Is it safe to run zone-wide NFS Primary Storage on the Management Server node itself? #12118
Description
Hi everyone,
We are running Apache CloudStack 4.22 with Ceph RBD as the main block storage for VM disks.
For simplicity, we have been hosting the zone-wide NFS Primary Storage share directly on the Management Server node itself (like /export/primary).
My question is very straightforward:
If the Management Server node goes down (reboot, crash, power loss, etc.) or simply the NFS service stops or becomes unreachable, what exactly happens to the whole zone?
From our previous painful experience (when we accidentally put NFS Primary on a compute node and took it offline), the entire cluster froze because every KVM host has that NFS share mounted with hard,intr options.
Will the same thing happen if the NFS share lives on the Management node?
Will all compute nodes hang? Will running VMs (with disks on Ceph) freeze? Will System VMs die?
We want to understand the exact failure mode before deciding whether to keep this setup or move the NFS share to a dedicated/HA server.
Thank you in advance for any clarification!