[pgpool-general: 1636] Re: inserts sent to standby

[Michael Betts]

Tatsuo,

Thank you for confirming this.  I have experienced the same behavior again.  Is there something I can do to help provide information about this problem?

Thanks,

Mike

On Apr 16, 2013, at 7:41 PM, Tatsuo Ishii wrote:

>> I have been running pgpool 3.2.3 in production and everything has been working great.  However, I recently had an incident where pgpool started sending inserts to the the read-only standby and not the master.  I'm using master/slave mode with streaming replication and both the master and slave were up and running and when I checked the pgpool status, everything seemed normal.  I'm also using delay_threshold = 100000.  When this incident occurred, it was after a large data update in which the lag exceed this threshold.  I believe that once the threshold was no longer exceeded, it started sending queries back to the standby and also started sending inserts there too.
>> 
>> Am I understanding this feature correctly?  If the lag exceeds the delay_threshold, it stops sending queries to the standby and when it drops below the threshold, it resumes.  Any idea as to why it would send inserts to the standby?  I did not have debug on, so I don't have much information to go on, except I do know that the delay_threshold was exceeded and shortly after it was no longer exceeded, this started happening.  Restarting pgpool cleared it up.
> 
> Your understanding of delay_threshold behavior is correct and sounds
> like the incident you observed is a bug. I will try to reproduce the
> problem to find out what is wrong.
> --
> Tatsuo Ishii
> SRA OSS, Inc. Japan
> English: http://www.sraoss.co.jp/index_en.php
> Japanese: http://www.sraoss.co.jp