Re: Making pg_rewind faster

John H <johnhyvr@gmail.com>

From: John H <johnhyvr@gmail.com>

To: Srinath Reddy Sadipiralla <srinath2133@gmail.com>

Cc: Robert Haas <robertmhaas@gmail.com>, wenhui qiu <qiuwenhuifx@gmail.com>, Japin Li <japinli@hotmail.com>, Michael Paquier <michael@paquier.xyz>, Andres Freund <andres@anarazel.de>, Alexander Korotkov <aekorotkov@gmail.com>, Justin Kwan <justinpkwan@outlook.com>, Tom Lane <tgl@sss.pgh.pa.us>, pgsql-hackers <pgsql-hackers@postgresql.org>, vignesh ravichandran <admin@viggy28.dev>, "hlinnaka@iki.fi" <hlinnaka@iki.fi>

Date: 2025-10-16T22:51:56Z

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 →

pg_rewind: Skip copy of WAL segments generated before point of divergence
- 5173bfd0443e 19 (unreleased) landed
pg_rewind: Extend code detecting relation files to work with WAL files
- 6ae08d9583e9 19 (unreleased) landed
Split TESTDIR into TESTLOGDIR and TESTDATADIR
- c47885bd8b69 16.0 cited

Attachments

v10-0001-Avoid-copying-WAL-segments-before-divergence-to-.patch (application/octet-stream) patch v10-0001

Hi,

On Wed, Oct 15, 2025 at 7:27 AM Srinath Reddy Sadipiralla
<srinath2133@gmail.com> wrote:

>
> TBH first it seemed to me a good coding practice for future proofing,
> then i checked if there's any place in postgres we are doing
> something similar ,then found 1 in src/include/access/gist.h
>
> typedef struct GISTDeletedPageContents
> {
> /* last xid which could see the page in a scan */
> FullTransactionId deleteXid;
> } GISTDeletedPageContents;
>

It makes more sense to me if we expect the struct or same set of arguments
to be reused in different places / want a stable API reference. I don't think
we want everything to be wrapped around a struct for functions just in case.

> ..
> thanks for updating, i have attached a diff patch on
> top of v9-0001 patch , where i tried to add more tests
> ...

Thank you for the diff! Your changes are a lot cleaner and I've included it in
the latest patch. One difference I've made is instead of additionally logging in
decide_wal_file_action

> + pg_log_debug("WAL segment \"%s\" is copied to target", fname);

I realized we are already logging the WAL segments that are copied over in
print_filemap so I've updated the test to check for that instead

> qr/pg_wal\/$corrupt_wal_seg \(COPY\)/

I also updated the error message for the last three checks.

Thanks,

-- 
John Hsu - Amazon Web Services