How to recover from Proxmox server physical hardware failure, also with a JBOD attached?

[tomgt]

Hello, I am building a Proxmox cluster with at least three nodes. I will install the PVE OS on a RAID1 mirror using Kingston DC500M enterprise SATA drives. That way if one of the boot drives fails, the server should keep running.

I will use ZFS storage with replication of my data/VM's between all the nodes every 15 minutes.
If one physical node goes down and I have to replace it with a completely new server, can I just swap the drives over to the new server and boot up again?
I would make sure the BIOS is set up the same as the original server.

I imagine there could be issues having that new physical server rejoining the cluster because of SSH certificate mismatch etc?

Also, I will have a JBOD attached to the server with a SAS cable and use ZFS to format the drives on it. Can I just plug that SAS cable into another Proxmox node and then will that node see the zpools and data?

Thank you for any advice

 

[Dunuin]

If you want HA and run 3 or more nodes I would use CEPH instead of ZFS. You basically get all the nice features like ZFS but CEPH is a shared/distributed storage so you won't loose any VM data if a node fails. With ZFS there is no shared storage. You just got multiple local storages that get synced using replication. But replication will never be perfectly in sync so if a node fails you will most likely loose some seconds or even minutes of data.

And you shoundn'T run ZFS ontop of any raid controller, even if it is configured as JBOD. ZFS needs direct access to the individual disks without any abstraction layer in between. You can read here why.

 

[tomgt]

Thanks @Dunuin Any suggestions on the best way to deploy CEPH for my 3 node cluster? I have not installed CEPH before..