[Linux-HA] Problem around stonith resources IP targets with fence_ipmilan
Andrew Beekhof
andrew at beekhof.net
Tue Oct 19 00:33:04 MDT 2010
It's because you're using PCMK_host_check=none
Only use that if every device can fence every host.
I don't have access at the moment, but iirc you need to tell each device which nodes it can kill with PCMK_host_list
Sent from my iPad
On 12 Oct 2010, at 17:20, "Alain.Moulle" <Alain.Moulle at bull.net> wrote:
> ( With releases : pacemaker-1.1.2-2.el6.x86_64 &
> corosync-1.2.1-2.el6.x86_64 )
>
> Hi,
>
> I have a problem that I don't understand :
> I have 4 nodes in a Pacemaker cluster,
> with 4 stonith resources likewise :
> on node1 : restofencenode2 with fence_ipmilan and <BMC-IP-node2>
> on node2 : restofencenode3 with fence_ipmilan and <BMC-IP-node3>
> on node3 : restofencenode4 with fence_ipmilan and <BMC-IP-node4>
> on node4 : restofencenode1 with fence_ipmilan and <BMC-IP-node1>
> and -INFINIY for each restofencenodeX on nodeX
>
> _OK let's say that node 1 is the DC :_
> if I execute "power off" on node 3, the node2 should call the
> restofencenode3 resource to fence it with
> the fence_ipmilan and <BMC-IP-node3> , but in fact in the syslog of node
> 2, I have :
>
> 1286884579 2010 Oct 12 13:56:19 node2 daemon debug stonith-ng [30671]:
> debug: stonith_command: Processing st_notify reply from node1
> 1286884579 2010 Oct 12 13:56:19 node2 daemon info stonith-ng [30671]:
> info: log_data_element: process_remote_stonith_exec: ExecResult
> <st-reply st_origin="stonith_construct_async_reply" t="stonith-ng"
> st_op="st_notify" st_remote_op="51fb5515-f9ec-4389-97d7-4a38dfa4f122"
> st_callid="0" st_callopt="0" st_rc="0" st_output="Rebooting machine @
> IPMI:_*<BMC-IP-node4>*_...Done
>
> So it is like if there is a mismatch of target IP addr when using the
> stonith resource ... and it effectively fences the node4 ! whereas
> it should have fenced the node3 with <BMC-IP-node3> configured in the
> restofencenode3 ...
>
> So I loss two nodes instead of only node3 where I have executed the
> poweroff.
>
> Even more strange : this "mistmatch" happens when I power off any node
> which is not DC , but
> if I power off the DC node, this one is fenced correctly by the
> restofencenode1 with fence_ipmilan and <BMC-IP-node1>
> from node4
>
> So, I admit this is very strange, and I have no idea how it can happen ...
> Just below is the stonith resources config ...
> Any idea ?
>
> Thanks
> Alain
>
> <primitive class="stonith" id="restofencenode1" type="fence_ipmilan">
> <instance_attributes id="restofencenode1-instance_attributes">
> <nvpair id="restofencenode1-instance_attributes-ipaddr"
> name="ipaddr" value="*<BMC-IP-node1>*"/>
> <nvpair id="restofencenode1-instance_attributes-login"
> name="login" value="<BMClogin>"/>
> <nvpair id="restofencenode1-instance_attributes-passwd"
> name="passwd" value="<BMCpass>"/>
> <nvpair
> id="restofencenode1-instance_attributes-pcmk_host_check"
> name="pcmk_host_check" value="none"/>
> <nvpair id="restofencenode1-instance_attributes-action"
> name="action" value="reboot"/>
> </instance_attributes>
> <meta_attributes id="restofencenode1-meta_attributes">
> <nvpair id="restofencenode1-meta_attributes-target-role"
> name="target-role" value="Started"/>
> </meta_attributes>
> </primitive>
> <primitive class="stonith" id="restofencenode2" type="fence_ipmilan">
> <instance_attributes id="restofencenode2-instance_attributes">
> <nvpair id="restofencenode2-instance_attributes-ipaddr"
