Hi,

On Tue, Apr 20, 2010 at 03:33:14PM +0200, Lars Marowsky-Bree wrote:
> On 2010-04-20T08:20:57, Florian Haas <[email protected]> wrote:
> 
> > > And the first two _are_ still used/maintained for the SLES10
> > > packages.
> > Which means they'll _never_ even install a resource-agents package
> > without risking to lose support. Why do we need to carry those RAs in
> > future upstream releases of resource-agents then?
> 
> Well, first, I'm entertaining the notion to upgrade SLES10 to a recent
> package build. Second, I was just pointing out that they are still being
> used by some, and I very much object to the notion to dropping RAs in
> minor releases with 1-6 months lead time.

Agreed. That time is too short. More, the warnings we had so far
were not sufficient. And there are certainly people who didn't
upgrade for a long time and perhaps won't in a long time. If
their clusters are running fine, I guess that people are
reluctant to do the upgrades anyway, until really forced.

> I find the policy per se inacceptable, sorry. This is _enterprise_ code,
> not J. Random Project people install for fun. Anything that is
> user-visible and has the potential to bring down configurations that
> work today but wouldn't after the update is a no-go.

Agreed again, though I'm not sure if we've always had the same
attitude.

> > > The LinuxSCSI RA is, to the best of my knowledge, not broken either, and
> > > does something entirely different from either SCSI reservations or sfex.
> > But is has BETA written all over it, and you were on the CC list of the
> > bug where we agreed that these should be deprecated and removed from
> > future releases. Why are you complaining now?
> > 
> > See comment
> > http://developerbugs.linux-foundation.org/show_bug.cgi?id=2244#c1
> > specifically. You could have corrected us _then_.
> 
> I _miss_ things. People do. If even I missed it, what is your guess how
> many users will not have noticed anything?
> 
> > > drbd - conceded, but shouldn't there then be a symlink to the new
> > > version (which will do work in 99% of all cases unless someone used my
> > > ill-conceived floating peer support)?
> > Fine with the idea of the distro doing that via RPM magic.
> 
> This is _irresponsible_ to our users of the community packaging. I'm at
> a loss for words.
> 
> > Aw c'mon, now don't _you_ start _me_ on rolling upgrades!
> 
> Yes, I do. Because I it is extremely painful if it is not possible.
> Trust me; I'm on the receiving end of those scenarios. We got away with
> it once from SLES10 to SLE11, because the jump is too massive
> technically for anything else to work, but here we are talking "x.y.z vs
> x.y.z+1".
> 
> Even pacemaker 0.6 to 1.0 is mostly automatic, unless the XSLT transform
> fails. (Which, except for one or two bugs, it hasn't.) Even then the
> cluster configuration was auto-converted.
> 
> > Besides like the other Lars said: welcome to maintenance mode.
> 
> If it is that simple, it should be automatic. We should not piss of our
> users.

I think that we need a different approach to the matter.

Cheers,

Dejan

> Regards,
>     Lars
> 
> -- 
> Architect Storage/HA, OPS Engineering, Novell, Inc.
> SUSE LINUX Products GmbH, GF: Markus Rex, HRB 16746 (AG Nürnberg)
> "Experience is the name everyone gives to their mistakes." -- Oscar Wilde
> 
> _______________________________________________________
> Linux-HA-Dev: [email protected]
> http://lists.linux-ha.org/mailman/listinfo/linux-ha-dev
> Home Page: http://linux-ha.org/
_______________________________________________________
Linux-HA-Dev: [email protected]
http://lists.linux-ha.org/mailman/listinfo/linux-ha-dev
Home Page: http://linux-ha.org/

Reply via email to