Chaos Experiments Catalogue (cHEX)
Recommended Scenario | Hypothesis | Gremlin Scenario Steps | Blast Radius / Targets | Link |
|---|---|---|---|---|
Validate Auto-Scaling with Status Checks | Status Checks will validate that your Cloud provider (like DigitalOcean) and a critical dependency (like GitHub) are in a steady state before launching attacks. When CPU usage ramps up and hits a set threshold, active instances will increase and decrease when CPU usage goes down. User sessions will remain active without throwing any errors. |
|
| https://app.gremlin.com/scenarios/recommended/validate-auto-scaling/hosts |
Validate Health Checks - Packet Loss | As packet loss degrades network communications, it should cause the targeted node or service to be marked as unhealthy. Your load balancer should distribute requests to other healthy resources. If an orchestrator is used, the unhealthy node or service should be replaced with a new one. |
|
| https://app.gremlin.com/scenarios/recommended/validate-health-checks-packet-loss/hosts |
Cache unavailable | When one cache AZ is made unavailable, the user experience will not experience an outage |
|
|
|
Cache CPU starvation | When CPU resources spike on the cache, additional cache instances will be added to the pool. |
|
|
|
Cache IO starvation | When IO resources spike on the cache instances, we will be alerted via monitors |
|
|
|
Cache Process Killer | When memcached is killed by the process killer attack. The instance will shutdown and a new instance will replace it. |
|
|
|