2

I have a machine running Ubuntu 16LTS with KVM/QEMU for a hypervisor. It has 1 guest running right now (A Ubuntu Server) that is using less than 1% of it's CPU and about 100MB of the 1000MB RAM allocated to it.

The Host is a 8core AMD with 16GB RAM and at idle with no guest running has a load avg of 0.00. With that one guest vm the host load climbs to a load avg of 2.50 - 2.70 and the fans are spinning at max RPMs and stays like that. This is a new problem noticed within the past day.

What would cause the host to have such a high load when that 1 guest vm is running and using so little resources?

update I'm starting to wonder if it has something to do with other machines connected to the guest vm via samba share. I just restarted the host, the guest, and all physical machines that were connected via samba to the guest vm and all has gone back to normal. Before a restart of the VM host or guest would not resolve anything. After they were both running again the problem would immediately start. However now the VM Host load is back to 0.00 and everything else is working normal. I'm not sure what caused the problem.

jtlindsey
  • 323
  • 1
  • 6
  • 16
  • Do you have virtualization extensions turned on for this host? It sounds like you're maybe using a binary translation instead, which is typically super slow. Try `grep svm /proc/cpuinfo` to see if you have these extensions enabled via the BIOS / UEFI. – Spooler May 24 '17 at 17:04
  • Yes virtualization extensions are on. I'm starting to wonder if it has something to do with other machines connected to the guest vm via samba share. I just restarted the host, the guest, and all physical machines that were connected via samba to the guestvm and all has gone back to normal. VM Host load is back to 0.00 and everything else is working. I'm not sure what caused the problem. – jtlindsey May 24 '17 at 17:58
  • wannacrypt trying to encrypt your samba share? :) – dyasny May 24 '17 at 18:07
  • @dyasny I'm not sure. Sometimes windows machines are connected but this time all the connected machines were newer Linux desktops which to my knowledge are not directly vulnerable to wannacrypt. – jtlindsey May 24 '17 at 18:21
  • I think we need more information about this to help you. Specifically, could we get some better output regarding your CPU usage? For example, this could be almost entirely I/O wait if your storage stack has been poorly optimized (as is typical when implementing a virtual storage appliance). If you could give some text output from `top` and `iotop` while this is running, that would be good. In addition, getting some additional details regarding your storage system, such as RAID (if applicable), partition structure, virtual disk structure, guest controller emulation type, and amount of clients. – Spooler May 24 '17 at 18:31
  • @SmallLoanOf1M I'll post that information if it comes up again. I updated my post because all systems are back to normal. I do remember that htop on the host was showing more than 20 KVM QEMU duplicate processes for the guest vm while it was running and the problem was happening. However now, even htop has normal output. So if it is a poorly optimized file share, I'll have to figure out how to duplicate the problem. – jtlindsey May 24 '17 at 18:45

0 Answers0