Re: Re: BUG #13685: Archiving while idle every archive_timeout with wal_level hot_standby

On 2016-02-11 09:25:30 +0530, Amit Kapila wrote:
> On Wed, Feb 10, 2016 at 1:42 PM, Michael Paquier <michael.paquier@gmail.com>
> wrote:
> >
> > On Wed, Feb 10, 2016 at 5:00 PM, Amit Kapila <amit.kapila16@gmail.com>
> wrote:
> > > Okay, but isn't it better that we remove the snapshot taken
> > > at checkpoint time in the main branch or till where this code is
> > > getting back-patched.   Do you see any need of same after
> > > having the logging of snapshot in bgwriter?
> >
> > But this one is necessary as well to allow hot standby faster to
> > initialize, no? Particularly in the case where a bgwriter snapshot
> > would have been taken just before the checkpoint, there may be up to
> > 15s until the next one.
> >
> 
> It could be helpful if checkpoint is done at shorter intervals, otherwise
> we anyway log it at 15s interval and if need faster initialisation
> of hot-standby, then it is better to reduce the log snapshot interval
> in bgwriter.

No. By emitting the first snapshot directly after the determination of
the redo pointer, we can get into STANDBY_SNAPSHOT_READY more
quickly. Especially if some of the snapshots are overflowed. During
startup 15s can be a *long* time; but on the other hand there's not much
benefit at logging snapshots more frequently when the system is up.

I don't think we should tinker with the frequency/logging points, while
fixing the issue here.