E2E persistence tests - Githubissues

improbable-eng / etcd-cluster-operator

A controller to deploy and manage etcd clusters inside of Kubernetes

MIT License

129 stars 35 forks source link

E2E persistence tests #69

Closed wallrj closed 4 years ago

wallrj commented 5 years ago

Part of #29

Don't set OwnerReference on PVC to prevent it being garbage collected for the peer it belongs to.
Test that etcd data can survive after the etcd process is restarted.
- Create a 1-node cluster
- Use etcdctl to add a key / value
- Delete all the cluster pods
- Use etcdctl to retrieve the value again.

Fixes: #50

wallrj commented 5 years ago

I notice that the ReplicaSet causes a new pod to be started before the existing pod has terminated. I'm not sure if that means that there may be two etcd processes reading (writing) the data.