Good Day,
After searching the forums and web in general, I think I know my answer, but would like to defer to the greater knowledge of the forum.
Apologies for the long post, but without all of the information, my question may not make sense.
I have convinced my workplace to give up its death grip on VMWare, and to go with PVE.
However, at this time, we have little to no budget for a new storage back end, so I need to reuse as much of the existing hardware as I can.
For the servers, that is no problem, I have similar class hardware at home running PVE like a dream.
But the shared storage is another matter.
Currently, we have an HP MSA 2040 SAN connected via FC (Fibre Channel) to the hosts via an FC Switch.
From what I am seeing on the Storage page of the wiki, and from numerous posts about similar situations, specifically bbgeek17's answer here, it seems my options for shared storage are limited as follows:
NFS, CIFS, GlusterFS, CephFS, and ZFS over iSCSI: All of these require a computer as the front end of the storage to offer those services. (The MSA SAN offers only FC or iSCSI)
IMO, this would only add an additional translation layer and unneeded latency.
Ceph/RBD looks very nice, but we would have to triple our 30TB array to have 3 nodes of identical storage, that's a big ouch on the pocketbook.
Therefore I see two possible pathways for me:
A) LVM over FC
B) iSCSI (either kernel or libiscsi)
Option B would require replacing the HBA cards, SFP modules, and FC Switch.
Which leaves Option A = LVM over FC.
Based on the thread mentioned above, this is doable with the following caveats:
- Thick Provisioning only as Thin Provisioning is not supported.
- No Snapshot support with Thick LVM.
Thick Provisioning I am completely ok with for our use case.
I will be setting up Proxmox Backup Server on separate bare metal, with its own separate ZFS storage array, and it will be backing up all of the VM's in the PVE Cluster.
Now comes the actual question:
Given that PBS will be backing up all of the VMs to its own storage, is the "no snapshots on the shared storage" caveat even an issue??
It feels like I might be missing a detail that I have not uncovered yet, but I am not really seeing a down side to this setup.
Thank you in advance for your time.
After searching the forums and web in general, I think I know my answer, but would like to defer to the greater knowledge of the forum.
Apologies for the long post, but without all of the information, my question may not make sense.
I have convinced my workplace to give up its death grip on VMWare, and to go with PVE.
However, at this time, we have little to no budget for a new storage back end, so I need to reuse as much of the existing hardware as I can.
For the servers, that is no problem, I have similar class hardware at home running PVE like a dream.
But the shared storage is another matter.
Currently, we have an HP MSA 2040 SAN connected via FC (Fibre Channel) to the hosts via an FC Switch.
From what I am seeing on the Storage page of the wiki, and from numerous posts about similar situations, specifically bbgeek17's answer here, it seems my options for shared storage are limited as follows:
NFS, CIFS, GlusterFS, CephFS, and ZFS over iSCSI: All of these require a computer as the front end of the storage to offer those services. (The MSA SAN offers only FC or iSCSI)
IMO, this would only add an additional translation layer and unneeded latency.
Ceph/RBD looks very nice, but we would have to triple our 30TB array to have 3 nodes of identical storage, that's a big ouch on the pocketbook.
Therefore I see two possible pathways for me:
A) LVM over FC
B) iSCSI (either kernel or libiscsi)
Option B would require replacing the HBA cards, SFP modules, and FC Switch.
Which leaves Option A = LVM over FC.
Based on the thread mentioned above, this is doable with the following caveats:
- Thick Provisioning only as Thin Provisioning is not supported.
- No Snapshot support with Thick LVM.
Thick Provisioning I am completely ok with for our use case.
I will be setting up Proxmox Backup Server on separate bare metal, with its own separate ZFS storage array, and it will be backing up all of the VM's in the PVE Cluster.
Now comes the actual question:
Given that PBS will be backing up all of the VMs to its own storage, is the "no snapshots on the shared storage" caveat even an issue??
It feels like I might be missing a detail that I have not uncovered yet, but I am not really seeing a down side to this setup.
Thank you in advance for your time.
