On 7/16/07, Andrew Beekhof <[EMAIL PROTECTED]> wrote:
On 7/16/07, Serge Dubrouski <[EMAIL PROTECTED]> wrote:
>
> On 7/16/07, Andrew Beekhof <[EMAIL PROTECTED]> wrote:
> > On 7/16/07, Maxim Veksler <[EMAIL PROTECTED]> wrote:
> > >
> > > Hello list,
> > >
> > > I'm configuring a cib, such as the once provided below.
> > > I see that when I make one of the resources fail, the count reaches to
> > > the sum of all resources. In the provided cib a resource can fail up
> > > to 15 times before it migrates to a second node. I would to have a
> > > fixed "stickiness" value for each resource - that is I would like each
> > > of them to a maxim fail count of 5. How should I be configuring my cib
> > > ?
> >
> >
> >
> > if the current setup will fail 15 times before moving to the other node
> and
> > you want it to fail only 5 times... try multiplying
> > default_resource_failure_stickiness
> > by 3 :-)
>
> Does it work like that for groups? As far as I remember standard
> algorithm for failcount wasn't applicable for groups. Did it get
> fixed?
>
Regardless of what the algorithm is, if the observed is 3x more than what is
desired, then multiplying the failure stickiness by 3 will always give the
desired result.
I can confirm that the "aggregation" of resource_stickiness is in
effect for either resources "grouped" together or for resources that
are chained together by a collocation constraint.
Doing the math an dividing default_resource_stickiness would do the
trick, that's true but it's not the expected behaviour as far as I can
tell.
I'm just learning the system, I play to allow dynamic setup of
resources (for our setup) and doing the math for
default_resource_stickiness is just another pitfall I'd rather avoid
(I actually means re parsing the cib from my script.
_______________________________________________
Linux-HA mailing list
[email protected]
http://lists.linux-ha.org/mailman/listinfo/linux-ha
See also: http://linux-ha.org/ReportingProblems
--
Cheers,
Maxim Veksler
"Free as in Freedom" - Do u GNU ?
_______________________________________________
Linux-HA mailing list
[email protected]
http://lists.linux-ha.org/mailman/listinfo/linux-ha
See also: http://linux-ha.org/ReportingProblems
