Re: Reviving lost replication slots

On Wed, Nov 9, 2022 at 12:32 AM Kyotaro Horiguchi <horikyota(dot)ntt(at)gmail(dot)com>
wrote:

> I don't think walsenders fetching segment from archive is totally
> stupid. With that feature, we can use fast and expensive but small
> storage for pg_wal, while avoiding replciation from dying even in
> emergency.
>

Thanks! If there is a general agreement on this in this forum, I would like
to start working on this patch,

>
> At Tue, 8 Nov 2022 19:39:58 -0800, sirisha chamarthi <
> sirichamarthi22(at)gmail(dot)com> wrote in
> > > If it's a streaming replication slot, the standby will anyway jump to
> > > archive mode ignoring the replication slot and the slot will never be
> > > usable again unless somebody creates a new replication slot and
> > > provides it to the standby for reuse.
> > > If it's a logical replication slot, the subscriber will start to
> > > diverge from the publisher and the slot will have to be revived
> > > manually i.e. created again.
> > >
> >
> > Physical slots can be revived with standby downloading the WAL from the
> > archive directly. This patch is helpful for the logical slots.
>
> However, supposing that WalSndSegmentOpen() fetches segments from
> archive as the fallback and that succeeds, the slot can survive
> missing WAL in pg_wal in the first place. So this patch doesn't seem
> to be needed for the purpose.
>

Agree on this. If we add the proposed support, we don't need this patch.

>
>
> regards.
>
> --
> Kyotaro Horiguchi
> NTT Open Source Software Center
>