From: Jyrgal (@Jyrgal) <noreply+Jyrgal@github.com>
To: pgbackrest/pgbackrest <noreply+pgbackrest-pgbackrest@github.com>
Subject: Re: [pgbackrest/pgbackrest] issue #2913: Proposal: Checkpoint saves in resume feature override for S3 object lock
Date: Thu, 12 Nov 2026 05:03:22 +0000
Message-ID: <gh-pgbackrest-pgbackrest-2913-c6815153072@github.com> (raw)
In-Reply-To: <gh-pgbackrest-pgbackrest-2913@github.com>
References: <gh-pgbackrest-pgbackrest-2913@github.com>
Thanks! I've reworked it as you suggested.
- **`tmp` flag**: added to `storageNewWriteP()` and passed through to the driver, including over the remote protocol. Backup sets `.tmp = !final` for the checkpoint copies of `backup.manifest.copy`. Other drivers ignore it.
- **`repo-storage-tmp-retention`**: on tmp files, S3 sets an Object Lock retention and Azure sets a blob immutability policy. Both override the default retention.
- **Modes**: `repo-s3-tmp-retention-mode` (`governance`/`compliance`) and `repo-azure-tmp-retention-mode` (`unlocked`/`locked`). They're separate options because the config builder doesn't support a per-type `allow-list`. I'm happy to switch to a single option if you prefer.
- **GCS**: not supported. An object retention can't shorten a bucket retention policy (GCS keeps the object until both are satisfied), so it would only add retention. I've left it out of the possible options as it doesn't really make sense as the bucket policy would override it anyway if we try to shorten the retention.
I've opened a draft PR with the changes: https://github.com/pgbackrest/pgbackrest/pull/2918