Alexander Kukushkin
Alexander Kukushkin
> even if replication between "a" and "b" has already been interrupted, patronictl list will still show node "b" as "streaming". Yes, member status is stored in the Pod annotations....
The simplest and backward compatible way of doing it would be enhancing: https://github.com/patroni/patroni/blob/e0d43022013f6f47dc2631e67b186ed18ca01406/patroni/api.py#L334-L337 And start supporting additional filters like `GET /replica?lag=32MB&replication_state=streaming`. @bluthg do you want to contribute a patch? :)
@kvazimoda24 LGTM. Since you are on ZooKeeper topic, maybe you can take over https://github.com/patroni/patroni/discussions/3475 and implement it as a separate PR?
> Some of our colleagues think that patroni may not be able to handle the error: patroni.dcs.etcd3.Unknown: (above it the logs). And that was the reason why it changed its...
@admnitazh you need to check Etcd logs on all members. Without knowing details it is impossible to chose the right retry strategy. Maybe we need just to retry request on...
Do you really think it is possible to get something useful from less than 1s of logs? Problem never happens instantly, but usually develops over time. To get a full...
@foot3print the issue is not closed, no fixes were implemented because I didn't get info I asked in https://github.com/patroni/patroni/issues/3407#issuecomment-3073005943 And also here: https://github.com/patroni/patroni/issues/3407#issuecomment-3073656913
@foot3print this is exactly what I am asking. 1. logs from etcd nodes starting from at least a few seconds before the problem (maybe even a few minutes) 2. To...
> For etcd Logs, the only thing i see over journald is this: you need to check logs on all etcd nodes