Jelte Fennema-Nio
Jelte Fennema-Nio
To clarify, I agree that there is a bug here. But the we should not fix it in the way that this PR currently does. The suggestion from @petere seems...
I'm going to close this in favor of the approach here: https://github.com/pgbouncer/pgbouncer/pull/902 We've decided that (de-)serializing full PgBouncer in memory state is not maintainable and too likely to break things....
Reopening this as #817 didn't fix the problem after all, it got reported again in #968.
@drekthral yeah it's definitely not a fun bug, a PR to fix is is definitely appreciated.
I'm closing this in favor of #845.
Thank you for the PR, but #845 seems better because it does not require locking server connections to certain clients.
> Is there any downside when users include the peer itself? (to keep config consistent) No, that works just fine. When receiving a cancellation, the code first checks if the...
@achanda Yes, you misunderstood. This peering feature is useful if there's a load balancer in front of these different pgbouncer instances (or if you're using `so_reuseport=1`, which is effectively turns...
@veshant The #706 PR is not in a state where it can be merged.
Closing this because protocol-level prepared statements was merged in #845 and is now released in PgBouncer 1.21.0