Re: WIP: long transactions on hot standby feedback replica / proof of concept

On Mon, Sep 4, 2017 at 4:51 PM, Alex Ignatov <a(dot)ignatov(at)postgrespro(dot)ru>
wrote:

> In this situation this parameter (max_standby_streaming_delay) wont help
> because if you have subsequent statement on standby (following info is from
> documentation and from our experience ): Thus, if one query has resulted in
> significant delay, subsequent conflicting queries will have much less grace
> time until the standby server has caught up again. And you never now how to
> set this parameter exept to -1 which mean up to infinity delayed standby.
>
> On our experience only autovacuum on master took AccesExclusiveLock that
> raise this Fatal message on standby. After this AccessExclusive reached
> standby and max_standby_streaming_delay > -1 you definitely sooner or
> later get this Fatal on recovery .
> With this patch we try to get rid of AccessEclusiveLock applied on standby
> while we have active statement on it.
>

+1,
Aborting read-only query on standby because of vacuum on master seems to be
rather unexpected behaviour for user when hot standby feedback is on.
I think we should work on this problem for v11. Backpatching documentation
for previous releases would be good too.

------
Alexander Korotkov
Postgres Professional: http://www.postgrespro.com
The Russian Postgres Company