[pgpool-general: 5227] Re: Questionable Failover

[Muhammad Usama]

On Wed, Jan 11, 2017 at 10:06 PM, Yates, James C. -ND <
James.C.Yates.-ND at disney.com> wrote:

> I was not expecting that behavior from PgPool.  I was under the impression
> that a failover would not occur unless the primary/master server had
> issues.
>

The failover is not limited to the primary/master backend nodes only. But
each backend node can be individually configured to allow or disallow the
failover in case of a backend error (using backend_flag config), and the
tolerance level in case of connectivity issue with backend nodes is also
configurable using the Pgpool-II health check configurations
<http://www.pgpool.net/docs/pgpool-II-3.6.1/doc/en/html/runtime-config-health-check.html>
.

Also the user provided script (failover_command
<http://www.pgpool.net/docs/pgpool-II-3.6.1/doc/en/html/runtime-config-failover.html>)
is executed by the Pgpool-II at the time of failover, that is passed with
all the information of the master/primary node and the failing node. So in
the failover script a user can decide not to take any action when the
standby node failover is performed if that is intended.

I also want to mention about the enhancement made in the Pgpool-II version
3.6, now in the case of standby node failover the sessions not using the
failed standby node remains unaffected by failover.

Thanks
Best regards


> On Jan 11, 2017, at 3:32 AM, Muhammad Usama <m.usama at gmail.com> wrote:
>
> Hi
>
> It seems like there was a temporary network glitch/failure between the two
> data centers around "Jan 05 00:04:29" time triggered the failover of
> backend node0.
>
> If we look at the logs of each node around this time we can see that the
> Pgpool-II nodes in the first data center (dbserver1 and dbserver2) were not
> able to connect to PostgreSQL on dbserver3 that is located in the second
> data center, and at the very same time the Pgpool-II on dbserver3 was also
> having the connection problem with PostgreSQL located on the first data
> center on dbserver1. All this points towards the connectivity issue between
> the two data centers.
>
> Also, apparently the network issue seems to have resolved after a couple
> of minutes, but till that time the PostgreSQL node failure was already
> registered by Pgpool-II on dbserver3 and the failover was triggered.
>
> Thanks
> Best regards
> Muhammad Usama
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.sraoss.jp/pipermail/pgpool-general/attachments/20170112/63931612/attachment.html>