high cpu on centos 5.4 linux guest which isn't doing anything

[aviwollman]

over the passed week i'm finding some of my centos guests (both x86 and x64) which normally use 15% cpu jumping to over 50%.

top shows a load of <0.30 and there are no processes which are working hard.

rebooting has brought the guests to there norm

the only clues i see are high timer interupts and fork rate (from smtp)

is it possible that our auto yum updating (which upgraded from 5.3 to 5.4) is the culprit ?

i saw articles about

adding nosmp noapic acpi=off apm=off or divider=10 to the kernel boot parameters of the Centos 5 guest.

or disabling acpid and cpuspeed

does anyone know of a offical recommendation of this ?

any ideas ?

thank you

Avi Wollman

[AWo]

Depending on the kernel you use, the timer you use and frequency compiled in a Linux guest generates up to 1000 timer ticks per second plus 1000 per CPU (in this case virtual CPU).

That is work the host has to do and you can see the load there. The guest thinks everything is fine.

How many vCPU's does your guest have. if you have more than one reduce it to one. It is always recommened to start with as less vCPU's as possible.

Have you installed the VMware Toos inside the guest, already?

AWo
VCP / VMware vEXPERT 2009

\[:o]===\[o:]

=Would you like to have this posting as a ringtone on your cell phone?=

=Send "Posting" to 911 for only $999999,99!=

Edited by AWo

[aviwollman]

two questions.

1- usually theses systems are fine but only in the past few days they jumped in cpu usage.

2- changing the amount of virtual processors ater the guest OS has been installed ? even lowering it ?

Avi

[AWo]

two questions.

1- usually theses systems are fine but only in the past few days they jumped in cpu usage.

If nothing has changed, nothing should change...

2- changing the amount of virtual processors ater the guest OS has been installed ? even lowering it ?

Depends on the OS. There are Winodws HAL's which can be used with one or more processors and there are HAL's which only can handle one CPU. By looking under the Computer in the Device Manager you can figure that out.

Linux might need a new Kernel. I don't know if that is the case with every Kernel and with which kind of configuration. But yes, lowering might need a new Kernel.

Check KB 1006427 and http://www.vmware.com/pdf/vmware_timekeeping.pdf for more information on the interrupt issue. There's a boot parameter "divider" to lower the rate. But that is a guess of mine that the problem comes from the interrupts generated.

Have you installed the VMware Tools inside the guest?

AWo
VCP / VMware vEXPERT 2009

\[:o]===\[o:]

=Would you like to have this posting as a ringtone on your cell phone?=

=Send "Posting" to 911 for only $999999,99!=

[AWo]

Yes, if you reduce the number of vCPU's to one you should use the uniprocessor kernel.

AWo
VCP / VMware vEXPERT 2009

\[:o]===\[o:]

=Would you like to have this posting as a ringtone on your cell phone?=

=Send "Posting" to 911 for only $999999,99!=