I am running SuSE 7.3 on a new G4/800 (single processor) and have noticed that the clock is losing about 3 minutes every 24 hours or so. I have never had any problems like this on any other system. I booted up OS X last night planning to let it run overnight to see if it was losing time as well (I normally run Linux 24 hours a day). Upon booting OS X, the time was immediately corrected! I don't think I have NTP enabled (still running the factory install of OS X) but I should double check to be sure. Assuming that I don't have NTP enabled, doesn't that mean that the hardware clock is keeping correct time and thus it is the Linux system time that is drifting? I've been researching the issue a bit and have run across the /etc/adjtime file. Mine just has one line of zeroes (0.0 0 0.0) so there is nothing strange going on there (unless it is supposed to be non-zero, of course). I noticed that the TZ environment variable is not set for my normal user account nor the root account but I did set the timezone during installation and it is set properly in the Control Center. Does anyone have any suggestions on how to fix this? Do I really need to run ntpd? I've never had to before. Do I just need to do some adjustments with hwclock? Thanks, Ron
Ron, I was just brushing up on the subject against the mini-howto, http://www.ibiblio.org/pub/Linux/docs/HOWTO/mini/other-formats/html_single/C lock.html Or for all http://www.linuxdoc.org/, and the jist of it was, at startup, linux copies the time in the hardware clock to the system clock in memory, which runs off a timed software routine to keep the clock on time. It never references the hwclock again, unless a startup occurs, or manual intervention takes place. Adjtime allows you to give the software clock a tweak factor if for some reason the software drumbeat is not timed right, interjecting a correction factor. Hope that helps, Dave Gomez On 2/14/02 7:38 PM, "Ron McCall" <ronald.mccall@snet.net> wrote:
I am running SuSE 7.3 on a new G4/800 (single processor) and have noticed that the clock is losing about 3 minutes every 24 hours or so. I have never had any problems like this on any other system.
I booted up OS X last night planning to let it run overnight to see if it was losing time as well (I normally run Linux 24 hours a day). Upon booting OS X, the time was immediately corrected! I don't think I have NTP enabled (still running the factory install of OS X) but I should double check to be sure. Assuming that I don't have NTP enabled, doesn't that mean that the hardware clock is keeping correct time and thus it is the Linux system time that is drifting?
I've been researching the issue a bit and have run across the /etc/adjtime file. Mine just has one line of zeroes (0.0 0 0.0) so there is nothing strange going on there (unless it is supposed to be non-zero, of course). I noticed that the TZ environment variable is not set for my normal user account nor the root account but I did set the timezone during installation and it is set properly in the Control Center.
Does anyone have any suggestions on how to fix this? Do I really need to run ntpd? I've never had to before. Do I just need to do some adjustments with hwclock?
Thanks, Ron
Dave, Thanks for those pointers. I realize now that I misunderstood the hwclock man page and thought the only way for the adjustments to take effect was manually (or via a cron job or equivalent). That didn't seem much better than NTP to me at the time so I went ahead and set up xntpd instead (fortunately, my DSL connection is up 24/7 so I can run this without any pain and suffering). NTP is working well so far but I still can't get over the feeling that something else is wrong and this is just a hack. I couldn't find any SuSE RPMS for adjtimex but it looks like every other PPC distribution has one. I guess I should look into that option more in case it is a little cleaner. Thanks again, Ron On Fri, Feb 15, 2002 at 09:24:24PM -0700, Dave Gomez wrote:
Ron, I was just brushing up on the subject against the mini-howto, http://www.ibiblio.org/pub/Linux/docs/HOWTO/mini/other-formats/ html_single/Clock.html Or for all http://www.linuxdoc.org/, and the jist of it was, at startup, linux copies the time in the hardware clock to the system clock in memory, which runs off a timed software routine to keep the clock on time. It never references the hwclock again, unless a startup occurs, or manual intervention takes place. Adjtime allows you to give the software clock a tweak factor if for some reason the software drumbeat is not timed right, interjecting a correction factor.
Hope that helps, Dave Gomez
Ron, although 3 minutes a day seems extreme, I don't think it's way out of the ordinary either. After reading the manual, they state that adjtimex is more elegant, but I also think it would take a long time to get it to be accurate though. What I mean is that you would have to do a long time study to understand the amount of drift you have, make an adjustment ( or an adjustment to the adjustment) and start again, except with an even longer time period. Ntp is more expedient, and at multiple times a day, with an open connection, no pain. dg On 2/16/02 8:23 PM, "Ron McCall" <ronald.mccall@snet.net> wrote:
Dave,
Thanks for those pointers. I realize now that I misunderstood the hwclock man page and thought the only way for the adjustments to take effect was manually (or via a cron job or equivalent). That didn't seem much better than NTP to me at the time so I went ahead and set up xntpd instead (fortunately, my DSL connection is up 24/7 so I can run this without any pain and suffering). NTP is working well so far but I still can't get over the feeling that something else is wrong and this is just a hack. I couldn't find any SuSE RPMS for adjtimex but it looks like every other PPC distribution has one. I guess I should look into that option more in case it is a little cleaner.
