Ceph RBD Namespaces support

[scott.ehas]

Hello,

Ceph Nautilus introduced proper namespace support for RBD pools. Does Proxmox latest support Namespaces for RBD pools? The Qemu patch looks to be accepted https://lists.gnu.org/archive/html/qemu-devel/2019-12/msg04389.html.

Thanks
Scott

 

[t.lamprecht]

Hi,

The patch you linked will only become available in QEMU 5.0, which is not yet released.

Namespace support is not yet integrated into Proxmox VE's ceph server management, as it's not fully useable yet.
If, it may become a opt-in setting, as not all require this restriction - it needs some thoughts about how namespaces should be assigned (e.g., per VMID or manually) and how that then affects VM clones and the like.

Just out of interest, for what specific use would you like to have RBD namespaces?

 

[scott.ehas]

Hi,

We use the Proxmox Hyper-Coverged design with a few customers who each have licenses, and also have large Ceph storage clusters within our datacenters. The RBD namespace support allows us to use an isolated client pool instead of many pools. It would make the management of the Ceph clusters easier to handle as it's just basically a metadata tag on RBD.

Thanks,
Scott

 

[victorhooi]

We use Ceph to provide VMs for several internal teams.

We'd love this feature (namespaces) to so we can implement quotas, and prevent any one single team from using up all the storage (thread)

I saw that QEMU 5.0 is now in pve-test =).

 

[t.lamprecht]

Can you please open a enhancement request over at https://bugzilla.proxmox.com/
The main task here will be to define where and how this feature is integrated.

From your POV, without having my developer bias on the product, where would you expect it to be?
(I cannot really promise that we then honor that expectations, but good input can help to decide )

 

[victorhooi]

Of course - I filed https://bugzilla.proxmox.com/show_bug.cgi?id=2698 with some initial thoughts.

Let me know what you think!

Would you need me to do a mockup?

 

[t.lamprecht]

For now this will do, much thanks! We're bit busy right now but let's see