[Linux-HA] about the resource stickiness

[Andrew Beekhof]

On 10/9/06, TORRESANI, Roberto <roberto.torresani at unitn.it> wrote:
>
>
> >when you can post your CIB i'll have a look
>
> Thank you! ;-)
>
> The whole picture is this one:
> - ha 2.07
> - 2 physical nodes that normally will run the resources (ernesto,
> marcos)
> - 1 virtual node (karl running on vmware) that would run resources only
> if requested by the operator (with a crm_resource -M, eventually with a
> -M). In case both the virtual nodes fail, the virtual one would have to
> takeover all the res.
>
> And those the problems:
> 1) node score calc. (especially for the mysql_grp + apache_grp which are
> colocated by a constraint)
>  On ernesto it takes 11 (actually crm_failcount reports 10, but it's 11,
> I haven't applied any patche on the 2.0.7 base release)

the latest code in crm-stable produces the numbers and results you would expect.

> 2) If I take a physical node out of the cluster and a resource exceed
> the failcount score to run on it, the resource is NOT migrated on karl
> (virtual). Can you please give me some advise on it, too?

thats because as soon as "is_virtual" is set, karl always has a node
score of -1000.
any node with a score less than zero cannot run that resource.