LVM on shared storage

kmentzelos

New Member
Feb 19, 2026
5
0
1
Hi,


I see that many people use shared storage (FC, iSCSI, NFS, etc.) with LVM (thin or thick) and no filesystem — just VM disks as logical volumes.

How is data integrity guaranteed in this setup?

I understand that only one node runs a VM at a time, but since the LUN is visible to all nodes:

  • What prevents LVM metadata corruption?
  • How is locking handled?
  • During live migration, when both nodes briefly access the same disk, how is consistency ensured?

Thanks!
 
Right, yes, that was me asking the same question earlier


I’ve been testing different setups (Ceph, ZFS replication, DRBD shared storage), and at the moment I feel confident setting up a primary/primary DRBD (plain DRBD, not LINSTOR) with LVM on top, and then managing the LVs from Proxmox — essentially one DRBD device / one VG per VM.

I’m also testing hook scripts to handle smooth primary/secondary DRBD transitions for extra safety.

Does that approach make sense from your perspective, or am I overlooking something important?
 
Does that approach make sense from your perspective, or am I overlooking something important?
That will work if you have a quorum machine. We used DRBD+LVM in active/active heavily with XEN before we switched to PVE over 10 years ago, but we used clvm with XEN and in the beginning with PVE, but it was replaced with PVE internal logic to ensure LVM metadata consistency. Don't play around with LVM by hand without clvm.
 
  • Like
Reactions: Johannes S
That will work if you have a quorum machine. We used DRBD+LVM in active/active heavily with XEN before we switched to PVE over 10 years ago, but we used clvm with XEN and in the beginning with PVE, but it was replaced with PVE internal logic to ensure LVM metadata consistency. Don't play around with LVM by hand without clvm.
alright, what setup did you end up using with Proxmox VE?
Are you using any shared storage, or just relying on LVM/local disks?

Thanks!