[Pacemaker] stonith resources external/ipmi, where should they run on?

Carlos Xavier cbastos at connection.com.br
Mon Sep 24 16:45:34 EDT 2012


Hi.

First of all I wish to say tank you to everyone who helped on this topic. 
Taking a look at the
logs it seams everthing is working fine, stonith is able to read the status 
of the other machine.

But reading the stonith man page a doubt has arised. I tryed to test the 
configuration by hand using
# stonith -t external/ipmi ipaddr=172.31.0.240 userid=root 
password=somepass -S

and some errors got displayed
external/ipmi[15985]: ERROR: ipaddr, userid or passwd missing; check 
configuration
external/ipmi[15982]: ERROR: error executing ipmitool:
WARN: external_status: 'ipmi status' failed with rc 1
ERROR: external/ipmi device not accessible.

Although if I run the ipmitool command it works fine

# /usr/bin/ipmitool -I lan -U root -P genese -H 172.31.0.240 chassis power 
status
Chassis Power is on

Am I missing something?

Best regards,
Carlos.

On Sunday, September 23, 2012 4:39 PM
Volker Dormeyer <volker at ixolution.de> wrote:
> Hi
>
> On Fri, Sep 21, 2012 at 07:35:35PM -0300,
> Carlos Xavier <cbastos at connection.com.br> wrote:
> > Hi.
> >
> > I´m running a Pacemaker Cluster on Dell R610 machines, so I enabled
> > the IDrac6 and configured the following external/ipmi resources
> > primitive resIPMI-1 stonith:external/ipmi \
> >        params hostname=apolo ipaddr=172.31.0.240 \
> >        userid=root passwd=somepass interface=lan \
> >        op monitor interval=600 timeout=240
> > primitive resIPMI-2 stonith:external/ipmi \
> >        params hostname=diana ipaddr=172.31.0.241 \
> >        userid=root passwd=somepass interface=lan \
> >        op monitor interval=600 timeout=240
> >
> > When the resources are started i see one running at each host:
> > resIPMI-1       (stonith:external/ipmi):        Started apolo
> > resIPMI-2       (stonith:external/ipmi):        Started diana
> >
> > But as it can be noticed the resource resIPMI-1 that controls the
> > node 172.31.0.240 is running at its own host and so is the
> > resIPMI-2.
> >
> > Is that correct or should them be running on the oposite host or be
> > a clone resource? Do I need to set the location for those resources?
>
> In general, it depends on the fence-device and agent. If the
> fence-device can control both nodes, you can drive a cloned resource. In
> this case the parameters pcmk_host_list and pcmk_host_map, etc. for
> stonithd might be of interest for you.
>
> In your case, I think one device can control a single host only, which 
> means
> the resource for node 1 should run on node 2 and the one for node 2 should
> run on node 1. This implies, that you need to set the appropriate location
> contrains to prevent the situation you described, above. It could be
> like this:
>
>   location locIPMI-1 resIPMI-1 -inf: apolo
>   location locIPMI-2 resIPMI-2 -inf: diana
>
> Best Regards,
> Volker
> 




More information about the Pacemaker mailing list