[pgpool-committers: 4326] Re: pgpool: Fix for bug in watchdog where sometime failover is not reliably

[Muhammad Usama]

Hi Ishii-San

Sorry I was shifting my house and was away from an email so couldn't reply
earlier.

On Tue, Oct 17, 2017 at 10:08 AM, Tatsuo Ishii <ishii at sraoss.co.jp> wrote:

> Usama,
>
> Is this the fix for?
> https://www.pgpool.net/mantisbt/view.php?id=339


Actually there was another similar issue reported for Pgpool-II 3.6 but,
The commit should also fix this problem (bug 339) for 3.5 and 3.6 branches,
But as the this bug mentions the problem is also appearing in Pgpool-II
3.3, So older branches still need a fix and . because of completely
different architecture of watchdog in versions prior to 3.5 we may need a
completely different solution.

Thanks
Best Regards
Muhammad Usama


>
> Best regards,
> --
> Tatsuo Ishii
> SRA OSS, Inc. Japan
> English: http://www.sraoss.co.jp/index_en.php
> Japanese:http://www.sraoss.co.jp
>
> From: Muhammad Usama <m.usama at gmail.com>
> Subject: [pgpool-committers: 4322] pgpool: Fix for bug in watchdog where
> sometime failover is not reliably
> Date: Mon, 16 Oct 2017 15:40:59 +0000
> Message-ID: <E1e47Vj-0007Zn-Dz at gothos.postgresql.org>
>
> > Fix for bug in watchdog where sometime failover is not reliably performed
> >
> > Currently watchdog process only considers the node's watchdog state
> before
> > deciding if it can handle the failover and failover-locking requests.
> But the
> > problem with this technique is that, for the instance when the node has
> > announced itself as a master/coordinator of the cluster and is waiting
> for
> > the standby nodes to recognise it as a Master node. For that period of
> time
> > the watchdog state of the node is Master/coordinator, but it is yet not
> fully
> > capable of handling the failover and failover-locking requests.
> > So sometimes this leads to a situation where a failover is not reliably
> > performed in case when the failover request arrives while watchdog
> cluster
> > is in the process of electing a leader.
> >
> > The fix for the above situation is to make sure that the node has fully
> acquired
> > the Master status before accepting the failover and failover-locking
> requests.
> >
> > Branch
> > ------
> > V3_6_STABLE
> >
> > Details
> > -------
> > https://git.postgresql.org/gitweb?p=pgpool2.git;a=commitdiff;h=
> f6ac53063da219bf3b916d6e5bbeec7a00283a77
> >
> > Modified Files
> > --------------
> > src/watchdog/watchdog.c | 12 +++++++++---
> > 1 file changed, 9 insertions(+), 3 deletions(-)
> >
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.sraoss.jp/pipermail/pgpool-committers/attachments/20171018/005899b6/attachment.html>