Re: Hot standby and b-tree killed items

On Tue, 2008-12-23 at 23:59 -0500, Robert Treat wrote:
> On Friday 19 December 2008 19:36:42 Simon Riggs wrote:
> > Perhaps we should listen to the people that have said they don't want
> > queries cancelled, even if the alternative is inconsistent answers. That
> > is easily possible yet is not currently an option. Plus we have the
> > option I referred to up thread, which is to defer query cancel until the
> > query reads a modified data block. I'm OK with implementing either of
> > those, as non-default options. Do we need those options or are we ok?
> >
> 
> Haven't seen any feed back on this, but I think the two options of cancel 
> query for replay, and pause replay for queries, are probably enough for a 
> first go around (especially if you can get the query canceling to work only 
> when changes are made to the specific database in question)

Thanks for picking up on this. This question is the #1 usability issue
for Hot Standby, since at least May 2008. There are many potential
additions and we need to track this carefully over the next few months
to see if we have it just right. I'll take viewpoints at any time on
that; this door is never closed, though tempus fugit.

Greg and Heikki have highlighted in this thread some aspects of btree
garbage collection that will increase the chance of queries being
cancelled in various circumstances. If this is important enough to
trigger additional actions then we need to highlight that now so we have
time to take those corrective actions. 

I've listened to many different viewpoints on and off list. Everybody
takes a slightly different angle on it and I'm in favour of giving
everybody what they want with the right set of options.

-- Simon Riggs           www.2ndQuadrant.comPostgreSQL Training, Services and Support