Objective.
Kill the only instance of an application while customers use it, and measure the longest pause they see.
Scenario.
The App's only instance was killed while a client called it continuously through the edge.
Expected behaviour.
- Replace the instance automatically.
- Serve again within seconds.
Measured recovery.
- Recovery time RTO
- 1.3 s
- How long recovery took.
- Application health
- Pass
Customer impact.
- Application availability
- 6 errors in 1,217 requests; longest interruption 1.29 s
What happened.
- Instance killed.
What we learned.
- The longest observed request interruption was 1.29 seconds.
What we changed.
- No change needed.
Scope of this drill.
An application with one instance. The recovery figure is the longest observed request interruption, not the time until a replacement instance was ready.