January 15, 20251 yr Hey guys, For several months now, I've had an issue where my VMs would freeze after a small while. Over the months, I've narrowed it down to the following facts: - Only Windows VMs (I'm using Windows 10 Pro). My Ubuntu VM works perfectly. - Both my Windows VMs do the same. - Only when using QEMU (launching from Unraid VM tab). - Using Remote Desktop from my laptop never freezes. - Seems to happen usually when I start browsing the web. - If I just start the VM and let it sit in QEMU without doing anything, it doesn't freeze. - Also, kind of seems like if I just do non-internet tasks, it doesn't freeze. Since it started, I've updated Unraid a couple of times. Maybe once or twice. Things I've tried: - Isolating the CPU cores to just these VMs. - Changine machine from i440FX-7.2 to Q35-7.2 (based on solutions proposed here). - Changing network model from virtio-net to virtio. - Creating fresh VM. - Different web browsers within the VM as it happens most when browsing web. Not sure what else to try and it worked for like a year and then all of a sudden they started freezing. Maybe after an update? Not sure. Hoping someone here has heard of this issue! Thank you for any advice! Please find attached both my diagnostics log and a screenshot of the log from when my VM just froze a minute ago: devante-nas-diagnostics-20250115-0818.zip
January 16, 20251 yr Author Adding VM XML here: https://pastebin.com/FgmPZ879 Edited January 16, 20251 yr by DevanteWeary
January 17, 20251 yr Hi Chester! (reference to your avatar from sifl & olly show) After an upgrade to 7.0, encountered this with one of my VM. Started with a reboot of the VM not loading anymore - would just drop to EFI with "map: no mapping found" and not boot into windows bootloader. Messing about and changing Q35 version to q35-9.1 (edit: From an earlier Q35 version) restored the VM's boot capability, but performance was AWFUL, and freezing up entirely shortly after boot. I implemented the changes in this thread: Enabling hyper-v enlightenments as described in that article restored performance to what it use to be. Maybe give that a try with the <hyperv mode='custom'> block options, and check that your 'migratable' setting is off, and 'hypervclock' is enabled. I'm still having the same issue I've always had, where the redhat virtio ethernet driver appears to completely consume one CPU thread with kernel load any time the VM was receiving more than around 700mbit. (In windows task manager, you can observe this on the performance tab by right clicking you CPU graph, change graph to 'logical processors' and 'show kernel times' - I see 1 thread at 100% and it's the darker color, indicating that portion of CPU utilization is kernel mode - indicating that driver code is the culprit) I couldn't saturate my 10G connection at all. Upload seems to not be as affected, strangely. I'm using virtio, NOT virtio-net. Edited January 17, 20251 yr by bitcore
January 29, 20251 yr Author Hey @bitcore, thanks for the reply (just saw it). Thank you for the info and I tried the setting changes you suggested and loaded the VM, launched Discord and Brave on it, went to Unraid Dashboard and clicked Check all updates on the Docker tab. About a minute later, it froze again. The log shows nothing different than the one above and I attached the diagnostic file. So weird that this happens only with Windows VMs and only when accessing them through the Unraid dashboard (QEMU). devante-nas-diagnostics-20250129-1256.zip Edited January 29, 20251 yr by DevanteWeary
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.