Message-ID: <aaf2d9ed1a5e8b271bcc0d4bb6a577385a6bb45f.camel@cybertec.at>
Subject: Re: slave stops replica
From: Laurenz Albe <laurenz.albe@cybertec.at>
To: Pepe TD Vo <pepevo@yahoo.com>, Pgsql-admin
	 <pgsql-admin@lists.postgresql.org>
Date: Fri, 27 Aug 2021 17:09:58 +0200
In-Reply-To: <1948635055.176990.1630061761986@mail.yahoo.com>
References: <1415172523.18991.1630012160409.ref@mail.yahoo.com>
	 <1415172523.18991.1630012160409@mail.yahoo.com>
	 <1e79ea171debe2b3ed9c297989f110cfccdf129b.camel@cybertec.at>
	 <1948635055.176990.1630061761986@mail.yahoo.com>
Content-Type: text/plain; charset="UTF-8"
User-Agent: Evolution 3.40.4 (3.40.4-1.fc34) 
MIME-Version: 1.0
Content-Transfer-Encoding: 8bit
Archived-At: 
 <https://www.postgresql.org/message-id/aaf2d9ed1a5e8b271bcc0d4bb6a577385a6bb45f.camel%40cybertec.at>
Precedence: bulk

On Fri, 2021-08-27 at 10:56 +0000, Pepe TD Vo wrote:
> > > I had master slave replica and yesterday the slave server's device was full on /var
> > > where the archive copy from master.  After trying to figure out why the
> > > pg_archivecleanup doesn't work, SA removed all wal files except the  current
> > >
> > > I tried to create a test table on master and didn't see it pick up from slave.
> > > I found out again /var is full. 
> >
> > If you have deleted WAL that has no yet been applied to the standby, replication
> > is broken and you have to build the standby from scratch.
> 
> Can we delete the old WALs?

Only the ones you don't need for replication.
Of course, if you rebuild replication from scratch, you can delete the WAL archives.

> May I know what  is the retention to perform tuning on the server like Oracle to
> keep only 7 days and why pg_archivecleanup doesn't work?

I guess that pg_archivecleanup doesn't delete anything because it is only
called once replication has processed a WAL segment and doesn't need it any more.

If you are missing a WAL segment because you deleted it, replication gets stuck
at that point and will wait indefinitely for that WAL segment.  So nothing is
processedm and nothing is deleted.

A gap in the WAL will stop and break replication, as I said.

Yours,
Laurenz Albe
-- 
Cybertec | https://www.cybertec-postgresql.com