Hello,
maybe someone can set me straight on a few assumptions I’m making here, based on what I’ve managed to understand and test. I’m coming at this from an enterprise standpoint, not a home lab: data integrity and resilience are non-negotiable. Customers expect their data to be recoverable, their VMs restorable without jumping through hoops.
First off, Proxmox Backup Server flat-out doesn’t support SMB/CIFS. You can mount a CIFS share and slap a Datastore on it, and PBS won’t complain—but you’ll immediately notice glacial performance. Worse, data integrity apparently isn’t even guaranteed. On a system that relies on non-disableable deduplication, that’s a recipe for disaster.
NFS? It’s obsolete. NFSv4 with Kerberos is absurdly complex. Let’s be clear: NFS without Kerberos in the 2020s is a security nightmare, so you’re forced into Kerberos. After wrestling through endless complexity just to mount an NFSv4 export from a Synology NAS on PBS, you hit another brick wall: PBS insists the share be owned by the ‘backup’ user (UID 34). Synology only assigns UIDs ≥ 1024 to its users, and there’s no way to override PBS’s hardcoded UID34 requirement. That’s a show-stopper.
Why on earth can’t PBS leverage the flexibility of standard network protocols like SMB or NFS without contortion? Any SMB-tier company will have a NAS they expect to use for hypervisor backups.
Sure, an iSCSI LUN works—but you can only expand it, not shrink it, which sucks away a lot of flexibility.
This is my second rant on the Proxmox forums, and it’s still a rant because while the project charges off in grandiose new directions, the fundamentals remain broken. Personally, I can’t, won’t, and don’t dare put it into production: compared to Hyper-V paired with budget backup tools (Veeam Community, Hornet VM Backup, Nakivo, just to name a few), I’m struggling. Even XCP-NG with XO handles these basics without a fuss.
Am I missing something?
maybe someone can set me straight on a few assumptions I’m making here, based on what I’ve managed to understand and test. I’m coming at this from an enterprise standpoint, not a home lab: data integrity and resilience are non-negotiable. Customers expect their data to be recoverable, their VMs restorable without jumping through hoops.
First off, Proxmox Backup Server flat-out doesn’t support SMB/CIFS. You can mount a CIFS share and slap a Datastore on it, and PBS won’t complain—but you’ll immediately notice glacial performance. Worse, data integrity apparently isn’t even guaranteed. On a system that relies on non-disableable deduplication, that’s a recipe for disaster.
NFS? It’s obsolete. NFSv4 with Kerberos is absurdly complex. Let’s be clear: NFS without Kerberos in the 2020s is a security nightmare, so you’re forced into Kerberos. After wrestling through endless complexity just to mount an NFSv4 export from a Synology NAS on PBS, you hit another brick wall: PBS insists the share be owned by the ‘backup’ user (UID 34). Synology only assigns UIDs ≥ 1024 to its users, and there’s no way to override PBS’s hardcoded UID34 requirement. That’s a show-stopper.
Why on earth can’t PBS leverage the flexibility of standard network protocols like SMB or NFS without contortion? Any SMB-tier company will have a NAS they expect to use for hypervisor backups.
Sure, an iSCSI LUN works—but you can only expand it, not shrink it, which sucks away a lot of flexibility.
This is my second rant on the Proxmox forums, and it’s still a rant because while the project charges off in grandiose new directions, the fundamentals remain broken. Personally, I can’t, won’t, and don’t dare put it into production: compared to Hyper-V paired with budget backup tools (Veeam Community, Hornet VM Backup, Nakivo, just to name a few), I’m struggling. Even XCP-NG with XO handles these basics without a fuss.
Am I missing something?
- The node shutdown doesn’t do hibernate --todisk 1—not by default, and even manually it’s buggy. Really?
- The backup solution has non-disableable deduplication. Seriously?
- The backup solution isn’t compatible with the de facto network protocol SMB. Really?
