Description:
- SLES 11 SP2 VM encounters CPU crunch issue constantly on ESXi 5.0 U1
- There is only one VM running on the host.
- There is no CPU limit or reservation set.
System Configuration:
ESXI version:5.0 U1
Server make and Model:DELL R610 with Intel processor. HT enabled with 2 NUMA nodes./Poweredge M610
VM Configuration:
- VMware Tools - Latest version.
- Virtual Hardware 8
- 4096 MB of RAM
- 2 vCPUs (1 virt socket, 2 cores per socket)/Also tested 4vCPUs (2 virt socket, 2 cores per socket)
- SCSI Controller : LSI Logic Parallel
- HD1 - 60 GB Thick Provisioned Lazy zeroed
- 1 Network Adapters - VMXNET3
GOS configuration:kernel version : 3.0
To recreate the issue:
- Run a normal SLES 11 SP2 VM under ESXi 5.0 U1.
Troubleshooting done:
- - This issue does not happens on SLES 11 SP1 VM.
- - Customer is able to rebuild the VM from scratch and reproduce it.
- - We disabled the THP support on VM and that reduced the memory utilization on the VM.
- - The average CPU utilization is nearly 100%.
- - The issue also happens on ESXi 4.1 U2.
- - There is no snapshot running, no storage latencies, no network latency.
- - The GOS CPU utilization looks fine.
Resolution:
- GOS image had the following kernel parameter for SP2: “idle=poll” in /boot/grub/menu.lst.
- This parameter basically tells the kernel to poll more aggressively. However it seems to be causing issues for VMs.
- After taking it out everything worked normally.