[Linux-HA] Don't get back resource again

[Andrew Beekhof]

On 7/19/07, Adrian Chapela <achapela.rexistros at gmail.com> wrote:
> Andrew Beekhof escribió:
> > then you want:
> >
> > default_resource_failure_stickiness=-INFINITY (if it fails, move
> > immediately), and
> >
> > default_resource_stickiness=INFINITY (dont move it unless it fails or
> > the node is shutting down or similar)
> But If I assign default_resource_stickiness=INFINITY , the resource
> group executes in the first heartbeat node available no ??

it has no relation to the initial placement (since its not yet running)

> I want to execute a resource in one node and then it network fails or
> resource fails, the resource group must be moved to the secondary node.

Right

>
> Now I assign a 200 score to a node by a rule, could be this rule the
> problem ?? The rule is:

not really a problem, but it will of course have an effect

>        <rsc_location id="my_resource:connected" rsc="MySQL_GROUP">
>          <rule id="my_resource:prefer:portatil" score="200">
>            <expression id="my_resource:prefer:portatil:expr"
> attribute="#uname" operation="eq" value="portatil"/>
>          </rule>
>          <rule id="my_resource:connected:rule" score="-INFINITY"
> boolean_op="or">
>            <expression id="my_resource:connected:expr:undefined"
> attribute="pingd" operation="not_defined"/>
>            <expression id="my_resource:connected:expr:zero"
> attribute="pingd" operation="lte" value="0"/>
>          </rule>
>        </rsc_location>
>
> (It will test your theory now..)
> >
> >
> >> To do that I configure in cib the next:
> >>
> >> default_resource_failure_stickiness=-INFINITY -> with this the resource
> >> doesn't execute in any node.
> >
> > has it failed once on every node?
> No, I think not..

well the only time default_resource_failure_stickiness has any effect
on a node's score is if the resource has failed on that node.

unless its the pingd rule that is preventing some of the nodes from
taking over the resource

> > you'll need to clear out the failcount before it will be allowed to
> > run there again
> This is to execute the resource in a node wich has a failcount > 0 not
> ?? How can I know failcount value for a node ??

crm_failcount --help