| From: | JoongHyuk Shin <sjh910805(at)gmail(dot)com> |
|---|---|
| To: | "pgsql-hackers(at)lists(dot)postgresql(dot)org" <pgsql-hackers(at)lists(dot)postgresql(dot)org> |
| Subject: | Re: [PATCH] Extend MXactCache lifetime from per-transaction to per-session |
| Date: | 2026-10-06 05:04:58 |
| Message-ID: | CACSdjfOYrRYAPFX-3jA7vGJRApRJM8eyWwFB+di5pD7S9i6uKw@mail.gmail.com |
| Views: | Whole Thread | Raw Message | Download mbox | Resend email |
| Thread: | |
| Lists: | pgsql-hackers |
On Mon, Apr 20, 2026 at 6:38 PM JoongHyuk Shin <sjh910805(at)gmail(dot)com> wrote:
> Since MultiXact member lists are immutable once created, cached entries
> stay valid regardless of transaction boundaries.
MultiXactIds are assigned again after wraparound, so a backend that
stays connected across a wraparound can still hold a cache entry for an
old MultiXactId. If that backend then looks up a new MultiXact that
received the same MultiXactId, it may get the old member list back.
I think the patch needs additional design work to handle this case, and
after fixing that it would also need measurements showing a real
performance gain. So for now I have marked the CF entry as Withdrawn.
--
JoongHyuk Shin
>
| From | Date | Subject | |
|---|---|---|---|
| Next Message | jian he | 2026-10-06 05:40:23 | should pg_dumpall --clean apply to database in --exclude-database |
| Previous Message | shveta malik | 2026-10-06 04:55:35 | Re: Persist slot invalidations before publishing them |