Re: Remove Deprecated Exclusive Backup Mode

Greetings,

* Robert Haas (robertmhaas@gmail.com) wrote:
> On Mon, Feb 25, 2019 at 8:42 AM Stephen Frost <sfrost@snowman.net> wrote:
> > Alright, then how about we provide a bit of help for everyone who's got
> > a system built around recovery.conf today, instead of just completely
> > ripping that out?
> >
> > If we're happy to do that then I really can't agree with these arguments
> > that there's things we should try to maintain when it comes to
> > interfaces, and that's far from the only example of a pretty massive
> > change that just went into version X with zero notice to users about
> > what they're using today being deprecated.
>
> It's not the same thing.  First of all, the recovery.conf changes have
> been under discussion for quite a few years now, whereas this change
> has been under discussion for only a few months.

This argument doesn't work for me because it doesn't make *any*
difference to our users- the vast majority of whom don't follow
-hackers.  Our users will simply see that in v12 the system they've
built around recovery.conf won't work and they'll have to learn the new
way and adjust everything to deal with it.

If that argument did matter, we could go back and find the prior
discussions about the issues around the exclusive backup mode and about
removing it, or next year we could point to this thread about it, or the
year after, and say "well, we talked about it a lot, so now let's just
do it", but it all ends up looking the same to our users unless we
actually write into some kind of user-facing documentation or WARNINGs
or similar that something is going away.

> It is also worth
> noting that those recovery.conf changes would've been made years ago
> if people hadn't objected to breaking backward compatibility; people
> have just as much of a right to object to these changes on those
> grounds as they did to those changes.  Second, the recovery.conf
> changes are intended to provide a tangible benefit - namely, getting
> the recovery.conf settings into the GUC system, so that they can be
> queried and changed using methods available for other GUCs.  But the
> change proposed here breaks things for people who are using the
> mechanism in question and it doesn't give them anything else instead.

This isn't a fair argument either because we're having this *after* the
new API was implemented for backup- namely non-exclusive mode, which
*did* give people something better to use instead.

If we went back and added back into the v12 branch support for a
recovery.conf file, then we'd be having exactly this same argument next
year about if we should remove the recovery.conf file support or leave
it in- and you could make the same argument then that removing the
recovery.conf support doesn't give the user anything tangible to replace
it with- because that would have already been done.

> Indeed, it seems quite obvious that you don't have the LEAST interest
> in putting any effort into actually making things in this area better
> for users; it appears that you just want to break what they're doing
> now and shove the work of coping with those changes onto them.

What will make things better for users is actually moving to the new API
that's been available for years now, so that they can avoid the pitfalls
with the old API.

> Christophe is quite right to question whether that will cause users
> not to upgrade.  I think it's overtly user-hostile.  You have referred
> to the effort of maintaining this code, but you haven't pointed to any
> tangible improvement that you'd like to make in this area that is
> actually being blocked by the existence of exclusive backup mode.

I disagree quite a bit with this statement- the existing documentation
is absolutely horrid and needs to be completely ripped out and rewritten
and maintaining the exclusive backup mode in such a rewrite would
absolutely be a *lot* of additional work.

We actually took a shot at trying to improve the documentation while
continuing to cover both the exclusive and the non-exclusive mode and
it's hugely painful.

> It's all just griping about how terrible the whole thing is, and it
> seems that there are actually quite a few people here who find it less
> terrible than you think it is.

I used to be one of those people.  I know that it looks fine and it
certainly seems appealing but having gone through bad experiences with
it, and seen others stumble through those same experiences time and time
again, I've learned that it really is an issue, and one that I would
very much like to avoid causing future users to stumble over.

Thanks!

Stephen