Re: Monitoring of a hot standby with a largely idle master

2017-07-13 20:15 GMT+12:00 Michael Paquier <michael(dot)paquier(at)gmail(dot)com>:

> On Thu, Jul 13, 2017 at 5:26 AM, Jeff Janes <jeff(dot)janes(at)gmail(dot)com> wrote:
> >
> > I think that none of the recovery information functions
> > (https://www.postgresql.org/docs/9.6/static/functions-
> admin.html#FUNCTIONS-RECOVERY-INFO-TABLE)
> > can distinguish a hot standby which is connected to an idle master,
> versus
> > one which is disconnected. For example, because the master has crashed,
> or
> > someone has changed the firewall rules.
> >
> > Is there a way to monitor from SQL the last time the standby was able to
> > contact the master and initiate streaming with it? Other than trying to
> > write a function that parses it out of pg_log?
>
> Not directly I am afraid. One way I can think about is to poll
> periodically the state of pg_stat_replication on the primary or
> pg_stat_wal_receiver on the standby and save it in a custom table. The
> past information is not persistent as any replication-related data in
> catalogs is based on the shared memory state of the WAL senders and
> the WAL receiver, and those are wiped out at reconnection.
> --
> Michael
>
>
>
That works for me too! I do this way... cron job runs that every X minutes
and if the replication lag is higher than 1 second it sends me an email..

It works pretty well and I used bash.

Lucas