[Pacemaker] stonith-ng and old stonithd in pcmk 1.1

Andrew Beekhof andrew at beekhof.net
Thu Feb 11 15:51:46 EST 2010

On Thu, Feb 11, 2010 at 6:47 PM, Dejan Muhamedagic <dejanmm at fastmail.fm> wrote:
> Hi,
> On Thu, Feb 11, 2010 at 02:12:31PM +0100, Andrew Beekhof wrote:
>> On Thu, Feb 11, 2010 at 1:08 PM, Dejan Muhamedagic <dejanmm at fastmail.fm> wrote:
>> > Hi Andrew,
>> >
>> > Though you haven't made any announcement, it seems like the old
>> > stonithd is removed from Pacemaker 1.1 and replaced by a brand
>> > new stonith daemon. That's most probably the way to go, but I'm
>> > worried because the new code didn't see much field testing.
>> Which is why we're not pushing it as a stable release yet...
>> Though I have spent quite some time testing it myself, both manually
>> and via cts.
>> The code has already been through a number of iterations, I'm quite
>> confident in it.
> I somehow learned never to be confident until the code went out
> and has actually been used by others, not only me :)

I'm not claiming its perfect, but I am reasonably confident that I've
ironed out the major problems.
CTS does a pretty good job of exercising fencing.

>> > Is there a way to keep the old stonithd and allow users to choose
>> > which stonith daemon to use at runtime?
>> Nope.  People concerned by this should continue using 1.0 (which will
>> remain the current "frozen" stable series until 1.2 comes out) and
>> wait until others report its ok.
> That unfortunately leaves SLES11SP1 in an awkward situation,
> since it seems like we'll use v1.1.

Lars and I had a conversation about this a while back, he asked
questions, I gave honest answers.

But you have QA people right? And presumably a few betas before you
release SP1 right? And you've been testing it already right?
So what's the problem?  If its falling in a heap, you guys are being
mighty quiet about it.

Besides, this is what enterprise distros are all about, taking
upstream code and polishing it until its ready for paying customers.
They pay for the reassurance that it hasn't just been tested by the
idiot that wrote it.

More information about the Pacemaker mailing list