Horton
Menu

Re: Changing the state of data checksums in a running cluster

Daniel Gustafsson <daniel@yesql.se>

View thread
From: Daniel Gustafsson <daniel@yesql.se>
To: Heikki Linnakangas <hlinnaka@iki.fi>
Cc: Tomas Vondra <tomas@vondra.me>, Andres Freund <andres@anarazel.de>, Bernd Helmle <mailings@oopsware.de>, Michael Paquier <michael@paquier.xyz>, Michael Banck <mbanck@gmx.net>, PostgreSQL Hackers <pgsql-hackers@lists.postgresql.org>
Date: 2026-03-19T23:34:02Z
Lists: pgsql-hackers

Attachments

> On 19 Mar 2026, at 14:49, Daniel Gustafsson <daniel@yesql.se> wrote:
>> On 18 Mar 2026, at 09:06, Heikki Linnakangas <hlinnaka@iki.fi> wrote:

>> Well, if CREATE DATABASE computed the checksums, that would fix this too. Actually, at least in the default wal_log=true mode, doesn't it already do that? It goes through the buffer cache as usual, I presume the checksums will be computed too.
> 
> That's a very good point, if the CREATE DATABASE is issued with wal_log and not
> file_copy it will mark the buffers dirty during copying which will calculate
> and set the checkpoint.  If file_copy isn't allowed during inprogress-on then
> the logic around rescanning databases for new entries could be quite
> simplified.

I tried to implement this on top of a fresh rebase, the attached 0003 contains
a very lightly tested version of this simplified logic (there is a removal of a
useless debug logging as well which while unrelated happened to sneak into this
work).

It now works off a single databaselist and relies on wal_copy to handle any
created database regardless of which template is used (file_copy is not allowed
while inprogress-on).  This gives a chance to do better erroring out in case of
failed databases, the code is still vulnerable to a the DROP/CREATE with the
same Oid, but that seems doable to solve.

It needs more testing, and likely has bugs, but I wanted to share it as soon as
possible to see if was along what you were thinking.  Personally I think the
code a lot easier to follow with this.

--
Daniel Gustafsson

Commits

Same data as JSON: GET /api/v1/messages/:b64id/commits the thread's linked commits as JSON, with link sources. API reference →

  1. Use correct datatype for PID

  2. Improve comments in online checksums code

  3. Fix checksum state transition during promotion

  4. Fix regex searching for page verification failures in tests

  5. Apply data-checksum worker throttling parameters

  6. Skip WAL for unlogged main fork during online checksum enable

  7. Fix data_checksum GUC show_hook

  8. Improve database detection logic in datachecksumsworker

  9. Improve handling of concurrent checksum requests

  10. Typo and spelling fixups for online checksums

  11. Fix invalid checksum state transition in checkpoints

  12. Handle data_checksum state changes during launcher_exit

  13. Test improvements for online checksums

  14. Prevent pg_enable/disable_data_checksums() on standby

  15. Test stabilization for online checksums

  16. Make data checksum tests more resilient for slow machines

  17. Formalize WAL record for XLOG_CHECKPOINT_REDO

  18. Revert "Get rid of WALBufMappingLock"

  19. Get rid of WALBufMappingLock

  20. Improve grammar of options for command arrays in TAP tests