Horton

Re: Lots of memory allocated when reassigning Large Objects

Tom Lane <tgl@sss.pgh.pa.us>

View thread View on postgr.es
From: Tom Lane <tgl@sss.pgh.pa.us>
To: Justin Pryzby <pryzby@telsasoft.com>
Cc: Guillaume Lelarge <guillaume@lelarge.info>, pgsql-hackers@postgresql.org
Date: 2021-12-01T18:47:58Z
Lists: pgsql-hackers 
Justin Pryzby <pryzby@telsasoft.com> writes:
> @Guillaume: Even if memory use with the patch isn't constant, I imagine it's
> enough to have avoided OOM.

I think it's good enough in HEAD.  In the back branches, the sinval
queue growth is bad enough that there's still an issue.  Still,
this is a useful improvement, so I added some comments and pushed it.

			regards, tom lane

Commits

  1. Avoid leaking memory during large-scale REASSIGN OWNED BY operations.

  2. Reduce memory consumption for pending invalidation messages.