I don't think that's supposed to happen

Rockstor running on a hyper-v machine , boot drive is a VHD, other drives are passed direct through, 2GB ram allocated, Raid5 (that might have been a mistake) was sending it several large files.

The files were stored on the Host (From The Raid1 that has the rockstor VHD for the boot drive on actually, Had to use it for temp storage when I moved some files around to free up the other drives so I could pass them through to RockStor) So I can’t rule the VM’s OS drive being slow/unresponsive as a possible contributing factor.

can you try if the same thing happens with only one vcpu?

Maybe the problem is I’m not totally sure what causes it so I could change the vCPU config and never be sure if that fixed it or not.

I’ve only seen it happen twice, both times when the “Rockstor” appliance is having a load of data thrown at it over Samba (We’re talking 100’s of GB) for instance when it crashed that time it was having a 200GB virtual machine HDD image written to it.

Also the CPU cores are pretty slow (It’s a C2750 atom)

i have found some (older) post and post on the internet which suggest it might be related to the vcpu count as the hv_vmbus module had some bug in kernel 3.XX, ofc this has been fixed upstream at that time (august 2014) so it shouldnt be a problem as the kernel used by rockstor is newer, however to rule that out you can set it to 1 vcpu

i currently dont have hyperv running on one of my machines so i cant reproduce this right now

the c2750 should be more than enough in terms of cpu power, i myself have the smaller version (2550) which gives about 522MB/s aes encrypted raidz2 speeds (used in a freenas server), also a slow os drive shouldnt be the problem either, i used an old usb thumb drive in my testbench which gets roughly 10MB/s max, IOPS are worse, and the os bootup takes about 2 minutes, but havent seen any problems with that so far (pushed some 700GB to it)

if thats not the cause it might be some passed-through hba in combination with the new kernel

If I was going to build it again I’d probably go for a Linux Host and KVM, but I wanted to play with storage spaces at the time so picked windows.

Once I get the data on RockStor (And therefore Linux/BTRFS) I might be able to re-think using windows.

I like Windows as a desktop OS, i’ve always been more comfortable with Linux on servers for some reason, this one was the exception and I kinda regret it.

Just had it happen again when creating a new share, changing it to 1 vCPU now.

Edit: Nope still panic’d with a single VCPU…
Edit2: I think I’m in a position now to move away from using Hyper-V :wink:

This has happened once so far in Proxmox as well (Multiple Vcpus’s) I think there’s something that goes badly wrong when the VM/Host is under high load.

strange thing, for me also the virtio thing wasnt happening, might be something related to the drivers used on the host?

can you test if something happens if you do the same on baremetal?
should be doable by unplugging every drive, using a usb thumb drive for the rockstor os and the same harddrives used in the vm

Not easily, the machine isn’t local