The best choice

[luklow]

Hi everybody
Can you help me with the best choice for disks synchronization over LAN.
My first choice was drbd9 . But i have few problems with this technology and I decide to change this to another technology.
I have 3 nodes with the same disk configuration on each node.
What do you think what is the best choice for disks synchronization over LAN like drbd.
I want to have:
3 disk with synchronization
HA over 3 nodes.
heartbeat.

 

[nicolaiB]

Do you need real HA or could you live with Hot-Standby? We run several machines with ZFS-replication and don't miss anything.

 

[samontetro]

My config is quite similar to yours: 3 nodes with same hardware on each node. I started installing proxmox (I have 2 hard drives with hardware raid1 for the OS) and setup HA as I was doing ten years ago (drbd, heartbeat, mon...). I had two 8TB additonal hard drive in each node, set up in hardware raid1 too, for drbd synchronisation and the VMs.

Then I went to a Proxmox training session...

Back to the office I break all this config and setup a proxmox cluster with the 3 nodes, uses the 8TB disks in JBOD with CEPH (nice video tutorial to set up CEPH in Proxmox) and for HA all is yet in proxmox. This solution is more reliable and easy to setup and manage. It works for more than 1.5 year now without any trouble. Migrating VMs, HA... is very efficient in proxmox.

I ran many tests before going in production, unpluging network cable, power cable or creating "kernel panic" on the nodes or in the linux VM to check the HA. I just got a small problem when the dedicated ceph network fails and have to implement a small daemon to shut down the node if the ceph interface is not able to see the two other nodes. VMs are automagicaly restarted by the proxmox HA on a new node . Migrating is not working as corosync and ceph falls in readonly mode on the failing node.

I'm using proxmox4.

Patrick