RTC problems on Dell Poweredge Servers
Matt_Domsch at dell.com
Sun Jan 7 23:46:38 CST 2007
On Sun, Jan 07, 2007 at 06:57:28PM -0500, Francis Lessard wrote:
> Hi !
> (this is mostly my original post on Howtoforge)
> I really hope we come to a solution.
> I have installed a VMWare server on Debien Etch kernel 2.6.18-3-686 on my
> brand new Poweredge 860. I still have an issue related to RTC.
> On a standard server (non-dell from what I tested), issuing the command
> :> hwclock and then cat /proc/interrupts. The number would increase
> beside IO-APIC-edge rtc showing that hwclock receives correct interrupts
> form /dev/rtc...
> That's not the case on the Dell PowerEdge 860. And after looking through
> all my Dell Linux installations, they all exibit the same problem, all
> with different kernels ! (tested on 2 poweredge sc430)
> The symptom is corrected on Dell servers by adding --directisa option to a
> hwclock command or in /etc/init.d/hwclock.sh for booting slowdowns and
> shutdowns... But the this perticular issue is affecting directly VMware
> server. The vmware server need /dev/rtc for keeping up time correctly. So
> even though this solves the problem in hwclock, it does not fix my issue.
> This message is logged in syslog :
> Jan 3 16:08:41 pe860 kernel: rtc: lost some interrupts at 2048Hz.
doh! a 2048 Hz RTC is pretty often. VMware ESX recommends people
running 2.6 kernels to change their timer interrupt from 1000 Hz to
100 Hz exactly because a guest waking up the hypervisor so often is
terribly inefficient and you see big time skews this way. This is
covered in the VMware documentation pretty well. Adding a 1000 Hz
clock and a 2048 Hz RTC interrupt is bound to cause the system to
spend a lot of time accounting for time and not much time doing actual
Dell Linux Solutions linux.dell.com & www.dell.com/linux
Linux on Dell mailing lists @ http://lists.us.dell.com
More information about the Linux-PowerEdge