I would still love to get a version of this patch backported. And I just
thought of an idea to do so without breaking the Windows ABI, by slightly
modifying my previous idea. See the attached patch.
0001-Treat-ETIMEDOUT-as-indicating-a-non-recoverable-conn.patch
Description: 0001-Treat-ETIMEDOU
Jelte Fennema writes:
> Personally, I would love to see this backpatched. Since together with a
> second bug I reported[1] it's causing high query timeouts in Citus even if
> tcp_user_timeout is set to a low value. I do understand your worry though.
> Would a patch like the one I attached now b
Oops sorry, I did make that change locally but apparently didn't update my
.patch file after committing, so I uploaded an intermediary one...
Thanks for fixing that.
I saw you added this section to the commit message:
> Perhaps this should be back-patched, but I'm hesitant to do so given
> the l
Jelte Fennema writes:
> Attached is a new patch that I think addresses your concerns.
You missed TranslateSocketError ...
Pushed to HEAD only with that fix.
regards, tom lane
Attached is a new patch that I think addresses your concerns.
From: Tom Lane
Sent: Thursday, September 30, 2021 16:04
To: Jelte Fennema
Cc: pgsql-hack...@postgresql.org
Subject: [EXTERNAL] Re: Add ETIMEDOUT to ALL_CONNECTION_FAILURE_ERRNOS
Jelte Fennema writes
Jelte Fennema writes:
> Previously successfully opened TCP connections can still fail on reads
> with ETIMEDOUT. This should be considered a connection failure, so that
> the connection in libpq is marked as CONNECTION_BAD. The reason I got an
> ETIMEDOUT was, because I had set a low tcp_user_time
