Alexander Kukushkin

> Is that right? Isn't the Jepsen test supposed to verify it?

@wb14123 is it based on https://github.com/jepsen-io/jepsen?

> if I want to cleanup the cluster before each test, other than deleting all the resources in Kubernetes by kubectl delete -f file.yaml, is there anything else I need...

> After the test finished. I manually checked the state. 1 node is still killed while 2 other nodes are alive. However, the other 2 nodes cannot recover from the...

`standby_cluster.restore_command` in the global configuration does exactly that.

I am sorry, but I really can't understand anything from your explanation. Please try explaining it without using old/new/source/target words.

> Patroni attempted pg_rewind, it failed. The issue is that the restore_command in the standby_cluster config points to DC2 cluster's WALs, but pg_rewind needs to access DC1 cluster's archived WALs....

> Is there anything i am missing? You just don't understand all details about pg_rewind and docs don't tell the full story. Please check the source code: https://github.com/postgres/postgres/blob/635f580120b99f6df71d7c12749b22acde61c5ad/src/bin/pg_rewind/pg_rewind.c#L461-L488 Specifically, findLastCheckpoint()...

> we need target WALs during the pg_rewind process not the source WAL consider the following: 1. source writes a checkpoint to a file 1A and archives it 2. source...