High availability without SAN
- Thread starter decibel83
- Start date
Hi,
yes if you use DRBD to create storage which are in sync for both nodes. See the wiki about drbd.
Udo
Thank you very much!
I have some questions:
I have some questions:
- I have a three node cluster, but I cannot sync all VMs to all three nodes, only two, so I wish to have all VMs synced to a node of one of another two...
- Should the datastore have the same name on both nodes, right?
- How I can configure HA on Proxmox to use only one node for backup node and not both others two?
You could setup one DRBD volume between nodes A and B and another DRBD volume between nodes B and C
Yes the DRBD storage should be named the same on both nodes that have access to it.
In the Proxmox Interface you can also mark that this storage is only accessible on the two specific nodes.
You need to use failoverdomains to control how HA works, this does not work in the GUI, you have to edit cluster.conf manually.
http://forum.proxmox.com/threads/8693-is-there-an-alternative-to-fencing-for-HA?p=50230#post50230
Because Proxmox HA is unaware of DRBD status (is it split-brain?) I feel it is safest to not use HA with DRBD.
I prefer to allow a human to make the decision on where to start a VM that resides on DRBD.
Hi,
I have been using proxmox 1.9 + drbd for a cluster (no HA) and it works very well. Now I need to deploy proxmox on a new cluster and proxmox 1.9 is not supported. However I also feel that I don't need HA: I prefer to migrate/start/stop the VMs manually if a node fails.
e100, could you please share briefly your experience with proxmox VE 2 and drbd without HA? My intent is to keep all VM running on a node A, and keep node B idle. In case node A fails, I would like to move all VMs to node B. With proxmox 1.9 it's relatively easy, one can force a node to become primary node and start the VMs on the new primary. As I understand, with proxmox 2 there is no longer the concept primary/secondary node?
Let's say all VMs are running on node A, and it crashes hard. How can I start all the VMs on node B?
thanks,
Tony
I have been using proxmox 1.9 + drbd for a cluster (no HA) and it works very well. Now I need to deploy proxmox on a new cluster and proxmox 1.9 is not supported. However I also feel that I don't need HA: I prefer to migrate/start/stop the VMs manually if a node fails.
e100, could you please share briefly your experience with proxmox VE 2 and drbd without HA? My intent is to keep all VM running on a node A, and keep node B idle. In case node A fails, I would like to move all VMs to node B. With proxmox 1.9 it's relatively easy, one can force a node to become primary node and start the VMs on the new primary. As I understand, with proxmox 2 there is no longer the concept primary/secondary node?
Let's say all VMs are running on node A, and it crashes hard. How can I start all the VMs on node B?
thanks,
Tony
First if you only have two nodes you will not be able to start a vm because you will not have quorum, unless you change the expected votes to 1 so you again have quorum.
Once you have quorum it is simply a matter of moving the VM config file to the working node folder and then starting it.
ie:
mv /etc/pve/nodes/nodeA/qemu-server/101.conf /etc/pve/nodes/nodeB/qemu-server/101.conf
This is not much different than on 1.9 since on 1.9 you still need a copy of the VM config files in the proper place.
Once you have quorum it is simply a matter of moving the VM config file to the working node folder and then starting it.
ie:
mv /etc/pve/nodes/nodeA/qemu-server/101.conf /etc/pve/nodes/nodeB/qemu-server/101.conf
This is not much different than on 1.9 since on 1.9 you still need a copy of the VM config files in the proper place.
Another workaround is #systemctl stop pve-cluster and then start it again but forcing local mode # pmxcfs -l
How about Proxmox VE 5? they have a step by step procedure? it ok have different capacity of Server (server01 - 4TB & server02 8TB)
Thank you
Thank you
