3 node HA: stopping rgmanager causes stop/start VM and not migration.

[oernii]

Is it possible to configure the cluster to try live migration and only in the case of error stop the VM?

PS: pve 2.1, fence_apc tested, 3 nodes, iscsi+lvm shares storage

 

[dietmar]

I do not understand that suggestion. rgmanager only tries to migrate if the vm is already stopped.

 

[oernii]

Currently HA works like this: if node1 goes down gradefully, HA stops/saves VMs on node1 and starts them on another node. Is is possible to change the behavior to live migration instead, so that VM users do not experience downtime?

 

[dietmar]

No, you need to live migrate manually before you shut down a node.