local-zfs that no warning is set, when the storage gets full

[adminkc]

Hi all,

we face some strange issue with the local-zfs that no warning is set, when the storage gets full.
The command zpool list rpool shows the amount of free rpool storage space.
Now I created a VM (see attached picture) with a virtual disk which is larger than the remaining rpool storage space and tip in again zpool list rpool and the size is still the same.
I think that local-zfs only fills up the storage when something is written onto the physical disks?
Why I am asking this question is because we faced last week an I/O error and the error was that we add a very big HDD to one VM and Proxmox didn’t warn about a possible lack of storage space on the node. Later, a lot of data has been copied to this large virtual disk and as a consequence the node ran out of storage space (local-zfs was full) and all machines somehow “died”.
So my question is how can I see or check how much storage space I have free on local-zfs bevor creating a VM and to avoid that the local-zfs storage is filled up completely?

BR,
KC IT-Team

 

[fabian]

you can decide yourself whether you want ZFS zvols to be thin or not (configurable in the storage settings, but it will only affect newly allocated volumes).

 

[adminkc]

But how it can be that Proxmox does not have implemented something that says " can not create VM local-ZFS will be filled up"
Do you maybe have some websites where we can see how to set up zvols?

BR

 

[fabian]

But how it can be that Proxmox does not have implemented something that says " can not create VM local-ZFS will be filled up"
Do you maybe have some websites where we can see how to set up zvols?

BR

if you configure a storage that is sparse, you risk running out of space. that's just the way it is, and not specific to ZFS. see https://pve.proxmox.com/pve-docs/pve-admin-guide.html#storage_zfspool for how to enable "thick" volumes.

 

[adminkc]

configurable in the storage settings, but it will only affect newly allocated volumes

What do you mean with that?

 

[fabian]

exactly what I wrote - you can change the behaviour via storage.cfg, but it will only affect new volumes created on that storage, any existing volumes will remain as they are.