Tobias Appel wrote:
> On Mon, 2009-01-26 at 13:04 +0100, Andrew Beekhof wrote:
>> On Mon, Jan 26, 2009 at 12:10, Tobias Appel <[email protected]> wrote:
>>> Well I've got a lot of questions today as you can see :)
>>>
>>> I have a group of resources which is ordered and colocated (due to drbd
>>> master / slave constraint). I added a monitor operation to nearly all
>>> members of this group with a on_fail restart setting.
>>> Now I'm wondering do I have to add resource_stickiness for each resource
>>> or just for the group?
>>> And how is it calculated if I add it to the group. Right now I have a
>>> resource_stickiness with a value of 100 and a failure_stickiness
>> Don't go there.  Seriously.  We made a big mistake implementing it like that.
>> Get the latest version of Pacemaker, look up migration-threshold, and
>> pretend you've never heard the phrase "failure stickiness".
> 
> Well now you got me. I'm running Heartbeat 2.1.4 I finally understood
> most of the configuration but I did try to move to pacemaker once and it
> was horrible because I could not even get a single resource to work. I
> really don't want to upgrade now, unless you are telling me that it will
> not work with my current version. The link Michael send me made sense to
> me.

You can propably make it work with your version, but you can propably
not stick to that doc Michael linked. We had 2.1.3 when I started
writing that ScoreCalculation page and there have been lots of changes
in the whole scoring system, especially in groups. I can honestly not
remember for which version that information is actually correct as it
changed quite a few times (due to numerous bug reports).

It is totally worth learning pacemaker 1.0, which gets rid of all the
(quoting Andrew) "failure-stickiness nonsense". Way easier to configure
and understand.

Read the configuration explained pdf and check out the crm shell. You'll
not regret it. http://www.clusterlabs.org/wiki/Documentation

Regards
Dominik
_______________________________________________
Linux-HA mailing list
[email protected]
http://lists.linux-ha.org/mailman/listinfo/linux-ha
See also: http://linux-ha.org/ReportingProblems

Reply via email to