[pgpool-general: 3153] Re: vacuum analyze caused node degeneration

[Tatsuo Ishii]

If you Google with "found 0 removable, 0 nonremovable row versions in
0 out of 0 pages", you will find lots of search results explaining why
this happen. Probably you have long running transaction(s) on node 1
and VACUUMM reclaim dead tuples.

I suggest to run VACUUM without VERBOSE option to suppress the "found
0 removal..." message, which should avoid unnecessary failover.

From: Sean Hogan <sean at compusult.net>
Subject: [pgpool-general: 3152] Re: vacuum analyze caused node degeneration
Date: Tue, 09 Sep 2014 09:04:30 -0230
Message-ID: <540EE5C6.2090905 at compusult.net>

> Does anyone have any thought on this?
> 
> Thanks,
> Sean
> 
> On 14-09-02 10:07 AM, Sean Hogan wrote:
>> Hi,
>>
>> Recently during a test event we had an unexpected node degeneration,
>> and I need to provide some reasonable explanation of why it happened.
>> Our system is two pgpool-II 3.3.3 nodes fronting two PostgreSQL 9.2.7
>> nodes.
>>
>> We selected a table in pgAdmin3 and ran VACUUM with the ANALYZE
>> option.  Auto-vacuum is enabled on the database so the manual vacuum
>> is a bit nonsensical, but I still didn't expect a failure.
>>
>> The logging tells the story:
>>
>> 2014-08-19 13:56:53 LOG: pid 7375: pool_send_and_wait: Error or notice
>> message from backend: : DB node id: 0 backend pid: 7404 statement:
>> VACUUM VERBOSE ANALYZE sch.tbl message: vacuuming "sch.tbl"
>> 2014-08-19 13:56:53 LOG: pid 7375: pool_send_and_wait: Error or notice
>> message from backend: : DB node id: 1 backend pid: 314 statement:
>> VACUUM VERBOSE ANALYZE sch.tbl message: vacuuming "sch.tbl"
>> 2014-08-19 13:56:53 ERROR: pid 7375: read_kind_from_backend: 1 th kind
>> C does not match with master or majority connection kind N
>> 2014-08-19 13:56:53 ERROR: pid 7375: kind mismatch among
>> backends. Possible last query was: "VACUUM VERBOSE ANALYZE sch.tbl"
>> kind details are: 0[N: "pg_toast_170592198": found 0 removable, 0
>> nonremovable row versions in 0 out of 0 pages] 1[C]
>> 2014-08-19 13:56:53 LOG: pid 7375: degenerate_backend_set: 1 fail over
>> request from pid 7375
>>
>> Web searches for potential causes were not helpful.  Any ideas why
>> node 1 would just say "command completed" without all the other
>> information garbage that came from node 0?
>>
>> Thanks for any assistance,
>>
>> Sean
>>
>>
>> _______________________________________________
>> pgpool-general mailing list
>> pgpool-general at pgpool.net
>> http://www.pgpool.net/mailman/listinfo/pgpool-general
>