Having a weird issue with a new Proxmox Install. I’m using a ZFS pool for my VM’s, and snapshotting without including RAM works just fine, and runs near instantly. But if I check the box for include RAM, it gets hung and does nothing, no output whatsoever. It is actually creating space on the ZFS partition, but, it never finishes. I end up having to stop the task, unlock the VM via qm unlock (It is stuck locked), then delete the snapshot via commandline via the .conf file to get it to disappear from the GUI. Then I have to zfs destroy the zfs storage that gets made for that snapshot via zfs destroy. I have updated all the way, and have images below showing what’s happening. Has anyone dealt with this before?
Proxmox ZFS Snapshots w/ RAM not working
- Thread starter ryan.kistler
- Start date
Snapshots without RAM are instantaneous and do not require space. On the backend, this only needs a couple of milliseconds.
Snapshots with RAM are another monster. In addition to the storage backend, the contents of the RAM are written to disk, in your case 32 GB.
I would assume your storage is just very slow, but your screenshot implies NVMe, so that should not be the case.
If you snapshot with RAM, could you monitor your zpool with zpool iostat 5 and report back the numbers? It would be interesting to see how fast/slow it is.
Snapshots with RAM are another monster. In addition to the storage backend, the contents of the RAM are written to disk, in your case 32 GB.
I would assume your storage is just very slow, but your screenshot implies NVMe, so that should not be the case.
If you snapshot with RAM, could you monitor your zpool with zpool iostat 5 and report back the numbers? It would be interesting to see how fast/slow it is.
It's like it's not even trying to write to the SSD's. These are 4 gen4 SSD's in a ZFS Raid 10. They're pretty fast according to crystaldiskmark numbers i ran in a Windows VM hosted in this very Proxmox install.
I don't get it. I'll try restarting the server I guess and see if anything happens, but I'm not confident.
UPDATE: I ran it on a different VM with only 2GB of ram, and it still hung, but this time I could atleast see the io jump correctly this time with zpool iostat 1. But the vm still stays locked and we the snapshot never makes it to done. So it definitely seems like it is in fact writing all of the currently used RAM correctly. After the spike in IO shown here, there is absolutely 0 more io usage, as I turned off every other VM for troubleshooting. Note the large IO indicative of successful writing.
It appears the VM is stuck in a "finish-migrate" status. Which is odd, as I'm not doing any migrating whatsoever.
