I need some advices for a new setup with VSA and Proxmox.
The setup :
1 cluster with 2 Proxmox node with 1 volume for the OS and 1 volume for the storage of the VM
The storage is done with Datacore Sansymphony (HCI solution).
1 VSA + 2 worker + 1 VIA repository for backup + 1 Synology NAS NFS for backup copy
I read Veeam documentation and limitation and I have 2 mainly questions :
The Snapshot storage must be file level storage. In my scenario, that mean I need to use the volume available on the same disk as the OS right ? Because the storage used for the VM will be block storage …
For the worker, same issue if I understand it correctly, it must be file-level storage that supports snapshots. That mean I also need to use the same volume right ?
How do you usually manage it ?
Finally, for the File system of it, do you use ZFS or XFS or EXT4 ?
Thanks for your answer
Best answer by eblack
I just read something I don’t like a lot, SAVC is “preview feature”.
Do I have an other solution maybe ?
I’m thinking about Datacore plugin for storage snapshot...
I’d think shared NFS or Ceph RBD then probably if skipping LVM /w SAVC.
Shared NFS is not possible with Sansymphony but Ceph RBD could.
Just for me to know, Why not ZFS here ?
I just read again the previous picture I added and I can see ZFS and Ceph RBD both provide block storage and snapshot support.
ZFS over iSCSI could work but local ZFS pools are not shared on PVE. To make ZFS/iSCSI work you’d need an appliance that supports the SSH connections from Prox etc. I guess you could give each node its own ZFS pool and use Prox replication but then you are also changing the design around.
With the storage you have SAVC is still my first choice. NFS with a gateway device would be my second. For Ceph you’ll need 3 nodes.
As I’m looking a bit into Proxmox lately, I wanna see if anyone chimes in on this @damien commenge . I hadn’t gotten that far into the weeds so to speak of implmenting PVE. I’ve only done intial configs, etc. I’m curious about this since we are considering moving to PVE from VMW.
The alternate snapshot storage does not have to be on the OS volume. It only needs to be configured in Proxmox as file level storage, and Veeam uses it only when the VM’s original storage cannot snapshot.
EXT4 or XFS is fine underneath a Proxmox directory store. QCOW2 provides the VM snapshot capability. ZFS is block so it won’t satisfy the req.
Worker system disks can reside in the same dir storage, no problem there.
I’d think the Datacore would be the biggest hitch. Veeam lists VMs on custom storage as unsupported. I don’t know much about it other than Datacore calls it a custom storage class.
@eblack - so, NFS then? No iSCSI obviously? Hmmm….
Not necessarily. NFS would satisfy the file level requirement for the workers and fallback snapshot storage. iSCSI can still provide the VM storage if it is presented as native LVM with SAVC enabled on Proxmox 9.
The alternate snapshot storage does not have to be on the OS volume. It only needs to be configured in Proxmox as file level storage, and Veeam uses it only when the VM’s original storage cannot snapshot.
Hello,
Thanks for your answer.
As you can see on the Proxmox server, I don’t have any other choice than locally…
When I say on OS storage, I mean a directory I add on Proxmox web interface for this need…
How can you see Datacore could not be supported in this scenario ? I’m curious because there is a plugin storage for Datacore on previous versions but maybe this change on V13 ? I’m really curious about it :)
@eblack - so, NFS then? No iSCSI obviously? Hmmm….
Not necessarily. NFS would satisfy the file level requirement for the workers and fallback snapshot storage. iSCSI can still provide the VM storage if it is presented as native LVM with SAVC enabled on Proxmox 9.
With LVM thin on top of the iSCSI storage it should work right ?
My guess is yes, it could. I don’t like using LVM to add additional layers of complexity to my systems. I want to manage storage at the storage level (SAN) only, not logically...in addition to storage level.
@eblack - so, NFS then? No iSCSI obviously? Hmmm….
Not necessarily. NFS would satisfy the file level requirement for the workers and fallback snapshot storage. iSCSI can still provide the VM storage if it is presented as native LVM with SAVC enabled on Proxmox 9.
With LVM thin on top of the iSCSI storage it should work right ?
That’s another confusing topic. LVM thin can’t be shared across Prox nodes.. But with Datacore you could thin the datacore lun.
The Datacore plug-in does support v13/13.1 But it’s the universal storage API integration. Or in other words, not an extension of Veeam Prox storage support. So it may work but it’s still considered a custom PVE storage setup.
My guess is yes, it could. I don’t like using LVM to add additional layers of complexity to my systems. I want to manage storage at the storage level (SAN) only, not logically...in addition to storage level.
I’m not sure we talked about the same thing here.
I’m not talking about LVM to use storage logically (pvcreate,vgcreate,lvcreate...) but based on this array. Are they the same ?
