High Availability Cluster issue

[timmy]

Hi,

I setup a HA cluster from reading: http://pve.proxmox.com/wiki/High_Availability_Cluster etc.

I have a 3 host setup running 2.1.

DRDB is on 2 of the servers mirroring a VMs disk.

I have set that VM up as a cluster managed machine, but it seems to try to bring the machine up on the 3rd machine which doesnt have the DRDB disk. Can you specify which machines a HA managed VM will come up on? It would make sense to me.

Thanks
Tim

 

[tom]

you need to define a failoverdomain, search the forum, e.g. http://forum.proxmox.com/threads/9423-Failover-Domains-question

 

[timmy]

Thanks for the prompt reply

 

[timmy]

Hi,

I ran a test and It didnt seem to auto fence a machine when I powered it down. Hence didnt start it on the 2nd machine.

If I manually fence the primary server then the 2nd one does take over correctly so its just a fencing issue.. any ideas?

config:

<?xml version="1.0"?>
<cluster config_version="25" name="xxx">
<cman expected_votes="2" keyfile="/var/lib/pve-cluster/corosync.authkey"/>
<fencedevices>
<fencedevice agent="fence_ipmilan" ipaddr="10.20.3.1" login="root" name="ipmi_host11" passwd="1234"/>
<fencedevice agent="fence_ipmilan" ipaddr="10.20.3.2" login="root" name="ipmi_host12" passwd="1234"/>
</fencedevices>
<clusternodes>
<clusternode name="host5" nodeid="5" votes="1"/>
<clusternode name="host11" nodeid="11" votes="1">
<fence>
<method name="1">
<device name="ipmi_host11"/>
</method>
</fence>
</clusternode>
<clusternode name="host12" nodeid="12" votes="1">
<fence>
<method name="1">
<device name="ipmi_host12"/>
</method>
</fence>
</clusternode>
</clusternodes>
<rm>


<pvevm autostart="1" vmid="201" domain="xxx"/>
<failoverdomains>
<failoverdomain name="xxx" restricted="1" ordered="1" nofailback="1">
<failoverdomainnode name="host11" priority="1"/>
<failoverdomainnode name="host12" priority="10"/>
</failoverdomain>
</failoverdomains>
</rm>
</cluster>

 

[dietmar]

I ran a test and It didnt seem to auto fence a machine when I powered it down.

Does fencing work if the node is down? Maybe the fence device has no power too, so fencing fails?

 

[timmy]

The fenc device works .. I can use it to power up and down the machine and check status from cli.

How long should it take to realise its offline?

 

[tom]

HA clustering and fencing is a challenging setup and you should think of going for a commercial support subscription - so we can work directly with you to get things sorted out.