[Linux-cluster] Clusvcadm doesn't behave as it should

[Agnieszka Kukałowicz]

> Subject: Re: [Linux-cluster] Clusvcadm doesn't behave as it should
> 
> On Thu, 2008-03-06 at 15:15 +0100, Agnieszka Kukałowicz wrote:
> 
> > Clusvcadm -e httpd_w11 -F on w12.local or w21.local causes on all
nodes
> > error:
> > Mar  6 13:36:59 w11 clurgmgrd[2176]: <crit> Watchdog: Daemon died,
> > rebooting...
> > After that all hosts w11.local, w12.local, w21.local are rebooted.
> >
> > I have installed rgmanager-2.0.31-1.el5 and the cluster runs on the
Xen
> > guest.
> 
> https://bugzilla.redhat.com/show_bug.cgi?id=430220

It works ok after installing rgmanager-2.0.31-1.el5.bz430272. There is
no critical error. 

But I still not sure if the command "enable service" in Conga works
properly in restricted failover domain. Doing that for service with
restricted failover domain returns 

"A ricci error occurred on w12.local:11111: clusvcadm failed to start
httpd_w11..."

I think if a service is configured to run only in restricted failover
domain the command "enable service" in Conga should check the failover
domain for that service and start the service in that domain. That not
happens. Conga tries to start service on the local node (in this example
w12.local) instead of starting the service in its restricted failover
domain (in my example in w11.local). 

This behavoiur of Conga is a little bit confusing (or am I wrong?)
because I should remember what are failover domain for my service and to
enable service I need to run command "Start this service on ..." and put
in this place the name of the host from failover domain.

Cheers
Agnieszka Kukalowicz