[Linux-HA] Ethernet link down NO failover;(
Dejan Muhamedagic
dejanmm at fastmail.fm
Fri Sep 7 05:21:30 MDT 2007
On Fri, Sep 07, 2007 at 12:24:14PM +0200, Dominik Klein wrote:
> > <constraints>
> > <rsc_location id="rsc_location_group_NFS" rsc="group_NFS">
> > <rule id="prefered_location_group_NFS" score="100">
> > <expression attribute="#uname"
> >id="prefered_location_group_NFS_expr" operation="eq"
> >value="nfs00001"/>
> > </rule>
> > </rsc_location>
> > <rsc_location id="pingd:connected" rsc="group_NFS">
> > <rule id="pingd:connected:rule"
> >score_attribute="default_ping_set">
> > <expression id="pingd:connected:expr:defined"
> >attribute="default_ping_set" operation="defined"/>
>
> I think this is your problem. pingd will still run with bond0 being
> completely disconnected. So it will still set the attribute (propably to 0).
> You need to check for the value >0 as a second expression.
>
> Moreover, i dont see a score for this rule?!
That constraint is OK. The score is calculated as a value of the
score_attribute (in this case "default_ping_set"). What this
literally means is that the node with the best connectivity gets
the resource.
> > </rule>
> > </rsc_location>
> > </constraints>
>
> I think it should look something like this:
>
> <rsc_location id="connected" rsc="group_NFS">
> <rule id="connected_rule" score="-INFINITY" boolean_op="or">
> <expression id="connected-rule-1" attribute="pingd"
> operation="not_defined"/>
> <expression id="connected-rule-2" attribute="pingd"
> operation="lte" value="0"/>
> </rule>
> </rsc_location>
This one would work too, but it has a slightly different meaning:
never run a resource on the node which has no activity. Hence,
it is more restrictive than the one above. If all nodes where to
lose connectivity, the resource would not run anywhere.
> Regards
> Dominik
> _______________________________________________
> Linux-HA mailing list
> Linux-HA at lists.linux-ha.org
> http://lists.linux-ha.org/mailman/listinfo/linux-ha
> See also: http://linux-ha.org/ReportingProblems
More information about the Linux-HA
mailing list