J. Friedrich wrote:
> Hi Michael,
> HI group,
> 
> I expanded my configuration and added the attributes
> resource-failure-stickiness and resource-stickiness to one resource of
> the "bar group" . I used "-30" for resource-failure-stickiness and
> "100" for resource-stickiness. 

In old versions it actually is "_" instead of "-".

> If I correctly understand heartbeat's
> principles, the resource should be started on the 2nd node after 3
> failures on the 1st node (the 4th problem with the resource will make
> it switch).

Sounds right.

> To simulate a problem with the resource, I stopped the service 

Assuming you have monitor operations for your resource, this should
cause a failure.

> and
> moved the binary, 

This (if your RA is implemented properly) will cause a permanent failure
by returning OCF_ERR_INSTALLED and force the resource off the node
immediately, as the program is no longer installed on the node.

> so that the monitor detects that the service is not
> running, but can not start the service again. Heartbeat start the
> other resources in the resource group, is not able to start the
> resource with the moved binary twice.... and does nothing! It does not
> switch.

I guess two things then (read above): s/-/_/ for your stickiness values
and correct your RA to correctly handle not installed software.

<snip>

> Do I have to configure Stonith if I want to use these stickiness values?

No.

Regards
Dominik
_______________________________________________
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