Re: Clear logical slot's 'synced' flag on promotion of standby

From: Ashutosh Sharma <ashu(dot)coek88(at)gmail(dot)com>
To: shveta malik <shveta(dot)malik(at)gmail(dot)com>
Cc: Ajin Cherian <itsajin(at)gmail(dot)com>, pgsql-hackers <pgsql-hackers(at)postgresql(dot)org>
Subject: Re: Clear logical slot's 'synced' flag on promotion of standby
Date: 2025-09-24 04:48:42
Message-ID: CAE9k0PmT5EOi7g8o18RHySM1yqOOtNUGVN=Wkb9mVKSKPja52A@mail.gmail.com
Views: Whole Thread | Raw Message | Download mbox | Resend email
Thread:
Lists: pgsql-hackers

On Wed, Sep 24, 2025 at 9:42 AM shveta malik <shveta(dot)malik(at)gmail(dot)com> wrote:
>
> On Tue, Sep 23, 2025 at 6:41 PM Ashutosh Sharma <ashu(dot)coek88(at)gmail(dot)com> wrote:
> >
> > > 3)
> > > + ereport(LOG,
> > > + (errmsg("reset synced flag for replication slot \"%s\"",
> > > + NameStr(s->data.name))));
> > >
> > > a) Shall we change it to DEBUG1?
> > > b) Shall the msg be:
> > > synced flag reset for replication slot \"%s\" during promotion
> > >
> >
> > I think this can stay as a LOG message, it only runs once at startup
> > and applies just to logical slots, so it won’t be noisy.
>
> It does not run at startup, it runs during promotion. Having said
> that, if there are a lot many slots, and we have one message per slot,
> overall logs can still be more. I somehow find DEBUG better here. But
> we can leave it as LOG and can revisit later when others review.
>
> > I’d also
> > avoid mentioning “during promotion,” since the flag might accidentally
> > be set on the primary and then reset later during startup, making that
> > description inaccurate.
>
> This function is called only during promotion (see check 'if
> (promoted)') and thus the suggested message ( “during promotion")
> seems better to me.
>

ResetSyncedSlots might be called during promotion, but
RestoreSlotFromDisk (which runs during standard PostgreSQL startup)
performs the same functionality. This creates a scenario where the
sync flag could be reset during either promotion or regular startup. I
think we should either remove it, or ensure it is present in both
places for consistency.

--
With Regards,
Ashutosh Sharma.

In response to

Browse pgsql-hackers by date

  From Date Subject
Next Message Rahila Syed 2025-09-24 05:01:29 Re: Use WALReadFromBuffers in more places
Previous Message shveta malik 2025-09-24 04:42:47 Re: Report bytes and transactions actually sent downtream