VM migration when server fail.

[recusant]

Hello everybody!

There is a cluster of two nodes (Proxmox 4.3, HP DL380 G9 Servers), HP MSA 2040 SAN connected via FC 8Gb. LVM shared for both nodes. Online and offline migration goes perfectly.

The problem is when we are testing HA and restart to the one node or make disconnection to the optical cable, VM machine must migrate to the next node, but result it's restart to the both servers.

I understand that this is the result of the fencing. Proxmox 4.x configured automatically and I can not use fence_tool and fence_node or the like.

Do I understand correctly that the hardware watchdog is using instead fence now? If so, is there any idea how to connect hardware watchdog with Proxmox? Can anyone share a link to guide for setting up watchdog with HP servers? In iLO I do not see anything like the watchdog.

On https://pve.proxmox.com/wiki/High_Availability_Cluster_4.x wiki says that the module to work with the watchdog from HP (hpwdt) is not working right and use it undesirable. I would like to know can I use the ipmi_watchdog module with HP servers? And where I can look HA logs?

Sorry for my English =)

Best regards.

 

[dietmar]

You need at least 3 nodes for HA.

 

[Alessandro 123]

The problem is when we are testing HA and restart to the one node or make disconnection to the optical cable, VM machine must migrate to the next node, but result it's restart to the both servers.

Side question: what happens when the storage connection is lost on a node? Is PVE smart enough to restart all affected VMs ?

 

[recusant]

Side question: what happens when the storage connection is lost on a node? Is PVE smart enough to restart all affected VMs ?

When the storage connection is lost on a node we have automatic restart both servers after 60 seconds.

 

[recusant]

You need at least 3 nodes for HA.

Somehow I missed this point. I thought that it is left only in the history. Thank You.
Can I use a different configuration third server for HA?

 

[Alessandro 123]

When the storage connection is lost on a node we have automatic restart both servers after 60 seconds.

I was talking about PVE, non on your specific infrastructure.