[ClusterLabs] Q: What is the default default-resource-stickiness? ; -)
Ken Gaillot
kgaillot at redhat.com
Tue Sep 4 14:02:14 EDT 2018
On Mon, 2018-09-03 at 13:57 +0200, RaSca wrote:
> On 9/3/18 1:14 PM, Ulrich Windl wrote:
> > Hi!
> > I wonder: If the default-resource-stickiness is not set, and no
> > stickiness for a resource is set, what is the stickiness of that
> > resource? Meaning what is the default for default-resource-
> > stickiness?
> > In case it's 0, why does "crm_attribute --query --name=default-
> > resource-stickiness --type=rsc_defaults --quiet" exit with code 6
> > then?
> > Regards,
> > Ulrich
>
> Hi Ulrich,
> default value is not set, or null, if you prefer:
>
> $ sudo pcs property show --all | grep stickiness
> default-resource-stickiness: (null)
>
> But the error 6 you get from the return code of crm_attribute applies
> to
> any other property valued with "null" or whatever value.
I'd add one point: The default-resource-stickiness cluster property has
been deprecated for a decade (along with the even older
default_resource_stickiness, default-resource-failure-stickiness, and
default_resource_failure_stickiness), and is removed as of 2.0.0.
The "modern" way is to put resource-stickiness in the rsc_defaults
section.
The default stickiness if unspecified is 0, except for clone instances,
which get a default stickiness of 1 so that (identical) instances don't
constantly shuffle from one node to another.
> In this specific case null and zero might count the same. This means
> that the general (or default) stickiness will make each resource come
> back on the node it was running on:
>
> 1 - RES1 is running on NODE1
> 2 - NODE1 fails
> 3 - RES1 move to NODE2
> 4 - NODE1 resurrect
> 5 - RES1 go back to NODE1
>
> with a default-resource-stickiness of INFINITY you'll not get step 5,
> with any other value the cluster will deal with node weights while
> deciding to move back or not the resource.
>
> Hope this helps,
>
--
Ken Gaillot <kgaillot at redhat.com>
More information about the Users
mailing list