[pgpool-general: 1376] Re: watchdog enabled delegate_IP on multiple nodes simultaneously

[Lonni J Friedman]

On Tue, Feb 5, 2013 at 6:02 PM, Yugo Nagata <nagata at sraoss.co.jp> wrote:
> On Tue, 5 Feb 2013 13:00:57 -0800
> Lonni J Friedman <netllama at gmail.com> wrote:
>
>> On Thu, Jan 31, 2013 at 10:18 PM, Yugo Nagata <nagata at sraoss.co.jp> wrote:
>> > Hello Lonni,
>> >
>> > On Fri, 2 Nov 2012 13:17:24 +0900
>> > Yugo Nagata <nagata at sraoss.co.jp> wrote:
>> >
>> >> I guess the situation you said is as follows.
>> >>
>> >> 1. Detach all the nodes from each pgpool using pcp_detach_node etc.
>> >> 2. Attach a node to each pgpool using pcp_attach_node etc.
>> >>  -> Both pgpools bring up the virtual IP.
>> >>
>> >> If so, this is a restriction of watchdog functionality, and is not bug.
>> >>
>> >> (see Restictions in
>> >>  http://www.pgpool.net/docs/latest/pgpool-en.html#watchdog )
>> >>
>> >> If all the nodes is detached from pgpool, you have to restart the pgpool.
>> >>
>> >> # The situation of raw mode you reported previously is a bug. I'll fix it.
>> >
>> > Sorry for delay. I attached a patch resolving the split-brain problem
>> > (situation that there are multiple pgpool holding delegate IP).
>> > Could you try it?
>> >
>> > In this fix, when once all backend DB nodes are detached from pgpool (by
>> > pcp_detach_node or DB server down etc.), the pgpool goes to DOWN status
>> > until this is restarted. The pgpool in DOWN status cannot escalate to
>> > delegate IP holder, so split-brain situation is avoided. All the pgpool
>> > in DOWN status should be restarted to recover correct status.
>>
>> This patch doesn't seem to apply cleanly to wd_lifecheck.c from
>> pgpool-3.2.1.  I get the following rejected hunk:
>
> Could you please apply this patch, after applying the patch attached in
> [pgpool-general: 1356] fixing segfaults on Fedora17.
>
> If it doesn't work well, try to use patch command.
>
>  $ patch < patch.file
>
> Or, you can try the snapshot on git repogitory which is available from here.
>
> http://git.postgresql.org/gitweb/?p=pgpool2.git;a=snapshot;h=a19a49dedddddbd0d3f60f1309daeba724ca1a28;sf=tgz

Ah, sorry.  I wasn't aware that the two patches were dependent upon
eachother.  I've confirmed that applying them in the correct order
worked, and I've verified that the original bug appears to be fixed.

thanks!