[pve-devel] ha cluster with dell servers and drac fencing config + some remarks
dietmar at proxmox.com
Sat Jan 14 17:54:06 CET 2012
> It would by great to fence manually a dead node with pve-manager, if something
> goes wrong with fencing device or for users without fencing devices.
> I can be done with command : "fence_ack_manual deadnodename"
> So maybe add a right click option on a dead node,to manually fence it.
Unfortunately, fence_ack_manual only works if executed on the fence master.
So it is not trivial to implement. I am also unsure if we want to put
that on the GUI - it's considered very dangerous (a simple button press can
result I a split brain!).
> If I stop manually a vm managed by ha. (for maintenance by exemple, or maybe
> if don't need to have it started all the day), the node is restarted by HA
> automatically. I could be great if we could really stop it without remove the vm
> from ha management.
Yes - I simply had no time to implement that.
> Not related to HA, but if a node goes down, we can't restart or move the vms
> manually on another node (with pve-manager).
Yes. Maybe we should implement a 'qm mirtate-from' command?
More information about the pve-devel