Linux clock glitch - drift jump on reboot - NTP
This is a discussion on Linux clock glitch - drift jump on reboot - NTP ; I'm running Fedora Core 6 on a 2.8 GHz P4/Xeon (2 hyperthreaded CPUs).
I'm running a 2.6.19 kernel - my own config file, no changes to the
official sources.
I just updated a whole bunch of software (yum update) and ...
-
Linux clock glitch - drift jump on reboot
I'm running Fedora Core 6 on a 2.8 GHz P4/Xeon (2 hyperthreaded CPUs).
I'm running a 2.6.19 kernel - my own config file, no changes to the
official sources.
I just updated a whole bunch of software (yum update) and rebooted
my system. (same kernel)
NTP's drift went from 151 ppm to 115 ppm.
Has anybody seen anything like that? Any hints on where I should
start looking?
I'm pretty sure I saw something like this a while ago, probably in
the other direction. but I didn't investigate. I think I have
enough log files to find it if that is likely to help. There may
have been a kernel switch.
--
These are my opinions, not necessarily my employer's. I hate spam.
-
Re: Linux clock glitch - drift jump on reboot
Hal Murray wrote:
> I'm running Fedora Core 6 on a 2.8 GHz P4/Xeon (2 hyperthreaded CPUs).
> I'm running a 2.6.19 kernel - my own config file, no changes to the
> official sources.
>
> I just updated a whole bunch of software (yum update) and rebooted
> my system. (same kernel)
>
> NTP's drift went from 151 ppm to 115 ppm.
>
> Has anybody seen anything like that? Any hints on where I should
> start looking?
>
> I'm pretty sure I saw something like this a while ago, probably in
> the other direction. but I didn't investigate. I think I have
> enough log files to find it if that is likely to help. There may
> have been a kernel switch.
>
This happened to me once. I think I was using acpi_pm as my clocksource
in the Linux kernel, then I compiled ACPI out of the Linux kernel and
rebooted. Oops.
Each Linux clocksource has its own frequency. On my PC, it's about 33.2
PPM for acpi_pm, 35.5 PPM for pit, and 95.3 for tsc. Whether tsc works
differently on a 2-CPU system, I don't know. If ntpd was stable on 151
PPM, and now it's stable on 115 PPM, maybe that's what happened.
Michael