[Linux-HA] Absolute values of stickiness parameters

[Dominik Klein]

Hi

Zakh, Rami wrote:
> Hi,
> 
> Sorry if i am being redundant here, but i could not locate a clear-cut answer to this query.
> 
> As far as i understand, a resource will be failed over and declared unable to run on a given node as soon as the multiply of its "failcount" and its "resource failure stickiness" value exceeds (or is equal?? can you please clarify?) its "resource stickiness value".

A resource is able to run on a node if the cluster computes a positive
score for that particular resource/node tuple.

> However, i still do not fully understand the importance of the absolute values of the stickiness parametes. For example, if i want a resource to be failed over and declared invalid on the node it is running on after a third failure, i can set "resource stickiness value" to 20 and "resource failure stickiness" to 7, but also to 200 and 67 resepectively. In this example, is there a difference between the value pairs (20,7) and (200,67)? Kindly clarify the importance of the absolute values to me.

As soon as the score becomes negative, the resource is not allowed on
that node. Whether that's -1 or -15412 or -241532 does not matter.

You will not have to worry about failure stickiness if you upgrade to
pacemaker 1.0.x. Instead of the failure stickiness, you can just specify
a "migration-threshold" of (eg) 2, which translates to: after 2
failures: do not allow the resource here anymore and move it somewhere else.

Read "migrating due to failure" in
http://clusterlabs.org/mediawiki/images/f/fb/Configuration_Explained.pdf

Regards
Dominik