G
G
gep20052015-12-05 18:40:32
Windows
gep2005, 2015-12-05 18:40:32

Why do Windows clients periodically freeze in KVM (Proxmox)?

Hello!
I have a server up with Proxmox 3.4. Four disks (SATA) in ZRAID (installation using Proxmox itself). Memory 32GB.
Two Xeon E5620s.
Virtual machines: a couple of containers with Debian, one with OwnCloud and several KVMs with Windows from 7 to 10.
10 rarely starts, but it works stably. Although it is possible that with a longer operation it may freeze - I did not check it.
But 7 and 2008R2 periodically hang. It may take several days, it may take several hours. They hang strangely: Everything seems to work, but when the program tries to access the disk, it freezes. The rest work, but as soon as they try to access the disk, they also hang.
In the case of a server, I usually connect to it via RDP. In the session, everything hung. I go to the console (via proxmox) - everything works there (the mouse moves). Poke Ctrl-Alt-Del - hangs. Everything is just a reboot.
I also noticed that at such moments the DiskIO schedule in Proxmox breaks. Those. here comes the load on the disk, and after freezing - 0. At the same time, the processor and memory graphics are in order.
It feels like something is happening to the virtual disk. But at the same time, Windows does not fall into a BSOD, but continues to work as best it can.
Disks in both cases are IDE / SATA (not virtio), cache - writeback, although at the beginning there was writethrough and it was the same.
By the way, I noticed that I have Windows 10 on virtio - maybe this is the case?

Answer the question

In order to leave comments, you need to log in

2 answer(s)
P
Puma Thailand, 2015-12-06
@opium

I have everything on virtio and does not hang. Does it make sense to bet not on virtio in your case?

D
dyasny, 2015-12-08
@dyasny

The IDE is inherently a bug. First you need to get rid of it. virtio_blk or virtio_scsi is already a more difficult question, except for testing on your own loads there is no way to decide which one to use.
Further, after the replacement, you need to look at the rest of the disk stack. firstly, virtio drivers, you can try different options, secondly, qemu and kvm_trace logs from the host, thirdly, drag disks from zfs to a simple lvm and see if there is a difference. If all else fails, try recreating the problem on the same hardware, but using centos 7 and libvirt rather than a mixture of bulldog and rhinoceros named proxmox.

Didn't find what you were looking for?

Ask your question

Ask a Question

731 491 924 answers to any question