Re: pg_upgrade instructions involving "rsync --size-only" might lead to standby corruption? - Mailing list pgsql-hackers

From Bruce Momjian
Subject Re: pg_upgrade instructions involving "rsync --size-only" might lead to standby corruption?
Date
Msg-id ZPoN7TbB7CL8pvsa@momjian.us
Whole thread Raw
In response to Re: pg_upgrade instructions involving "rsync --size-only" might lead to standby corruption?  (Nikolay Samokhvalov <nik@postgres.ai>)
Responses Re: pg_upgrade instructions involving "rsync --size-only" might lead to standby corruption?
List pgsql-hackers
On Mon, Jul 10, 2023 at 02:37:24PM -0700, Nikolay Samokhvalov wrote:
> Maybe. It will require changes in other parts of this doc.
> Thinking (here: https://gitlab.com/postgres/postgres/-/merge_requests/18/diffs)
> 
> Meanwhile, attached is v2
> 
> thanks for the comments

I looked over this issue thoroughly and I think I see the cause of the
confusion.  In step 8 we say:

    8. Stop both servers
    Streaming replication and log-shipping standby servers can remain
                                                           ---
    running until a later step.

Of course this has to be "must" and it would be good to explain why,
which I have done in the attached patch.

Secondly, in step 9 we say "verify the LSNs", but have a parenthetical
sentence that explains why they might not match:

    (There will be a mismatch if old standby servers were shut down before
    the old primary or if the old standby servers are still running.)

People might take that to mean that it is okay if this is the reason
they don't match, which is incorrect.  Better to tell them to keep the
streaming replication and log-shipping servers running so we don't need
that sentence.

The instructions are already long so I am hesitant to add more text
without a clear purpose.

-- 
  Bruce Momjian  <bruce@momjian.us>        https://momjian.us
  EDB                                      https://enterprisedb.com

  Only you can decide what is important to you.

Attachment

pgsql-hackers by date:

Previous
From: Markus Winand
Date:
Subject: Re: Wrong command name in writeable-CTE related error messages
Next
From: Bruce Momjian
Date:
Subject: Re: MergeJoin beats HashJoin in the case of multiple hash clauses