We have a Stolon cluster setup on AWS with persistent volumes, which we destroy each night. In order to have the cluster resume the next day, we start the keepers with the existing flag: exec stolon-keeper '{"initMode":"existing","existingConfig":{"keeperUID": '"$STKEEPER_UID"'}}'. However, that in itself is not enough. Since the Etcd store is started empty, the Keepers cannot get the cluster data: https://github.com/sorintlab/stolon/blob/057389f7e484ee1d5c1e1a7020256020e7413c87/cmd/keeper/cmd/keeper.go#L1048
When I export the clusterdata from the cluster upon destroy and insert that value on startup, the cluster does come up.
I wonder if this is the (/a) correct way of doing. Having a persistent volume on AWS has no operational overhead on our Stolon management. However, the Etcd bit I mention above does! we need the insert a step to get the Etcd key on destroy, and one to insert it on create. So ideally I would cut that step out.