Re: POC: make mxidoff 64 bits

Heikki Linnakangas <hlinnaka@iki.fi>

From: Heikki Linnakangas <hlinnaka@iki.fi>
To: Maxim Orlov <orlovmg@gmail.com>, Ashutosh Bapat <ashutosh.bapat.oss@gmail.com>, Alvaro Herrera <alvherre@alvh.no-ip.org>, Alexander Korotkov <aekorotkov@gmail.com>
Cc: wenhui qiu <qiuwenhuifx@gmail.com>, Postgres hackers <pgsql-hackers@lists.postgresql.org>
Date: 2025-11-17T16:35:46Z
Lists: pgsql-hackers

Commits

Same data as JSON: GET /api/v1/messages/:b64id/commits the thread's linked commits as JSON, with link sources. API reference →
  1. Fix partial read handling in pg_upgrade's multixact conversion

  2. Increase timeout in multixid_conversion upgrade test

  3. Improve sanity checks on multixid members length

  4. Clarify comment on multixid offset wraparound check

  5. Never store 0 as the nextMXact

  6. Add runtime checks for bogus multixact offsets

  7. Widen MultiXactOffset to 64 bits

  8. Move pg_multixact SLRU page format definitions to a separate header

  9. Convert confusing macros in multixact.c to static inline functions

  10. Index SLRUs by 64-bit integers rather than by 32-bit integers

  11. Cope with possible failure of the oldest MultiXact to exist.

Attachments

Here's yet another patch version. I spent the day reviewing this in 
detail and doing little cleanups here and there. I squashed the commits 
and wrote a proper commit message.

One noteworthy refactoring is in pg_upgrade.c, to make it more clear (to 
me at least) how upgrade from version 9.2 and below now works. It was 
actually broken when I tested it. Not sure if I had broken it earlier or 
if it never worked, but in any case it works now.

I also tested upgrading a cluster from an old minor version, < 9.3.5, 
where the control file has a bogus oldestMultiXid==1 value (see commit 
b6a3444fa6). As expected, you get a "could not open file" error:

> Performing Upgrade
> ------------------
> Setting locale and encoding for new cluster                   ok
> ...
> Deleting files from new pg_multixact/members                  ok
> Deleting files from new pg_multixact/offsets                  ok
> Converting pg_multixact files                                 
> could not open file "/home/heikki/pgsql.93stable/data/pg_multixact/offsets/0000": No such file or directory
> Failure, exiting

I don't think we need to support that case. I hope there are no clusters 
in that state still in the wild, and you can work around it by upgrading 
to 9.3.5 or above and letting autovacuum run. But I wonder if a 
pre-upgrade check with a better error message would still be worthwhile.


Ashutosh, you were interested in reviewing this earlier. Would you have 
a chance to review this now, before I commit it? Alexander, Alvaro, 
would you have a chance to take a final look too, please?

- Heikki