I have a cluster with 3 ESXi 5.0 hosts with DRS (fully automated) enabled, and one of these hosts have a high cpu usage by the hypervisor (checked with vcops).
Do you know if ther is any way to detect why the hypervisor is using so many cpu resources.
how are the vms on the host?
one of them or a combonation of them are probably hitting the server up for cpu cycles
and keep in mind DRS is not meant to balance the load across all nodes in the cluster but to ensure the VMs are getting the necessary resources so you can have a singlre host that is heavily loaded but no VMs will be migrated if all are receiving their necessary resources -
The problem actually is that some ESXi process is using almost all of the cpu cycles, not the VM´s...
please get the report of esxi process which is consuming more CPU. you can find by ps command.
the process that is consuming the CPU resources is hostd.
Just when we restart the management agents the % of use down.
Regards,
Francisco Valenzuela
do you have any dead lun links that the host keeps tryign to reconnect to?
Have you checked your hostd.log file? Can you upload it?
/var/log/hostd.log
is the vmware webclient server also instaleld on this machine?
and is that activly used by many users?
just a thought, since I saw a lot of http requests.
this is a host ESXi part of a vCloud Director provider vDC. The VMware webclient is installed on the vcenter server machine, wich is hosted in another cluster (admin cluster).
Thanks,