On Mon, Nov 20, 2017 at 1:23 PM, Simon Riggs <simon@2ndquadrant.com> wrote:
> I believe you just made up "fake news" about 2ndQuadrant.
>
> Please explain your source of this information, or retract and apologise now.
I also think that JD misrepresented 2ndQPostgres, which I gather is a
set of backpatches for each stable release (it would be nice if that
was open source, but that's beside the point).
I think it's reasonable for a small number of users to want to get
this or that critical performance feature without upgrading major
version. I did this on an ad-hoc basis for 2ndQuadrant several years
ago, usually for customers that really needed it.
Isn't this kind of customization a big advantage of open source? When
a user is screaming for smoother I/O from checkpoints on their
enormous mission critical database, for example, are you really going
to tell them that they're wrong for wanting something like this? That
can amount to real downtime.
Some full upgrades require a lot of planning, on account of changes to
the catalogs and possibly on disk representation, but a lot of things
don't need that. We've made huge progress in the last number of
releases on performance, often by adding things that don't care about
on-disk representation at all. While we as a community are better at
judging risk than users collectively, that isn't necessarily true of
individual users. I find it easy to believe that something like
2ndQPostgres could make sense in a small number of individual special
cases, where there are already real problems, and the user really
knows what they're doing.
--
Peter Geoghegan