On 25/06/14 12:45 PM, Riccardo Murri wrote:
Hello,
we are using IPMI to fence off nodes in the cluster (yes, we know it's
a bad idea but we do not yet have proper STONITH HW so we have to
sitck with it).
IPMI is just fine as a fence method. It's probably the most commonly
used one, actually. The only trick is to understand that IPMI draws its
power from the host, so it will fail if power is entirely lost to the
node. For that, I use stonith topology to order IPMI -> switched PDUs.
http://clusterlabs.org/wiki/Fencing_topology
We would like to co-locate STONITH with a functional ethernet
interface on the IPMI network (`eth0.617`). So we have the following
rules::
Unless I am wrong (and I am not a pacemaker expert), you no longer need
to tell stonith where to run (on pacemaker 1.1.10+, anyway).
primitive ipmi_net_up ethmonitor \
params interface=eth0.617 name=ipmi_net_up \
op monitor interval=5s timeout=60s \
op start interval=0 timeout=60s \
op stop interval=0
clone ipmi_net_up_clone ipmi_net_up \
meta globally-unique=false ordered=false notify=false
interleave=true clone-node-max=1
colocation stonith-lustre-oss1-with-ipmi inf: stonith-lustre-oss1
ipmi_net_up_clone
colocation stonith-lustre-oss2-with-ipmi inf: stonith-lustre-oss2
ipmi_net_up_clone
...
colocation stonith-lustre-oss8-with-ipmi inf: stonith-lustre-oss8
ipmi_net_up_clone
Now the ethmonitor probe is successful::
Jun 25 18:34:59 lustre-mds1 crmd[3255]: notice:
process_lrm_event: Operation ipmi_net_up_monitor_5000: ok
(node=lustre-mds1.ften.es.hpcn.uzh.ch, call=725, rc=0, cib-update=392,
confirmed=false)
However, STONITH resources appear as "Stopped", and `crm_simulate`
says they cannot be located anywhere::
# crm_simulate -sL | fgrep stonith
stonith-lustre-oss1 (stonith:fence_ipmilan): Stopped
stonith-lustre-oss2 (stonith:fence_ipmilan): Stopped
stonith-lustre-oss3 (stonith:fence_ipmilan): Stopped
stonith-lustre-oss4 (stonith:fence_ipmilan): Stopped
stonith-lustre-oss5 (stonith:fence_ipmilan): Stopped
stonith-lustre-oss6 (stonith:fence_ipmilan): Stopped
stonith-lustre-oss7 (stonith:fence_ipmilan): Stopped
stonith-lustre-oss8 (stonith:fence_ipmilan): Stopped
native_color: stonith-lustre-oss1 allocation score on
lustre-mds1.ften.es.hpcn.uzh.ch: -INFINITY
native_color: stonith-lustre-oss1 allocation score on
lustre-mds2.ften.es.hpcn.uzh.ch: -INFINITY
native_color: stonith-lustre-oss1 allocation score on
lustre-oss1.ften.es.hpcn.uzh.ch: -INFINITY
native_color: stonith-lustre-oss1 allocation score on
lustre-oss2.ften.es.hpcn.uzh.ch: -INFINITY
native_color: stonith-lustre-oss1 allocation score on
lustre-oss3.ften.es.hpcn.uzh.ch: -INFINITY
native_color: stonith-lustre-oss1 allocation score on
lustre-oss4.ften.es.hpcn.uzh.ch: -INFINITY
native_color: stonith-lustre-oss1 allocation score on
lustre-oss5.ften.es.hpcn.uzh.ch: -INFINITY
native_color: stonith-lustre-oss1 allocation score on
lustre-oss6.ften.es.hpcn.uzh.ch: -INFINITY
native_color: stonith-lustre-oss1 allocation score on
lustre-oss7.ften.es.hpcn.uzh.ch: -INFINITY
native_color: stonith-lustre-oss1 allocation score on
lustre-oss8.ften.es.hpcn.uzh.ch: -INFINITY
native_color: stonith-lustre-oss2 allocation score on
lustre-mds1.ften.es.hpcn.uzh.ch: -INFINITY
native_color: stonith-lustre-oss2 allocation score on
lustre-mds2.ften.es.hpcn.uzh.ch: -INFINITY
native_color: stonith-lustre-oss2 allocation score on
lustre-oss1.ften.es.hpcn.uzh.ch: -INFINITY
...
What are we doing wrong?
Thanks,
Riccardo
--
Digimer
Papers and Projects: https://alteeve.ca/w/
What if the cure for cancer is trapped in the mind of a person without
access to education?
_______________________________________________
Pacemaker mailing list: Pacemaker@oss.clusterlabs.org
http://oss.clusterlabs.org/mailman/listinfo/pacemaker
Project Home: http://www.clusterlabs.org
Getting started: http://www.clusterlabs.org/doc/Cluster_from_Scratch.pdf
Bugs: http://bugs.clusterlabs.org
