HA cluster

[otrotabi]

I am having a really hard time trying to understand the concept behind corosync and clustering. Maybe some charitable soul will feel sorry for me and explain it simple enough.

So, we have two identical servers running proxmox, and I want to have redundancy for some virtual machines, so that in case one of them fails, the other server will take over and start working. Each server has plenty of local disk space, so I mounted the vm in the local disk (actually a raid11 disk setup). I then added a third smaller proxmox machine for quorum to have HA.

Then I migrated the VM from the first server to the second server. I thought (and I now think I am completely wrong) that this migration process would duplicate the information on the first server to the second server, keeping them synced from then on. But actually that does not seem to be the case, once the vm is migrated, the vm disappears from the first server.

From further reading, I now "seem" to understand that a cluster needs shared storage for that to be true. Please confirm if I finally get the idea.

However, this makes me wonder: why would I need to use a cluster if there´s a single point of failure anyway there ? Except that the point of failure is not the server but the shared storage.

Thanks for your comments.

 

[bbgeek17]

From further reading, I now "seem" to understand that a cluster needs shared storage for that to be true. Please confirm if I finally get the idea.

You are correct. The other option is sync or async replication.

However, this makes me wonder: why would I need to use a cluster if there´s a single point of failure anyway there ? Except that the point of failure is not the server but the shared storage.

You have a 3 node HA compute cluster, that should connect to a set of redundant Network switches. If it does not - the network becomes a SPOF. The Storage layer should have its own HA. The types of storage HA can differ, some examples can be found here: https://www.blockbridge.com/architectures/

When building an HA environment - every layer has to have HA.

---

Blockbridge : Ultra low latency all-NVME shared storage for Proxmox - https://www.blockbridge.com/proxmox

 

[otrotabi]

Thanks for your prompt reply. I actually do not care about HA, I can wait to have either a switch or other component replaced. What I do care about is either a disk or a CPU going wrong, in that case I do risk losing information.

I read that Ceph and hardware raid do not like each other very much. What other simple software structure could I use other than Ceph to have my virtual machine replicated in my servers ?

 

[bbgeek17]

I read that Ceph and hardware raid do not like each other very much. What other simple software structure could I use other than Ceph to have my virtual machine replicated in my servers ?

"Simple" is a relative term. Perhaps this would work for you: https://pve.proxmox.com/wiki/Storage_Replication

---

Blockbridge : Ultra low latency all-NVME shared storage for Proxmox - https://www.blockbridge.com/proxmox

 

[esi_y]

So, we have two identical servers running proxmox, and I want to have redundancy for some virtual machines, so that in case one of them fails, the other server will take over and start working.

You are basically limited to replication, but do note that with 2 nodes only, the HA cannot work. You need a third vote or special configuration of the "corosync" engine. That third vote may come from another node or a quorum device - this is all to consider on top of having redundant storage (e.g. the mentioned replication).

https://pve.proxmox.com/wiki/Cluster_Manager

https://manpages.debian.org/unstable/corosync/votequorum.5.en.html#SPECIAL_FEATURES

 

[bbgeek17]

You need a third vote or special configuration of the "corosync" engine. That third vote may come from another node or a quorum device - this is all to consider on top of having redundant storage (e.g. the mentioned replication).

All good points, I think OP already realized it:

I then added a third smaller proxmox machine for quorum to have HA.

tacked away in the middle of everything

---

Blockbridge : Ultra low latency all-NVME shared storage for Proxmox - https://www.blockbridge.com/proxmox