Re: BUG #14171: Wrong FSM file after switching hot standby to master

> Date: Thu=2C 2 Jun 2016 07:42:32 -0700
> From: andres@anarazel.de
> To: michael.paquier@gmail.com
> CC: timofeid@outlook.com=3B pgsql-bugs@postgresql.org
> Subject: Re: [BUGS] BUG #14171: Wrong FSM file after switching hot standb=
y to master
>=20
> On 2016-06-02 16:42:35 +0900=2C Michael Paquier wrote:
> > > Sometimes we have troubles with fsm-files. In case:
> > >  =95 master instance is switching to another node(failover or switcho=
ver) on
> > > highload
> > >  =95 Hot standby node restart and run as master succesfully.
> > >  =95 After that we sometimes get FSM files pointing to non-existent b=
locks in
> > > the table=2C so subsequent insert operations on such tables fails wit=
h error
> > > message like following: 'could not read block XX in file "base/YYYY/Z=
ZZZZ"'.
> > > The issue can be resolved by either deleting of wrong FSM file (while
> > > database is stopped) or performing VACUUM FULL on erroneous table. Th=
e
> > > problem is usually observed on relatively small tables (e.g. up to 30
> > > blocks) which are often cleaned out (having most rows deleted).
> > > Does anybody already faced such behavior? What can be the root cause =
of such
> > > problems? Are there any recommendations on how to avoid them?
> >=20
> > Andres=2C do you think that c6ff84b0 can help here? Those symptoms look
> > rather similar to some missing invalidation messages on the standby.
>=20
> If there was a restart involved=2C it seems unlikely that that'll be
> relevant. Timofei=2C do I understand correctly that the problem persists
> across restarts?
>=20
> Regards=2C
>=20
> Andres

Yes=2C problem persists across restarts. We can resolve problem only by per=
forming VACUUM FULL or delete inconsistent FSM file.

Regards=2CTimofei Dynikov=20
                           =