RE: BUG #18165: Could not duplicate handle for "Global/PostgreSQL.xxxxxxxxxx": Bad file descriptor

Categorization:
Unclassified

Thank you all! :) I changed the setting back to 0 and we're not getting the error anymore. I have another error from pg11 I'll submit a new bug for.

Maxime

-----Original Message-----
From: Thomas Munro <thomas(dot)munro(at)gmail(dot)com>
Sent: October 21, 2023 5:29 PM
To: Boyer, Maxime (he/him | il/lui) <Maxime(dot)Boyer(at)cra-arc(dot)gc(dot)ca>
Cc: Alexander Lakhin <exclusion(at)gmail(dot)com>; Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us>; Andres Freund <andres(at)anarazel(dot)de>; pgsql-bugs(at)lists(dot)postgresql(dot)org
Subject: Re: BUG #18165: Could not duplicate handle for "Global/PostgreSQL.xxxxxxxxxx": Bad file descriptor

On Sun, Oct 22, 2023 at 4:38 AM Boyer, Maxime (he/him | il/lui) <Maxime(dot)Boyer(at)cra-arc(dot)gc(dot)ca> wrote:
> You guys are good! I did set min_dynamic_shared_memory to 2 GB in the new config, which didn't exist on 11. I can set it to 0 and give it a try on Tuesday. Does 2 GB seem overkill? We have plenty of memory to go by.

Apparently it has never worked on Windows and you're the first to notice and report that, so we can't expect anyone to have platform-specific experience with that. But on other OSes we've seen it improve performance of large parallel joins and parallel vacuum, especially when 'huge pages' are configured. (Note that we had reports that our huge page support for windows stopped working on recent Windows versions, addressed in PostgreSQL 16 by commit fdd8937c, so YMMV on PostgreSQL 14.)

Thanks for the report, and to Alexander for testing. I've pushed the fix now, so it will ship in the next point release in a few weeks. In the meantime, the workaround is to set it to 0.