Has anyone the same problem with NFW Qcow2 Snapshots not being external?Hi!
Regarding the QCOW2 external Snapshots:
I can't get them to work....
I updated our Test-Cluster to PVE9 Beta and added a NFS volume as storage.
The "Allow Snapshots as Volume-Chain" checkbox was visible and i activated that.
Its also visible in the storage.cfg:
nfs: lnx_test_01_rz0
export /vol_proxmox_lnx_test_01
path /mnt/pve/lnx_test_01_rz0
server 10..X.X.X
content images
prune-backups keep-all=1
snapshot-as-volume-chain 1
I cloned a VM to that volume and made a snapshot.
Sadly it was still the internal QCOW2 snapshot.
Then i tried it with a newly created VM, but also sadly the same... no external qcow2 snaps...
View attachment 88644
Am i missing something?
I eagerly awaited this feature, because snapshot creation and deletion on NFS shares is sometimes very painful with large disks and some disk pressure...
Proxmox VE 9.0 BETA released!
- Thread starter t.lamprecht
- Start date
Hi, I can reproduce the problem, seems like
snapshot-as-volume-chain 1 on an NFS storage is not respected for online snapshots (seems to work for offline snapshots though). I will look into it.EDIT: Sent a patch: https://lore.proxmox.com/pve-devel/20250731082538.31891-1-f.weber@proxmox.com/T/
Last edited:
Hi, it fixed, thank you.
one more thing, does `cannot rollback to middle snapshot` is expected result for thick-LVM?
TASK ERROR: can't rollback, 'ss11' is not most recent snapshot on 'vm-100-disk-0.qcow2'
yes, this is an intentional limitation for now. while technically we could extend the feature to support arbitrary snapshot trees, the code (and in particular the failure scenarios) would get a lot more complex, so we limited it to snapshot chains and rollback to the most recent snapshot for the time being.
Sounds interesting - but not necessarily for everyone - maybe there should be a way to toggle the view of those metrics on or off depending on need? What about the Server Load ? Being deprecated due to the PSI now ??
The patch is included in
libpve-storage-perl >= 9.0.12, could you please update and check whether volume-chain snapshots on NFS/CIFS work as expected now?
I think the "Server Load" graph should still be visible in the Node Summary on PVE 9 beta. Can you double-check (your screenshot shows the VM summary page, where there is no "Server Load" graph)?
My BAD - I was so distracted by the "Pressure Stall" charts - I didn't notice that I was looking at a VM - when I go to node level, I see server load ;-)
I still suggest the ability to toggle the more granular PSI charts on or off either individually or as a different "view"
Thx!
Please update (
apt update and apt dist-upgrade), the new graphs were only integrated today and pve-manager 9.0.0~12 does not include them yet.Imo the more detailed graphs are an excellent addition. For those who don’t need them, they can just not scroll down to look at them.
Last edited:
Linux 6.14.8-2-pve (2025-07-22T10:04Z)
Running currently consumer grade hardware (Ryzen PRO 4650G, MB AsRock B450M Pro4, 2x Micron MTA18ASF4G72AZ-3G2B1Z1 DDR4-3200), it supports ECC (ras-mc-ctl --error-count shown 0, driver enabled) and works normally when I boot into Win11, passed 3 hours memtest without errors.
It seems that with this kernel and enabled ECC it randomly freezes.
I recall I had 6.8 kernel before on Proxmox 8 and these were happening after over 2 weeks or less often, I could not find the reason, I thought it's GPU passthrough.
It does not leave a panic log, does not leave dmesg, nothing to track back, but after upgrading I had to disable ECC in BIOS settings to stop freezes on Proxmox 9 as it was happening pretty much in the first 10 minutes after boot, sometimes when I kept memory low, it could get to about an hour uptime. I disabled all IOMMU, GPU passthrough options, but in the end it was ECC.
While PVE9 is still in beta, could you please take a look? If there is a way I can enable some extra debug logs to trace what's wrong, I can provide anything if needed.
Running currently consumer grade hardware (Ryzen PRO 4650G, MB AsRock B450M Pro4, 2x Micron MTA18ASF4G72AZ-3G2B1Z1 DDR4-3200), it supports ECC (ras-mc-ctl --error-count shown 0, driver enabled) and works normally when I boot into Win11, passed 3 hours memtest without errors.
It seems that with this kernel and enabled ECC it randomly freezes.
I recall I had 6.8 kernel before on Proxmox 8 and these were happening after over 2 weeks or less often, I could not find the reason, I thought it's GPU passthrough.
It does not leave a panic log, does not leave dmesg, nothing to track back, but after upgrading I had to disable ECC in BIOS settings to stop freezes on Proxmox 9 as it was happening pretty much in the first 10 minutes after boot, sometimes when I kept memory low, it could get to about an hour uptime. I disabled all IOMMU, GPU passthrough options, but in the end it was ECC.
While PVE9 is still in beta, could you please take a look? If there is a way I can enable some extra debug logs to trace what's wrong, I can provide anything if needed.
I tried installing this, i was already running kernel 6.14 prior to upgrading and at first the upgrade said it finished without error but after a reboot found it basically gutted the system and quit.
so i ran the it again to try to force install of all the missing packages and it again said it finished but now the system just won’t boot at all and has a kernel panic screen and the only option is to turn the system off…..
I really would like to avoid a reinstall if possible in any way….
Is it at all possible to use the VE 9 iso debugging console to install the packages from the ISO? To repair/recover the install without a full reinstall?
This is my main homelab server and I don’t exactly have the spare drives / space to make full backups of everything to do the full reinstall properly and don’t want to start over.
Any help would be greatly appreciated.
so i ran the it again to try to force install of all the missing packages and it again said it finished but now the system just won’t boot at all and has a kernel panic screen and the only option is to turn the system off…..
I really would like to avoid a reinstall if possible in any way….
Is it at all possible to use the VE 9 iso debugging console to install the packages from the ISO? To repair/recover the install without a full reinstall?
This is my main homelab server and I don’t exactly have the spare drives / space to make full backups of everything to do the full reinstall properly and don’t want to start over.
Any help would be greatly appreciated.
I get the logs filled with this:
2025-08-02T09:15:16.507629+02:00 pm106 kernel: audit: type=1400 audit(1754118916.502:2747): apparmor="DENIED" operation="create" class="net" info="failed protocol match" error=-13 profile="/usr/sbin/ntpd" pid=865846 comm="ntpd" family="unix" sock_type="dgram" protocol=0 requested="create" denied="create" addr=none
It's a couple of lines every second. It were no problems in 8.4
Have the latest as of now installed.
2025-08-02T09:15:16.507629+02:00 pm106 kernel: audit: type=1400 audit(1754118916.502:2747): apparmor="DENIED" operation="create" class="net" info="failed protocol match" error=-13 profile="/usr/sbin/ntpd" pid=865846 comm="ntpd" family="unix" sock_type="dgram" protocol=0 requested="create" denied="create" addr=none
It's a couple of lines every second. It were no problems in 8.4
Have the latest as of now installed.