KVM: events/# process loads 100% of one guest OS core (Ubuntu)

Question

System administration

I rent a server from a foreign hosting provider SingleHop. Their own virtualization system is preinstalled on the server and there is no access to it. The data storage is located separately on the NAS.
Clients have access only to the virtual machines themselves, which can be created through their control panel within the available resources of the leased server. They call these servers dynamic servers and the main “feature” of their use is the ability to simply transfer a virtual machine to another, for example, more powerful hardware, and other clients will not eat up your resources, as is the case with VPS.

According to /proc/cpuinfo (processor model "QEMU Virtual CPU") they use KVM virtualization. This also explains the possibility of installing Windows OS in a virtual machine.

The point is the following. The VM is running Ubuntu 10.04.4. After the reboot of the virtual machine, the events / # system process hangs in the processes for about an hour and loads one core at 100%. This is not very annoying, because. quickly passes, but after adding them to the R1Soft Backup agent server by the system administrators (this agent adds its module to the kernel), this process began to hang constantly.

Contacting SingleHop support required buying their General Management package, but SingleHop's sysadmin couldn't resolve it and concluded that I had added some left driver to the Ubuntu kernel.

Server resources allow you to add another VM for the test. I did this to demonstrate to the support team that this is some kind of incompatibility issue between their KVM virtualization and Ubuntu 10.04.4. Indeed, after installing the 10.04.2 system image, the events/# process did not load the CPU. But the installation script for this image at the end runs "aptitude -y safe-upgrade", which successfully upgrades the OS to 10.04.4 and after that the CPU load from the ill-fated events/# appears.

Restarting the new VM does not help. By the way, the OS boot process lasts about 7 minutes. Unusually long.

I don't see anything unusual (errors, etc.) in the syslog, dmesg logs.

I have already sent them a message that they can test a fresh install of their Ubuntu image and finally fix the issue. But, to be honest, I am pessimistic about this.

Line from top:

PID USER      PR  NI  VIRT  RES  SHR S %CPU %MEM    TIME+  COMMAND
10  root      20   0     0    0    0 R   71  0.0   3:18.13 events/1

Maybe someone faced a similar problem? Where to dig if their system administrators cannot solve the problem? I don't want to pay them to fix bugs in their own software.

Answer 1

Run away from this host.