CPU usage over "overcommitment" value

Hello,

Some days ago, one of my three OpenNebula servers exceeded the value of “overcommitment” in host configuration. As you could see in these two screenshots, my host was at 106% of assigned CPUs, but host was at configured to share its 104 cores, but not more than them:

So my question is if it is a normal behavior of “overcommitment” parameter. My OpenNebula version is 6.8.0-1 and my servers are running Ubuntu Server 22.04.3.

Thanks.

Hi,

Are all VMs deployed to the host via the OpenNebula scheduler? Wondering because a VM instantiated or migrated to the host by an Administrator could override the limits (oneadmin for example has no limits…)

BR,
Anton Todorov

Hi @atodorov_storpool

Possibly, in that moment of 106%, some VMs were created using API as a “normal” users (not oneadmin) and not from Sunstone.

Could be that the problem? API? In that case (waiting for your confirmation), is there any way to disable that behavior? How could I force to avoid overcommitment?

Thanks.