2-Node Cluster Failover not completely working

[jandoe]

Hi I have a big problem with my HA-Cluster. I do fencing on IPMI.
If i stop rgmanager service on one node, the second takes the vms. If i shut down one node this works too.
But if i take away the powercord the webinterface shows me that the host is down but no migration is done.
Also if i unplug the network cable (drbd and ipmi network cable still plugged). When network cable is plugged again the node gets the power off signal from fencing and in the same moment sends a power off to the other node so tha both are going down. What did i missconfigure? Hope anybody can help me..


cluster.conf:

<?xml version="1.0"?>
<cluster name="CL123" config_version="3">

<cman keyfile="/var/lib/pve-cluster/corosync.authkey" two_node="1" expected_votes="1"></cman>
<fencedevices>
<fencedevice agent="fence_ipmilan" name="ipmi1" lanplus="1" ipaddr="192.168.77.xx1" login="root" passwd="xxx" power_wait="5"/>
<fencedevice agent="fence_ipmilan" name="ipmi2" lanplus="1" ipaddr="192.168.77.xx2" login="root" passwd="xxx" power_wait="5"/>
</fencedevices>
<clusternodes>
<clusternode name="kearney" votes="1" nodeid="1">
<fence>
<method name="1">
<device name="ipmi1"/>
</method>
</fence>
</clusternode>
<clusternode name="Nelson" votes="1" nodeid="2">
<fence>
<method name="1">
<device name="ipmi2"/>
</method>
</fence>
</clusternode>
</clusternodes>
</cluster>
<?xml version="1.0"?> <cluster name="CL123" config_version="3"> <cman keyfile="/var/lib/pve-cluster/corosync.authkey" two_node="1" expected_votes="1"></cman> <fencedevices> <fencedevice agent="fence_ipmilan" name="ipmi1" lanplus="1" ipaddr="192.168.xx.x2" login="root" passwd="wirduell" power_wait="5"/> <fencedevice agent="fence_ipmilan" name="ipmi2" lanplus="1" ipaddr="192.168.xx.x1" login="root" passwd="wirduell" power_wait="5"/> </fencedevices> <clusternodes> <clusternode name="kearney" votes="1" nodeid="1"> <fence> <method name="1"> <device name="ipmi1"/> </method> </fence> </clusternode> <clusternode name="Nelson" votes="1" nodeid="2"> <fence> <method name="1"> <device name="ipmi2"/> </method> </fence> </clusternode> </clusternodes> </cluster>

 

[dietmar]

Well, I guess IPMI device need power, else it will not work.

 

[jandoe]

Hi. Thanks for the answer. Okay but what if my mainboard is defect? Computer goes down and the second node sends the power off to the first but the defect mainboard cannot receive. Now my vms from the defect node are arrested because they get not started on the running node. Can i migrate them manually?

 

[tom]

use external power fencing and go for a 3 node cluster. there are a lot good reasons why we always recommend this.

 

[dietmar]

Now my vms from the defect node are arrested because they get not started on the running node. Can i migrate them manually?

You can always override fencing manually with 'fence_ack_manual'.