and a cache for 24 hours under steady use, from outside, and count every error.
Scenario.
The App was called four times a second, the database written once a second and the Plexivo API polled every 30 seconds, for 24 hours.
Expected behaviour.
- 0 failed database writes.
- App errors at or below 0.1%.
- Every API call answered.
Measured recovery.
- Application health
- Pass
- Database consistency
- Pass
Customer impact.
- Application availability
- 344,879 requests, 99.9986% answered; p95 66 ms
- Management operations
- Every API call answered
- Committed data
- Fully preserved
What happened.
- Steady use started.
- 24 hours completed: 79,319 writes, 0 failed.
Correctness checks.
- Committed data
- Fully preserved
What we learned.
- A full day of steady use with 79,319 successful writes.
What we changed.
- No change needed.
Scope of this drill.
One App, one database and one cache under steady, moderate load.