Re: Impact of checkpointer during pg_upgrade

On Fri, Sep 8, 2023 at 5:37 AM Michael Paquier <michael@paquier.xyz> wrote:
>
> On Thu, Sep 07, 2023 at 03:33:52PM +0530, Amit Kapila wrote:
> > I think if we just make max_slot_wal_keep_size to -1 that should be
> > sufficient to not let any slots get invalidated during upgrade. Do you
> > have anything else in mind?
>
> Forcing wal_keep_size while on it may be a good thing.
>

I had thought about it but couldn't come up with a reason to force
wal_keep_size for this purpose.

> > If we do (b) either via GUCs or IsBinaryUpgrade check we don't need to
> > do any of (a), (b), or (d). I feel that would be a minimal and
> > sufficient fix to prevent any side impact of checkpointer on slots
> > during an upgrade.
>
> I could get into the addition of a post-upgrade check to make sure
> that nothing got invalidated while the upgrade was running, FWIW.
>

This validation tries to ensure that we don't have any bugs/issues in
our patch. It may be a candidate for assert but I feel even if we
encounter any bug it is better to fix the bug.

--
With Regards,
Amit Kapila.