Objective.
Kill the platform agent on a customer machine and measure how quickly it reconnects. Running applications must not notice.
Scenario.
The agent process was killed without warning.
Expected behaviour.
- Restart automatically and reconnect within 10 seconds.
- Leave running applications untouched.
Measured recovery.
- Recovery time RTO
- 2.1 s
- Target 10 s · met
Customer impact.
- Application availability
- Unaffected
- Management operations
- Back after 2.1 s
What happened.
- Agent killed.
- Agent restarted and reconnected.
What we learned.
- The agent restarts and reconnects in about two seconds.
What we changed.
- No change needed.
Scope of this drill.
An agent process failure on one machine.