Craig Ringer

I'm going to withdraw this, as I've retired my use of the opentelemetry-operator in favour of directly managed manifests. Would you like the issue left open in case someone else...

Wouldn't this typically be done on k8s via deployment of a `LoadBalancer` endpoint?

A workaround can be to deploy a very low `LimitRange` to your namespace, then make sure everything else specifies limits and requests explicitly. E.g. ``` apiVersion: v1 kind: LimitRange metadata:...

Possibly related to https://github.com/hadolint/hadolint/issues/355

Haven't had time to look yet, there's a great deal going on and priorities are spread a bit thin. A test case SQL script would be handy.

What was the function? Written in what language? Is there any chance it set `session_replication_role=replica` ? Or configured the replication origin? If so, that would be the expected outcome. I...

Thanks very much for reaching out. Sync is astonishingly difficult to get correct and race-free even when running one at a time. Parallel sync has mostly not been implemented for...

Hm, I got email notification that you commented, but nothing here. To be honest I don't remember where the deadlocks can occur, only that they can in a simple resync...

Please save the full logs. The `epoll_ctl() failed` error just indicates a broken socket, most likely terminated by the other end, so look for corresponding output there. Did you by...

Looks unrelated. The issue there seems to pertain to pglogical's dump/reload during `pglogical.create_subscription` instead; you're looking for issues around `pglogical_sync_tmpfile_cleanup_cb`. Adding a `%m` would be a big help, so I...