Weblog entry #14 for dkg
See all weblog entries made by dkg.
I have a machine which i'm trying to do this on, though, and while it's up and running, i'm getting serious clock skew: something like 10x real time. The machine is running stock debian etch, as up-to-date as you can get it, with a 2.6.18-3-686 kernel.
Interestingly, the hardware clock is stable, and very close to real time. I think i might just be getting way too many timer interrupts or something. Any suggestions about what to do? Somehow, i don't think ntp was designed to handle a skew of 10 seconds per second.
Here's a debugging attempt where you can see the skew:
0 orangutan:~# hwclock --hctosys; for foo in 1 2 3 4; do > date +%c > hwclock --show > ntpdate -q -u sundial.columbia.edu > sleep 10 > done Thu 01 Feb 2007 09:05:53 PM EST Thu 01 Feb 2007 09:05:54 PM EST -0.997837 seconds server 128.59.59.177, stratum 2, offset -2.570964, delay 0.40991 1 Feb 21:05:55 ntpdate[6909]: step time server 128.59.59.177 offset -2.570964 sec Thu 01 Feb 2007 09:06:05 PM EST Thu 01 Feb 2007 09:05:55 PM EST -0.256843 seconds server 128.59.59.177, stratum 2, offset -13.593056, delay 0.41589 1 Feb 21:06:07 ntpdate[6913]: step time server 128.59.59.177 offset -13.593056 sec Thu 01 Feb 2007 09:06:17 PM EST Thu 01 Feb 2007 09:05:56 PM EST -0.252818 seconds server 128.59.59.177, stratum 2, offset -24.668502, delay 0.41322 1 Feb 21:06:19 ntpdate[6917]: step time server 128.59.59.177 offset -24.668502 sec Thu 01 Feb 2007 09:06:29 PM EST Thu 01 Feb 2007 09:05:57 PM EST -0.256849 seconds server 128.59.59.177, stratum 2, offset -35.685214, delay 0.41321 1 Feb 21:06:31 ntpdate[6921]: step time server 128.59.59.177 offset -35.685214 sec 0 orangutan:~#Any suggestions? What should i be looking for?
Comments on this Entry
[ Send Message | View dkg's Scratchpad | View Weblogs ]
0.000000 1170381475 0.000000 1170381475 UTCI'm afraid i'm not sure how to interpret it.
[ Parent | Reply to this comment ]
I was getting a terrible clock skew problem while running Debian Sarge as a virtual machine with MS Virtual Server 2005 R2.
The time skew problem was solved by adding clock=pit to the bootloader.
Details of the clock parameter can be found in kernel-parameters.txt.gz from the linux-doc package.
clock= [BUGS=IA-32, HW] gettimeofday clocksource override.
[Deprecated]
Forces specified clocksource (if avaliable) to be used
when calculating gettimeofday(). If specified
clocksource is not avalible, it defaults to PIT.
Format: { pit | tsc | cyclone | pmtmr }
[ Parent | Reply to this comment ]
[ Send Message | View dkg's Scratchpad | View Weblogs ]
0 orangutan:~# hwclock --hctosys; for foo in 1 2 3 4; do > date +%c > hwclock --show > ntpdate -q -u sundial.columbia.edu > sleep 10 > done Fri 02 Feb 2007 10:11:37 AM EST Fri 02 Feb 2007 10:11:38 AM EST -0.997767 seconds server 128.59.59.177, stratum 2, offset -0.898535, delay 0.04990 2 Feb 10:11:38 ntpdate[2520]: step time server 128.59.59.177 offset -0.898535 sec Fri 02 Feb 2007 10:11:48 AM EST Fri 02 Feb 2007 10:11:49 AM EST -0.768946 seconds server 128.59.59.177, stratum 2, offset -0.898470, delay 0.04987 2 Feb 10:11:49 ntpdate[2524]: step time server 128.59.59.177 offset -0.898470 sec Fri 02 Feb 2007 10:11:59 AM EST Fri 02 Feb 2007 10:12:00 AM EST -0.776830 seconds server 128.59.59.177, stratum 2, offset -0.898535, delay 0.04991 2 Feb 10:12:00 ntpdate[2565]: step time server 128.59.59.177 offset -0.898535 sec Fri 02 Feb 2007 10:12:10 AM EST Fri 02 Feb 2007 10:12:11 AM EST -0.776747 seconds server 128.59.59.177, stratum 2, offset -0.898532, delay 0.04984 2 Feb 10:12:11 ntpdate[2569]: step time server 128.59.59.177 offset -0.898532 sec 0 orangutan:~#I had suspected a virtualized environment earlier (the hosting company is claiming that this is a real machine, but i've never seen it). Your comment makes me even more suspicious. Do you know how to detect if you are running within an MS Virtual Server? I've tried scoopy doo, which is billed as being able to detect a VMWare instance, but it doesn't detect anything.
Thanks again!
[ Parent | Reply to this comment ]
[ Send Message | View dkg's Scratchpad | View Weblogs ]
A year ago, Ben Armstrong of MS was claiming that then MS Virtual Server environment could be detected by a hardware scan, but it's not clear to me how to translate his VB into something i can run on debian, though i've tried a little bit:
0 orangutan:~# lshw -class system
orangutan
description: Desktop Computer
product: MS-7191
vendor: MSI
version: 1.0
serial: To Be Filled By O.E.M.
width: 32 bits
capabilities: smbios-2.3 dmi-2.3
configuration: chassis=desktop uuid=00020003-0004-0005-0006-000700080009
0 orangutan:~#
[ Parent | Reply to this comment ]
[ Send Message | View dkg's Scratchpad | View Weblogs ]
[ Parent | Reply to this comment ]
vmdebian:~# lshw -class system
vmdebian
description: Desktop Computer
product: Virtual Machine
vendor: Microsoft Corporation
version: VS2005R2
serial: 6241-9319-3528-4190-6131-5224-30
width: 32 bits
capabilities: smbios-2.3 dmi-2.3
configuration: boot=normal chassis=desktop uuid=4482F20C-B3F4-3F42-9665-4F3F5C30E6A2
Bad news is that still leaves you in the dark with the cause of your clock skew :(
[ Parent | Reply to this comment ]