On Wed, Aug 13, 2008 at 20:51, Eric Blau <[EMAIL PROTECTED]> wrote:
> * Andreas Kurz <[EMAIL PROTECTED]> [2008-08-13 14:34 -0400]:
>> >
>> > I tried the crm_resource command, but it adds a permament location
>> > constraint to the CIB.  I was trying to avoid that.
>> >
>> > For example, with a resource named IPaddr2_1, when I ran that command I
>> > had this constraint added to my CIB:
>> >
>> > <constraints>
>> >     <rsc_location id="cli-prefer-IPaddr2_1" rsc="IPaddr2_1">
>> >         <rule id="cli-prefer-rule-IPaddr2_1" score="INFINITY">
>> >             <expression id="cli-prefer-expr-IPaddr2_1" attribute="#uname"
>> > operation="eq" value="cmlab8" type="string"/>
>> >         </rule>
>> >     </rsc_location>
>> > </constraints>
>> >
>> >
>> > crm_standby also adds a permanent instance attribute forcing the node to
>> > standby until the attribute is deleted or set to false.
>>
>> try 'crm_resource -U -r IPaddr2_1'
>
> I'm trying to script this to do this in an automated fashion when two
> servers have scores within the stickiness threshold.  The user may want
> to push the resource to another server without affecting the score

You can't.  Adjusting scores is the only way to tell the cluster where
to put resources.

> if
> the scores are within the stickiness threshold.
>
> With crm_resource how long should I wait before running crm_resource
> again to remove the INFINITY scored constraint?  What happens if the
> scores of my resources change due to some monitor action while this
> INFINITY constraint is added?  I would want the failure to override my
> movement of the resource.  After all the point is just to tip the
> resource over to another equivalent server since the scores are both
> within the stickiness threshold.
>
> I don't want an INFINITY scored constraint ever entered into the CIB or
> any other effect on the score if possible.  Is this possible?
>
> Thanks,
> Eric
>
> _______________________________________________
> Linux-HA mailing list
> [email protected]
> http://lists.linux-ha.org/mailman/listinfo/linux-ha
> See also: http://linux-ha.org/ReportingProblems
>
_______________________________________________
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