[Linux-cluster] fence_ipmilan issue

[jim parsons]

On Fri, 2007-11-02 at 13:07 -0500, Stewart Stafford wrote:
> Hey all,
> 
> I have a 3 node cluster consisting of Sun X2200 M2's.  I am using the
> ELOM port for fencing the nodes.  The problem I am having is when I
> manually fence a node by calling fence_ipmilan with the reboot option,
> the service running on that node is stopped and never migrates to
> another node.  However, if I issue a reboot on the node running the
> service, the service migrates successfully.  Attached is the
> cluster.conf and below is a snippet of the messages log.
> 
> Thanks,
> 
> Stew
Try setting lanplus="1" in the device blocks. It is currently set to "".

How are you configuring this conf file? With system-config-cluster?
Conga? Vi?

-J

> 
> cluster.conf
> ===========================
> <?xml version="1.0"?>
> <cluster alias="ices_nfscluster" config_version="78" name="nfs_cluster">
>         <fence_daemon post_fail_delay="20" post_join_delay="3"/>
>         <clusternodes>
>                 <clusternode name="isc0" nodeid="1" votes="1">
>                         <fence>
>                                 <method name="1">
>                                         <device lanplus="" name="iisc0"/>
>                                 </method>
>                         </fence>
>                 </clusternode>
>                 <clusternode name="isc1" nodeid="2" votes="1">
>                         <fence>
>                                 <method name="1">
>                                         <device lanplus="" name="iisc1"/>
>                                 </method>
>                         </fence>
>                 </clusternode>
>                 <clusternode name="isc2" nodeid="3" votes="1">
>                         <fence>
>                                 <method name="1">
>                                         <device name="iisc2"/>
>                                 </method>
>                         </fence>
>                 </clusternode>
>         </clusternodes>
>         <cman/>
>         <fencedevices>
>                 <fencedevice agent="fence_ipmilan" auth="none"
> ipaddr="172.16.158.159" login="root" name="iisc0" passwd="changeme"/>
>                 <fencedevice agent="fence_ipmilan" auth="none"
> ipaddr="172.16.158.160" login="root" name="iisc1" passwd="changeme"/>
>                 <fencedevice agent="fence_ipmilan" auth="none"
> ipaddr="171.16.158.161" login="root" name="iisc2" passwd="changeme"/>
>         </fencedevices>
>         <rm>
>                 <failoverdomains>
>                         <failoverdomain name="fotest" ordered="1"
> restricted="1">
>                                 <failoverdomainnode name="isc0" priority="1"/>
>                                 <failoverdomainnode name="isc1" priority="1"/>
>                                 <failoverdomainnode name="isc2" priority="2"/>
>                         </failoverdomain>
>                 </failoverdomains>
>                 <resources>
>                         <ip address="172.16.127.15" monitor_link="1"/>
>                         <ip address="172.16.127.17" monitor_link="1"/>
>                 </resources>
>                 <service autostart="1" domain="fotest" name="nfstest"
> recovery="restart">
>                         <fs device="/dev/ices-fs/test" force_fsck="0"
> force_unmount="1" fsid="13584" fstype="ext3" mountpoint="/export/test"
> name="testfs" options="" self_fence="0"/>
>                         <nfsexport name="test_export">
>                                 <nfsclient name="test_export"
> options="async,rw,fsid=20" path="/export/test"
> target="128.83.68.0/24"/>
>                         </nfsexport>
>                         <ip ref="172.16.127.15"/>
>                 </service>
>                 <service autostart="1" domain="fotest" name="nfsices"
> recovery="relocate">
>                         <fs device="/dev/ices-fs/ices" force_fsck="0"
> force_unmount="1" fsid="44096" fstype="ext3"
> mountpoint="/export/cices" name="nfsfs" options="" self_fence="0"/>
>                         <nfsexport name="nfsexport">
>                                 <nfsclient name="nfsclient"
> options="async,fsid=25,rw" path="/export/cices"
> target="128.83.68.0/24"/>
>                         </nfsexport>
>                         <ip ref="172.16.127.17"/>
>                 </service>
>         </rm>
> </cluster>
> 
> Nov  2 13:05:48 isc1 fenced[2762]: fencing node "isc0"
> Nov  2 13:05:48 isc1 openais[2743]: [CLM  ] got nodejoin message 172.16.127.123
> Nov  2 13:05:48 isc1 openais[2743]: [CLM  ] got nodejoin message 172.16.127.124
> Nov  2 13:05:48 isc1 openais[2743]: [CPG  ] got joinlist message from node 2
> Nov  2 13:05:48 isc1 openais[2743]: [CPG  ] got joinlist message from node 3
> Nov  2 13:06:21 isc1 fenced[2762]: agent "fence_ipmilan" reports:
> Rebooting machine @ IPMI:172.16.158.159...ipmilan: Failed to con
> nect after 30 seconds Failed
> Nov  2 13:06:21 isc1 fenced[2762]: fence "isc0" failed
> Nov  2 13:06:26 isc1 fenced[2762]: fencing node "isc0"
> Nov  2 13:06:26 isc1 ccsd[2736]: process_get: Invalid connection
> descriptor received.
> Nov  2 13:06:26 isc1 ccsd[2736]: Error while processing get: Invalid
> request descriptor
> 
> --
> Linux-cluster mailing list
> Linux-cluster at redhat.com
> https://www.redhat.com/mailman/listinfo/linux-cluster