Design question for LVM thick backend and HA for the VMs

[Rappy]

Hello group,


I have a question regarding the PVE design with fibre channel connected backend to Storage Array - LVM thick backend design from Proxmox perspective.
How you deal with the HA of the VMs ?

AI gives me feedback that it is risky to map a LUN to more than one node, could lead to data corruption if another node activates VG and write metadata to the LUN.
What are your recommendations from experience, if anyone tried it.

 

[bbgeek17]

Hi @Rappy , welcome to the forum.

You have to distinguish two scenarios:
a) you have two or more bare-bones vanilla Linux hosts where you map the same LUN, and use that LUN as if it was a local disk without any coordination
b) you have cluster software package that "owns" this LUN and arbitrates appropriate access between the hosts

The PVE is the latter. It has built-in technology and intelligence to allow simultaneous single LUN shared storage usage. There are some inherit limitations since the LUN is used as RAW storage rather than as File System.
The shared FC/iSCSI/SAS/NVMe backend is officially supported by PVE, which means it is tested and approved for home or commercial use. There is official and unofficial documentation that you can peruse. Many thousands of people use the PVE/Shared storage combination daily for their critical infrastructure.

Some documentation:
https://pve.proxmox.com/wiki/Storage
https://pve.proxmox.com/wiki/Multipath
https://kb.blockbridge.com/technote/proxmox-lvm-shared-storage/

Cheers

---

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