[opensuse] Re: 10.2 system clock too fast and NTP
I have been following this thread and have seen several responders go off on tangents, proposing solutions that have little bearing on the problem as listed. The way I understood the original problem is that the OS clock routine in the installation in question is beserk. Nuts. Spastic. On speed. Or any other term you want to lay on it. It's like the wall clock with the hands spinning really fast in spite of having a motor that is supposed to be synchronous with the input ac frequency, and the OS clock correction software pukes when trying to correct it. The solution is to determine how the system clock actually works and then determine what is so special about his particular installation that would cause the problem. Obviously it only happens with a few systems or we would all be bitching about it. So what is so different about his? It does seem to be an interesting puzzle. Fred -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
Stevens wrote:
I have been following this thread and have seen several responders go off on tangents, proposing solutions that have little bearing on the problem as listed.
The way I understood the original problem is that the OS clock routine in the installation in question is beserk. Nuts. Spastic. On speed. Or any other term you want to lay on it. It's like the wall clock with the hands spinning really fast in spite of having a motor that is supposed to be synchronous with the input ac frequency, and the OS clock correction software pukes when trying to correct it.
The solution is to determine how the system clock actually works and then determine what is so special about his particular installation that would cause the problem. Obviously it only happens with a few systems or we would all be bitching about it. So what is so different about his? It does seem to be an interesting puzzle.
Fred
Finally someone "other than possibly Anders Norrbring with his "clock=pit" suggestion" with some sense. A dmesg output from the OP might shed some light. Mark -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
A dmesg output from the OP might shed some light. That won't help at all, since nothing related to the clocks is logged to
On 2006-12-17 04:48, Mark Hounschell wrote: the system log (the default xntp configuration is logging to /var/log/ntp). The only clues available are from reading the kernel time variables directly (adjtimex, or ntpdc if ntpd is running), and from comparing the hardware of systems that are having the problem. -- The best way to accelerate a computer running Windows is at 9.81 m/s² -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
Darryl Gregorash wrote:
A dmesg output from the OP might shed some light. That won't help at all, since nothing related to the clocks is logged to
On 2006-12-17 04:48, Mark Hounschell wrote: the system log (the default xntp configuration is logging to /var/log/ntp).
The only clues available are from reading the kernel time variables directly (adjtimex, or ntpdc if ntpd is running), and from comparing the hardware of systems that are having the problem.
dmesg | grep clock Time: acpi_pm clocksource has been installed. All the clock source stuff is in kernel not ntp Duh... -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
Mark Hounschell wrote:
Darryl Gregorash wrote:
A dmesg output from the OP might shed some light. That won't help at all, since nothing related to the clocks is logged to
On 2006-12-17 04:48, Mark Hounschell wrote: the system log (the default xntp configuration is logging to /var/log/ntp).
The only clues available are from reading the kernel time variables directly (adjtimex, or ntpdc if ntpd is running), and from comparing the hardware of systems that are having the problem.
dmesg | grep clock Time: acpi_pm clocksource has been installed.
All the clock source stuff is in kernel not ntp
Duh...
Remember the original problem IS NOT an NTP problem. The guys clock runs to fast. Mark -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
On Sunday 17 December 2006 03:42, Mark Hounschell wrote:
Mark Hounschell wrote:
Darryl Gregorash wrote:
On 2006-12-17 04:48, Mark Hounschell wrote:
A dmesg output from the OP might shed some light.
That won't help at all, since nothing related to the clocks is logged to the system log (the default xntp configuration is logging to /var/log/ntp).
The only clues available are from reading the kernel time variables directly (adjtimex, or ntpdc if ntpd is running), and from comparing the hardware of systems that are having the problem.
dmesg | grep clock Time: acpi_pm clocksource has been installed.
All the clock source stuff is in kernel not ntp
Duh...
Remember the original problem IS NOT an NTP problem. The guys clock runs to fast.
