View previous topic :: View next topic |
Author |
Message |
smiffy Apprentice
Joined: 28 Jun 2006 Posts: 259 Location: SA.AU.AP.EARTH
|
Posted: Thu Aug 24, 2006 5:17 am Post subject: Blade 100 - very bad timekeeping |
|
|
My new Sun Blade 100 is a wonderful thing in many ways. The build quality is very good and I LOVE OpenBoot - after too many years of PCs, this is very nice.
I have taken the installed 20Gb disc out and have put in a pair of 80Gb Hitachi Deskstars, configured as RAID1. Only have 256Mb of RAM at the moment, but have just won another 512Mb on eBay.
Running kernel 2.4.16.17, everything seems to be working quite well so far. (I have the bootup file system check disabled due to some bug that everyone is complaining about.)
However, I have noticed one problem - timekeeping is really, really bad. I have ntpd running and it is having to make adjustments of 0.7 - 1.5 seconds every twenty minutes. Consequently, ntpq -c rv still puts it at Stratum 16.
Would this be something wrong in my kernel configuration or is this a hardware issue?
EDIT - Problem persists with all frequency scaling removed from kernel.
EDIT - Show relevant syslog entries.
Code: |
Aug 24 16:19:36 narsil ntpd[5533]: synchronized to 202.191.97.130, stratum 2
Aug 24 16:19:37 narsil ntpd[5533]: time reset +0.950675 s
Aug 24 16:19:37 narsil ntpd[5533]: kernel time sync disabled 0041
Aug 24 16:24:58 narsil ntpd[5533]: synchronized to 218.214.125.154, stratum 3
Aug 24 16:24:59 narsil ntpd[5533]: synchronized to 202.191.97.130, stratum 2
Aug 24 16:24:59 narsil ntpd[5533]: synchronized to 203.217.30.156, stratum 1
Aug 24 16:38:56 narsil ntpd[5533]: synchronized to 202.173.190.30, stratum 2
Aug 24 16:38:57 narsil ntpd[5533]: time reset +1.305468 s
Aug 24 16:44:20 narsil ntpd[5533]: synchronized to 202.191.97.130, stratum 2
Aug 24 16:44:22 narsil ntpd[5533]: synchronized to 203.217.30.156, stratum 1
Aug 24 16:52:58 narsil ntpd[5533]: time reset +0.529047 s
|
_________________ Matthew Smith |
|
Back to top |
|
|
simon_6162 n00b
Joined: 07 Jul 2005 Posts: 8
|
Posted: Wed Sep 06, 2006 3:36 pm Post subject: |
|
|
if it makes you feel any better I have exactly the same problem, on a blade 100. I have no idea why this happens either, i just leave ntp running and let it correct the problem for me. |
|
Back to top |
|
|
Weeve Retired Dev
Joined: 30 Oct 2002 Posts: 641
|
Posted: Wed Sep 06, 2006 10:57 pm Post subject: |
|
|
My experience has been that Sun machines tend to loose time quite rapidly. I don't recall if this is OS specific at all or not though as its been a while since I've used a non-Linux OS on one. |
|
Back to top |
|
|
smiffy Apprentice
Joined: 28 Jun 2006 Posts: 259 Location: SA.AU.AP.EARTH
|
Posted: Thu Sep 07, 2006 12:38 am Post subject: |
|
|
Was looking for a fix so that I could get rid of all the complaints that ntpd is making about the frequency instability!
Have read that problem exists with Solaris as well, although said to be a kernel bug.
Just wish I knew how the timekeeping worked - I could feed it a 1 PPS signal from GPS, if I a) thought it might help and b) knew how...
/me hates inaccurate clocks. _________________ Matthew Smith |
|
Back to top |
|
|
Fanfwe n00b
Joined: 28 Sep 2006 Posts: 6
|
Posted: Fri Sep 29, 2006 8:56 am Post subject: |
|
|
The SunBlade 100 clock runs fast by apx. 300ppm. (Sun inernal BugID 4460876, apparently affects all 'grover' systems) This causes it to gain 1.5 seconds per hour.
There is a patch for Solaris to fix this but I'm not sure on Linux there is such patch unfortunately... |
|
Back to top |
|
|
smiffy Apprentice
Joined: 28 Jun 2006 Posts: 259 Location: SA.AU.AP.EARTH
|
Posted: Fri Sep 29, 2006 9:59 am Post subject: |
|
|
This is how I understand it, too. I think that I may need to get in touch with the kernel developers to see if there is some fix that we can apply. _________________ Matthew Smith |
|
Back to top |
|
|
gust4voz Retired Dev
Joined: 09 Sep 2003 Posts: 373 Location: Buenos Aires, Argentina
|
Posted: Fri Sep 29, 2006 4:56 pm Post subject: |
|
|
If it's a precise known number it could probably be done in userland with a simple daemon. _________________ Gustavo Zacarias
Gentoo/SPARC monkey |
|
Back to top |
|
|
w.hill Tux's lil' helper
Joined: 23 Jan 2005 Posts: 133 Location: Perth, Western Australia
|
Posted: Sat Sep 30, 2006 12:40 am Post subject: |
|
|
My Sunblade 100 is the same. If I have it powered down for any prolonged period it boots up "way-out". The only solution that I've found is to use ntpd or chronyd.
I checked the RTC battery and ruled it out quite a while ago. |
|
Back to top |
|
|
smiffy Apprentice
Joined: 28 Jun 2006 Posts: 259 Location: SA.AU.AP.EARTH
|
Posted: Sat Sep 30, 2006 12:47 am Post subject: |
|
|
What I would like to know is if:
1) The clock in question (don't know if we're talking RTC or system clock issue here) can be calibrated by writing a value to a register somewhere.
2) If it's the RTC, whether there may be a hardware hack to fix it once and for all. (For instance if it has a faulty 32768Hz oscillator, cut the track and feed a signal from a decent TCXO, like the DS32KHZ.)
At the moment, I use ntpd, which HATES the clock - just look at ntp entries in /var/log/messages.
gust4voz - is using ntp or similar what you meant? _________________ Matthew Smith |
|
Back to top |
|
|
|