On 5 August 2016 at 09:16, Tatsuo Ishii <ishii@sraoss.co.jp> wrote:
>> On 27/07/16 18:54, Chris Travers wrote:
>>> Another one I think they obliquely referred to (in the subtle problems
>>> section) was the fact that if you have longer-running queries on the
>>> replica with a lot of updates, you can get funny auto-vacuum-induced
>>> errors (writes from autovacuum on the master can interrupt queries on
>>> the slave). BTW if there is interest in what could be done for that,
>>> something which allows autovacuum to decide how long to wait before
>>> cleaning up dead tuples would be a great enhancement.
>>
>> You mean something like hot_standby_feedback?
>>
>> https://www.postgresql.org/docs/current/static/runtime-config-replication.html#GUC-HOT-STANDBY-FEEDBACK
>
> I wonder if their problem could be fixed by using
> hot_standby_feedback. I have encountered similar problem but it seems
> hot_standby_feedback was not any help in this case:
>
> https://www.postgresql.org/message-id/20130829.164457.863984798767991096.t-ishii%40sraoss.co.jp
There have been various bugs and enhancements over the years, not all
of which were backpatched.
--
Simon Riggs http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services