I am receiving the same errors in the operator log as in the description of this bug:

plus, the replication does not happen.

what I did was:

1. setup replication with haproxy

2. make sure it works

3. just change proxysql -> enabled and haproxy -> disabled

4. setup replication again

5. it doesn't work

Using haproxy:

For replica in pxc-monit logs I see:

But

Replica cluster:
Operator log:

Both clusters in READY state

Replica status:

But data replicate from source cluster to the replica one.

Connect to proxysql in replica NS and can see all the data:

I was checking this change in proxysql-admin: https://github.com/percona/proxysql-admin-tool/pull/222
And we are no more hitting the same issue, but it seems there are other issues.
I propose we approve the proxysql-admin change (since it's simple and analog to what was already done) and continue fixing operator for this use case (if we decide that it is worth it).

For example, operator is printing this in the logs:

If you try to connect to read-only cluster through proxysql endpoint and execute some read-only query it will fail because it expects nodes in hostgroup 11, but there aren't any:

The replication starts to get into "Waiting for dependent transaction to commit" after some time working and at the current moment I'm not sure what is causing it (I was running sysbench load).

But the pods seem to be running: