Re: pg_upgrade_replica: avoid full re-clone of standbys after pg_upgrade

From: Marco Nenciarini <marco(dot)nenciarini(at)enterprisedb(dot)com>
To: pgsql-hackers(at)postgresql(dot)org
Cc: reshkekirill(at)gmail(dot)com
Subject: Re: pg_upgrade_replica: avoid full re-clone of standbys after pg_upgrade
Date: 2026-09-08 13:13:45
Message-ID: CA+nrD2fwcm9cgJB6icGbnhrfz7AQTFPYLH9Z0GjGt9a05z6VYw@mail.gmail.com
Views: Whole Thread | Raw Message | Download mbox | Resend email
Thread:
Lists: pgsql-hackers

> So, is all of this just a built-in analog for rsync with params?

Yes, but built for a place where rsync isn't an option.

My target is Kubernetes, specifically CloudNativePG. Pods have no ssh
between them, and no rsync binary. It's not that I avoid rsync, it
just isn't there.

On safety: --size-only trusts file size alone. My tool checks the old
replica's control data against the exact old primary checkpoint first,
then trusts only the blocks the new primary's WAL summarizer reports
unchanged since. Same size can still mean different content.

Marco Nenciarini
EnterpriseDB

In response to

Browse pgsql-hackers by date

  From Date Subject
Next Message Andres Freund 2026-09-08 13:14:03 Re: Assert in test_bms_membership();
Previous Message Kirill Reshke 2026-09-08 13:05:47 Re: pg_upgrade_replica: avoid full re-clone of standbys after pg_upgrade