Clusters nodes fault tolerance

[DRVTiny]

I'am using Proxmox on 2-nodes cluster with /dev/pve/data on both nodes glued in one mirror (RAID1) with glusterfs (so "local" storage on both nodes are the same and always in sync). All works as expected, live migration is very fast, thanks for your great product, i am really happy to find it as a usefull and near-comprehensive replacement for VMware ESX!
But now i want to have some HA Cluster benefits with my very robust two-heads virtual environment. So when one of the "heads" is going down it will be nice if all of virtual machines was running on it will appeare on the head that still alive in a "shutdowned" state. So i am understand that on-line synchronisation of VM's memory is impossible, but i still need to be able to restart all VM's which crashed on one head (node), on another head.
I am experienced in shell and perl programming, so if this functionality is not realized soon, i can do it, but for this i need to understand the logic of how Proxmox works better...
For example, who can tell me, how Proxmox node knows what VMs running on the local node and what - on the remote one(s)?

Thanks!

Andrey Konovalov,
Russia

 

[dietmar]

but i still need to be able to restart all VM's which crashed on one head (node), on another head.
I am experienced in shell and perl programming, so if this functionality is not realized soon, i can do it, but for this i need to understand the logic of how Proxmox works better...

This is a very simple view of the problem - you also need reliable membership and fencing. But we are working on a HA solution.

 

[xrisse]

hi,

i wanted exactly the same an now i sync the config-files from one node to a backup dir on second node:

node1:
rsync -aP --delete -e ssh /etc/qemu-server/ root@node2:/etc/qemu-server-node1/.

node2:
rsync -aP --delete -e ssh /etc/qemu-server/ root@node1:/etc/qemu-server-node2/.

if node2 crashes:

node1:~# cp /etc/qemu-server-node2/*.conf /etc/qemu-server/

and now i can start the machines via webgui

[UPDATE]
inserted --detele to rsync command
[/UPDATE]

 

[DRVTiny]

Thanks for your solution, it is excellent and simple. "Everything of Genius Is Simple"

 

[ZyTer]

DRVTiny, you say

" I'am using Proxmox on 2-nodes cluster with /dev/pve/data on both nodes glued in one mirror (RAID1) with glusterfs (so "local" storage on both nodes are the same and always in sync). All works as expected, live migration is very fast, thanks for your great product, i am really happy to find it as a usefull and near-comprehensive replacement for VMware ESX!"

If it's possible for you, can you share your GlusterFs configuration files ? I'am going to make your configuration whith 2 nodes in master / master.