> name="ipaddr" value="*<BMC-IP-node2>*"/>
> <nvpair id="restofencenode2-instance_attributes-login"
> name="login" value="<BMClogin>"/>
> <nvpair id="restofencenode2-instance_attributes-passwd"
> name="passwd" value="<BMCpass>"/>
> <nvpair
> id="restofencenode2-instance_attributes-pcmk_host_check"
> name="pcmk_host_check" value="none"/>
> <nvpair id="restofencenode2-instance_attributes-action"
> name="action" value="reboot"/>
> </instance_attributes>
> <meta_attributes id="restofencenode2-meta_attributes">
> <nvpair id="restofencenode2-meta_attributes-target-role"
> name="target-role" value="Started"/>
> </meta_attributes>
> </primitive>
> <primitive class="stonith" id="restofencenode3" type="fence_ipmilan">
> <instance_attributes id="restofencenode3-instance_attributes">
> <nvpair id="restofencenode3-instance_attributes-ipaddr"
> name="ipaddr" value="*<BMC-IP-node3>*"/>
> <nvpair id="restofencenode3-instance_attributes-login"
> name="login" value="<BMClogin>"/>
> <nvpair id="restofencenode3-instance_attributes-passwd"
> name="passwd" value="<BMCpass>"/>
> <nvpair
> id="restofencenode3-instance_attributes-pcmk_host_check"
> name="pcmk_host_check" value="none"/>
> <nvpair id="restofencenode3-instance_attributes-action"
> name="action" value="reboot"/>
> </instance_attributes>
> <meta_attributes id="restofencenode3-meta_attributes">
> <nvpair id="restofencenode3-meta_attributes-target-role"
> name="target-role" value="Started"/>
> </meta_attributes>
> </primitive>
> <primitive class="stonith" id="restofencenode4" type="fence_ipmilan">
> <instance_attributes id="restofencenode4-instance_attributes">
> <nvpair id="restofencenode4-instance_attributes-ipaddr"
> name="ipaddr" value="*<BMC-IP-node4>*"/>
> <nvpair id="restofencenode4-instance_attributes-login"
> name="login" value="<BMClogin>"/>
> <nvpair id="restofencenode4-instance_attributes-passwd"
> name="passwd" value="<BMCpass>"/>
> <nvpair
> id="restofencenode4-instance_attributes-pcmk_host_check"
> name="pcmk_host_check" value="none"/>
> <nvpair id="restofencenode4-instance_attributes-action"
> name="action" value="reboot"/>
> </instance_attributes>
> <meta_attributes id="restofencenode4-meta_attributes">
> <nvpair id="restofencenode4-meta_attributes-target-role"
> name="target-role" value="Started"/>
> </meta_attributes>
> </primitive>
> ....
> <constraints>
> <rsc_location id="loc1-restofencenode1" node="node2"
> rsc="restofencenode1" score="+INFINITY"/>
> <rsc_location id="forbiddenloc-restofencenode1" node="node1"
> rsc="restofencenode1" score="-INFINITY"/>
> <rsc_location id="loc1-restofencenode2" node="node3"
> rsc="restofencenode2" score="+INFINITY"/>
> <rsc_location id="forbiddenloc-restofencenode2" node="node2"
> rsc="restofencenode2" score="-INFINITY"/>
> <rsc_location id="loc1-restofencenode3" node="node4"
> rsc="restofencenode3" score="+INFINITY"/>
> <rsc_location id="forbiddenloc-restofencenode3" node="node3"
> rsc="restofencenode3" score="-INFINITY"/>
> <rsc_location id="loc1-restofencenode4" node="node1"
> rsc="restofencenode4" score="+INFINITY"/>
> <rsc_location id="forbiddenloc-restofencenode4" node="node4"
> rsc="restofencenode4" score="-INFINITY"/>
> </constraints>
>
>
>
> _______________________________________________
> Linux-HA mailing list
> Linux-HA at lists.linux-ha.org
> http://lists.linux-ha.org/mailman/listinfo/linux-ha
> See also: http://linux-ha.org/ReportingProblems
More information about the Linux-HA
mailing list