Re: Proposal: Allow excluding specific file patterns in pg_checksums

From: Horimoto Yasuhiro <horimoto(at)clear-code(dot)com>
To: Daniel Gustafsson <daniel(at)yesql(dot)se>, Greg Sabino Mullane <htamfids(at)gmail(dot)com>
Cc: XYenon <i(at)xyenon(dot)bid>, PostgreSQL Hackers <pgsql-hackers(at)postgresql(dot)org>, kou(at)clear-code(dot)com
Subject: Re: Proposal: Allow excluding specific file patterns in pg_checksums
Date: 2025-11-13 03:08:30
Message-ID: c7fdffb4-df49-49aa-89fa-34890f09424a@clear-code.com
Views: Whole Thread | Raw Message | Download mbox | Resend email
Thread:
Lists: pgsql-hackers

Hi all,

Thanks Greg and Daniel for the helpful feedback and suggestions.

Following your advice, we're considering moving PGroonga's database-specific
files out of the `base/` directory to avoid interference with `pg_checksums`.
Specifically, we're exploring placing them under a path such as
`pgroonga/${DB_OID}/pgrn*`.
This approach would likely resolve the checksum verification issue without
requiring any PostgreSQL changes.

Here's a brief summary of what we've verified so far:

- Files under `base/` indeed trigger `pg_checksums` verification failures.
- Moving the same files to `$PGDATA/pgroonga/` allowed checksum verification to
complete successfully.
- `pg_basebackup` includes files and directories placed directly under the
PostgreSQL data directory, not only `base/`. So our proposed new location
would still be included in backups.
- According to the PostgreSQL source code, `pg_checksums` checks only the
following directories. Therefore, files under `$PGDATA/pgroonga/` are safely
excluded.
- `global/`
- `base/`
- `pg_tblspc/`

We'll continue testing this approach to confirm compatibility and performance
before finalizing the change.

Thanks again for the discussion and insights that helped guide our direction.

Best regards,
Yasuhiro Horimoto

PGroonga Team

In response to

Browse pgsql-hackers by date

  From Date Subject
Next Message Mihail Nikalayeu 2025-11-13 03:13:00 Re: Logical replication: lost updates/deletes and invalid log messages caused by SnapshotDirty + concurrent updates
Previous Message Peter Smith 2025-11-13 03:00:19 Re: [PATCH] Add pg_get_subscription_ddl() function