@eblack - so, NFS then? No iSCSI obviously? Hmmm….
Not necessarily. NFS would satisfy the file level requirement for the workers and fallback snapshot storage. iSCSI can still provide the VM storage if it is presented as native LVM with SAVC enabled on Proxmox 9.
With LVM thin on top of the iSCSI storage it should work right ?
That’s another confusing topic. LVM thin can’t be shared across Prox nodes.. But with Datacore you could thin the datacore lun.
The Datacore plug-in does support v13/13.1 But it’s the universal storage API integration. Or in other words, not an extension of Veeam Prox storage support. So it may work but it’s still considered a custom PVE storage setup.
OK That means for LVM thin it’s not possible.
For LVM it could right ?
However, for Snapshot on LVM a second volume is needed ? It can’t be on the same one than the VMs ?
DataCore thin provisioned LUN > shared standard Proxmox LVM with SAVC enabled on 9.2+ > production VMs
Local Directory storage > Veeam workers and fallback snapshots
Separate Directory/NFS storage > Instant Recovery target
Thanks for your answer.
For the local directoy storage, I think I was right.
For the Shared storage, I was wront because LVM need to be used and not LVM thin (but it’s not possible in shared so ok ! ) and I’m going to look at how to implement SAVC ok !
For instant recovery I need an other directory or external storage ok !
I just read something I don’t like a lot, SAVC is “preview feature”.
Do I have an other solution maybe ?
I’m thinking about Datacore plugin for storage snapshot...
I’d think shared NFS or Ceph RBD then probably if skipping LVM /w SAVC.
Shared NFS is not possible with Sansymphony but Ceph RBD could.
Just for me to know, Why not ZFS here ?
I just read again the previous picture I added and I can see ZFS and Ceph RBD both provide block storage and snapshot support.
ZFS over iSCSI could work but local ZFS pools are not shared on PVE. To make ZFS/iSCSI work you’d need an appliance that supports the SSH connections from Prox etc. I guess you could give each node its own ZFS pool and use Prox replication but then you are also changing the design around.
With the storage you have SAVC is still my first choice. NFS with a gateway device would be my second. For Ceph you’ll need 3 nodes.
I just read something I don’t like a lot, SAVC is “preview feature”.
Do I have an other solution maybe ?
I’m thinking about Datacore plugin for storage snapshot...
I’d think shared NFS or Ceph RBD then probably if skipping LVM /w SAVC.
Shared NFS is not possible with Sansymphony but Ceph RBD could.
Just for me to know, Why not ZFS here ?
I just read again the previous picture I added and I can see ZFS and Ceph RBD both provide block storage and snapshot support.
ZFS over iSCSI could work but local ZFS pools are not shared on PVE. To make ZFS/iSCSI work you’d need an appliance that supports the SSH connections from Prox etc. I guess you could give each node its own ZFS pool and use Prox replication but then you are also changing the design around.
With the storage you have SAVC is still my first choice. NFS with a gateway device would be my second. For Ceph you’ll need 3 nodes.
Really thank you for your help about my issue :)
I can see it’s not so easy to use Veeam with Proxmox than it’s with vsphere / Hyper-v :)
I just read something I don’t like a lot, SAVC is “preview feature”.
Do I have an other solution maybe ?
I’m thinking about Datacore plugin for storage snapshot...
I’d think shared NFS or Ceph RBD then probably if skipping LVM /w SAVC.
Shared NFS is not possible with Sansymphony but Ceph RBD could.
Just for me to know, Why not ZFS here ?
I just read again the previous picture I added and I can see ZFS and Ceph RBD both provide block storage and snapshot support.
ZFS over iSCSI could work but local ZFS pools are not shared on PVE. To make ZFS/iSCSI work you’d need an appliance that supports the SSH connections from Prox etc. I guess you could give each node its own ZFS pool and use Prox replication but then you are also changing the design around.
With the storage you have SAVC is still my first choice. NFS with a gateway device would be my second. For Ceph you’ll need 3 nodes.
Really thank you for your help about my issue :)
I can see it’s not so easy to use Veeam with Proxmox than it’s with vsphere / Hyper-v :)
No problem! I think we are all still working out the gotchas with PVE this year. I’m just excited to see replication in Veeam added :)
Thanks for the explanation of SAVC. Trying to wrap my brain around how that works. 🤔😅
I’m looking more into PVE this yr as well because I think we “may” be going to it from VMW next yr. Not looking fwd to a full migration, but it is what it is. Thanks guys.
Thanks for the explanation of SAVC. Trying to wrap my brain around how that works. 🤔😅
I’m looking more into PVE this yr as well because I think we “may” be going to it from VMW next yr. Not looking fwd to a full migration, but it is what it is. Thanks guys.
I did a 12 part series on PVE 9.1. Might be worth your time. - 🤝