No it doesn't run too fast. Byte specifically stated his clock was just find on prior releases and in other OSs. Its only the latest version of Suse that is giving him fits. -- _____________________________________ John Andersen
On 2006-12-17 21:04, John Andersen wrote:
On Sunday 17 December 2006 03:42, Mark Hounschell wrote:
Mark Hounschell wrote:
Remember the original problem IS NOT an NTP problem. The guys clock runs to fast.
No it doesn't run too fast.
Well, yes, it does. The system clock is far more than just a hardware counter. It also consists of a kernel module that takes a raw count from some counter and translates that into real time.
Byte specifically stated his clock was just find on prior releases and in other OSs. Its only the latest version of Suse that is giving him fits.
Which suggests a bug in the kernel clock module. See also jdd's reply to my "what is clock=pit" query, specifically the URL mentioned. -- The best way to accelerate a computer running Windows is at 9.81 m/s² -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
On 2006-12-17 06:31, Mark Hounschell wrote:
Darryl Gregorash wrote:
On 2006-12-17 04:48, Mark Hounschell wrote:
A dmesg output from the OP might shed some light.
That won't help at all, since nothing related to the clocks is logged to the system log (the default xntp configuration is logging to /var/log/ntp).
The only clues available are from reading the kernel time variables directly (adjtimex, or ntpdc if ntpd is running), and from comparing the hardware of systems that are having the problem.
dmesg | grep clock Time: acpi_pm clocksource has been installed.
All the clock source stuff is in kernel not ntp Of course it is, but not much of any significance gets logged -- apparently just a line or two to say what source is being used. How is
dmesg|grep clock <null> All those dropped SYN packets since the last reboot may have a lot to do with that, but "grep clock /var/log/boot.msg" also shows nada -- whereas I do have this: <6>Using pmtmr for high-res timesource and nothing else related to the clock. that helpful, unless all you're interested in doing is changing which timesource is used? ntpd logs more meaningful stuff into the syslog than what I find in boot.msg. duh -- The best way to accelerate a computer running Windows is at 9.81 m/s² -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
Darryl Gregorash wrote: When everything is working right all you will see is a message indicating what clocksource is installed. It's when the kernel thinks there is a problem with one or more of the clock sources it will spit out meaningful info. Mark -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
Darryl Gregorash wrote:
Of course it is, but not much of any significance gets logged -- apparently just a line or two to say what source is being used. How is that helpful, unless all you're interested in doing is changing which timesource is used? ntpd logs more meaningful stuff into the syslog than what I find in boot.msg.
duh
No NTP messages are going to tell you why your clock source is not letting you keep reasonably accurate time. Don't you think a machine without NTP can keep time? NTP has nothing to do with this problem. It is only been suggested as a possible mask of the problem. Mark -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
No NTP messages are going to tell you why your clock source is not letting you keep reasonably accurate time. Don't you think a machine without NTP can keep time? NTP has nothing to do with this problem. It is only been suggested as a possible mask of the problem. Mark, I never *said* this was an ntp problem. Neither did I suggest that ntp messages will tell you much about the internal time; I merely said
On 2006-12-17 07:05, Mark Hounschell wrote: that ntpd logs more information. Of course I think a machine without ntp can keep time. I did, after all, suggest the OP stop his ntpd daemon before fetching the kernel time parameters and clock behaviour.
When everything is working right all you will see is a message indicating what clocksource is installed. It's when the kernel thinks there is a problem with one or more of the clock sources it will spit out meaningful info.
Though I don't know what checks the kernel makes, when the clock keeps a consistent rate, I doubt it will be logging much information. OP's system clock seems to be keeping a very consistent rate -- unfortunately, it's 178,900 ppm fast. It's nearly 2007, and we still don't have a decent high-speed clock <sigh> -- The best way to accelerate a computer running Windows is at 9.81 m/s² -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
participants (4)
-
Darryl Gregorash -
John Andersen -
Mark Hounschell -
Stevens