Fencing, HA with Dell M1000e. Stop Power and doesn't work.

N

nandex

Member
Proxmox Subscriber
Feb 6, 2013
47
0
6
41
Trujillo
Hi,

I'll try to put my cluster with HA. I have 4 nodes. I changed cluster.conf and validate. My cluster.conf: http://pastebin.com/BcymtFCx
If I launch fence_drac5: fence_drac5 -m server-1 -l root -p dc4CEYaB -a 192.168.3.44 -o list -x.... it works correctly.
But if I execute "fence_node cetamox-01 -vv", I obtain this:
fence cetamox-01 dev 0.0 agent fence_drac5 result: error from agent
agent args: nodename=cetamox-01 agent=fence_drac5 ipaddr= 192.168.9.173 login=root module_name=server-1 passwd=password secure=1
fence cetamox-01 failed

When cluster is running and I stop the RGManager Service in a node... it works correctly and VW are start on another node. But if I interrupt the power in a node with VM it doesn't works and VM stops.

I have 2 LVM groups like storage elements and are shared correctly.

Have you got any idea about my problem??.

Thanks.
 
Last edited:
nandex said:
..
But if I execute "fence_node cetamox-01 -vv", I obtain this:
fence cetamox-01 dev 0.0 agent fence_drac5 result: error from agent
agent args: nodename=cetamox-01 agent=fence_drac5 ipaddr= 192.168.9.173 login=root module_name=server-1 passwd=w6TCwxyU secure=1
fence cetamox-01 failed

...
Click to expand...

fencing must works, otherwise HA will not work. check your cluster.conf, as far as I see you configured the same IP address for all 4 fencing devices - this looks not ok for me. also the password is not set, ...
 
Hi Tom,

I put the different IP of Blades and pass.... it doesn't work.
Then I put de iDrac IP and DRAC pass.... it doesn't work.

The software is IDRAC7. I think that it doesn't matter.

Buf if I launch

fence_drac5 -m Server-1 -l root -p password -a 192.168.9.173 -o list -x.... it works correctly.

Have you got any idea??
 
Is it 'server-1' or 'Server-1'?

- - - Updated - - -

nandex said:
fence_drac5 -m Server-1 -l root -p password -a 192.168.9.173 -o list -x.... it works correctly.
Click to expand...

And reboot also works?

fence_drac5 -m Server-1 -l root -p password -a 192.168.9.173 -o reboot -x
 
I have the new configuration for Fencing (with ipmilan). And I have the VM with HA activated.
When I tryed to launch fence_node, NODE; the node goes to reboot and VMs starts in other NODE. But when I shutdown a node with power off directly AC, the VM's doesn't restart in other NODE.

is it normal?
 
ok, I think that the fencing is OK now. I need to put on cluster.conf, in argument for fence_ipmilan -> option=monitor and fence works perfectly. But now, I have other problem to do HA. When I shutdown by power off A/C one node, the other 3 nodes starts VMs from that NODE. But if I shutdown other NODE.... I only have a cluster with two nodes and appears a message that says: Quorum Dissolved and I cant starts this VM's.
Could I solved it??

I'll try with expected_votes="1" or/and two_node="1", but it doesn't work.
 
nandex said:
ok, I think that the fencing is OK now. I need to put on cluster.conf, in argument for fence_ipmilan -> option=monitor and fence works perfectly. But now, I have other problem to do HA. When I shutdown by power off A/C one node, the other 3 nodes starts VMs from that NODE.
Click to expand...

yes, thats the idea of our HA implementation. as soon as you mark a VM as HA, the rgmanager make sure that the VM is always on.

if you do not want that VMs are managed in that way, don´t mark them as HA.

nandex said:
But if I shutdown other NODE.... I only have a cluster with two nodes and appears a message that says: Quorum Dissolved and I cant starts this VM's.
Could I solved it??

I'll try with expected_votes="1" or/and two_node="1", but it doesn't work.
Click to expand...

If you have 3 nodes in total, the cluster is operational as long as two nodes are up. if you loose two nodes, then only one is left and therefor you lost quorum.
 
You must log in or register to reply here.
Top Bottom