cpu and vcpu utilization in cloudstack

[TheKunalSen]

I have created an offering of 1cpu, 0.3ghz, and 1 GB RAM, and if I create a vm using this, the vm takes up a whole CPU core, as shown in the dashboard. Now, the thing is, I have set the cpu.overprovisioning.factor to 5 in the global settings

Now the thing is I have only 72 cores in total, which are there in my 3 hosts, each with 24 core 2.1ghz

So the dashboard is showing 1vm is using 1 physical CPU core, so that means I will only be able to create 72 vms

How can I create more vms cause if 1 cpu core is getting occupied per vm, then it's really compute-heavy.

[TheKunalSen]

Do I need to change the cpu.overprovisioning.factor before adding the hosts ?

Then I can see the vcpu count!

Cause this is getting really confusing at this point

[rajujith]

Sharing this here, I hope it answers https://docs.cloudstack.apache.org/en/latest/adminguide/hosts.html#over-provisioning-and-service-offering-limits

[TheKunalSen]

Thanks, it's is working, even after surpassing the max core count showed in the dashboard the vms are getting created until the GHz are available

[chunkyen]

I agree with you that it can be confusing, but based on your specs, 2.1x72x5 gives you exactly 756GHz. You can check the GHz used after your 0.3 GHz VM offering was used. I suspect that the overprovisioning is more to do with GHz vs number of cores.

[TheKunalSen]

So if the 72/72 cores get filled , still i would be able to add vms?

[SviridoffA]

@TheKunalSen Yes, you can run virtual machines; when it comes to overprovisioning, GHz is the key factor rather than the core count. For now, I usually run at 130% to 150% of used/total cores in dashboard.

[TheKunalSen]

One thing to note that I have changed the cpu overprovisioning factor of the cluster and the global settings after adding the hosts,

Do I have to remove the hosts and add again?

[SviridoffA]

@TheKunalSen Of course not. You don’t need to remove and re-add the hosts. This setting only affects resource orchestration and doesn't impact hypervisor operation; as a rule, it is applied only at the management server level. You can change overprovisioning parameters multiple times during cluster life cycle, even in very large-scale environments.

[TheKunalSen]

Okay thank you