Thanks again, Ron
Dave, Have you heard about many other people that have had to use NTP to keep decent time? This is the first machine I've ever encountered that has had this problem. I think you're right, NTP is pretty painless, especially since my net connection is up all the time. It has made 38 corrections in just under 6 hours so far ranging from about 0.25 seconds to about 4.5 seconds depending on the execution interval which has varied from about 4 minutes to about 31 minutes. Thanks again, Ron On Sat, Feb 16, 2002 at 09:48:30PM -0700, Dave Gomez wrote:
Ron, although 3 minutes a day seems extreme, I don't think it's way out of the ordinary either. After reading the manual, they state that adjtimex is more elegant, but I also think it would take a long time to get it to be accurate though. What I mean is that you would have to do a long time study to understand the amount of drift you have, make an adjustment ( or an adjustment to the adjustment) and start again, except with an even longer time period. Ntp is more expedient, and at multiple times a day, with an open connection, no pain.
dg
Ron, yes, think the worse I saw was about 10min/week. I suppose you could do the rough adjust with adjtimex for the average, then ntp would only have to take care of the fine adjustment. As for no rpm, it's extremely easy just to pull it down, and build it. Most all programs are just 4 steps, "tar -xzvf tarball", "./configure", "make", and "make install". I've compiled what seems like hundreds of programs with nary a hitch. dave On 2/16/02 10:12 PM, "Ron McCall" <ronald.mccall@snet.net> wrote:
Dave,
Have you heard about many other people that have had to use NTP to keep decent time? This is the first machine I've ever encountered that has had this problem.
I think you're right, NTP is pretty painless, especially since my net connection is up all the time. It has made 38 corrections in just under 6 hours so far ranging from about 0.25 seconds to about 4.5 seconds depending on the execution interval which has varied from about 4 minutes to about 31 minutes.
Thanks again, Ron
On Sat, Feb 16, 2002 at 09:48:30PM -0700, Dave Gomez wrote:
Ron, although 3 minutes a day seems extreme, I don't think it's way out of the ordinary either. After reading the manual, they state that adjtimex is more elegant, but I also think it would take a long time to get it to be accurate though. What I mean is that you would have to do a long time study to understand the amount of drift you have, make an adjustment ( or an adjustment to the adjustment) and start again, except with an even longer time period. Ntp is more expedient, and at multiple times a day, with an open connection, no pain.
dg
Dave, That is some comforting news! I have been worried that there is something wrong with my new machine or that maybe I screwed something up during installation. My main concern with using a non-SuSE RPM (or building it from source) was that I am still coming up to speed on SuSE specific things like SuSEconfig (rc.config and rc.config.d). From what I have learned so far, I really like this scheme. It provides one stop shopping for most of your configuration needs. I am guessing that the adjtimex package probably just consists of programs and documentation and thus shouldn't require anything to be done with SuSEconfig. I am a bit concerned about the magnitude of the drift and thus the magnitude of the subsequent corrections affecting things like make in a really bad way. It doesn't seem terribly likely though and it is good that the clock is slow rather than fast so that there is no backward time jumping going on! I'll take a look at adjtimex today and see if I can drastically reduce the magnitude of the drift. Thanks again, I really appreciate the help! Ron On Sat, Feb 16, 2002 at 11:19:28PM -0700, Dave Gomez wrote:
Ron, yes, think the worse I saw was about 10min/week. I suppose you could do the rough adjust with adjtimex for the average, then ntp would only have to take care of the fine adjustment. As for no rpm, it's extremely easy just to pull it down, and build it. Most all programs are just 4 steps, "tar -xzvf tarball", "./configure", "make", and "make install". I've compiled what seems like hundreds of programs with nary a hitch.
dave
Hi, On Sat, Feb 16, 2002 at 10:23:48PM -0500, Ron McCall wrote: [...]
a hack. I couldn't find any SuSE RPMS for adjtimex but it looks like every other PPC distribution has one. I guess I should look into that option more in case it is a little cleaner.
Just to let you know, adjtimex will be included in further releases. Peter -- VFS: Busy inodes after unmount. Self-destruct in 5 seconds. Have a nice day...
On Sun, 17 Feb 2002, Peter Poeml wrote:
Just to let you know, adjtimex will be included in further releases.
What does the roadmap for SuSE on PPC look like? -j -- "Only whimps use tape backup: real men just save their important stuff unencrypted and let NSA mirror it."
Ron McCall wrote:
I am running SuSE 7.3 on a new G4/800 (single processor) and have noticed that the clock is losing about 3 minutes every 24 hours or so. I have never had any problems like this on any other system.
hey ron, you can get some minutes from my system... on the titanium g4 667 MHz my suse ppc 7.1 tells me, that we have already march. so i've got 1 month to much- if you take it all you could be happy for the next 39 years. (what the hell i should do with 43.200 minutes?) ;-) btw. both of my apple systems (9.2 and X) show the correct time. greetz from ben
Ben, Wow, I guess I don't feel quite so bad now! It is interesting that your clock is running way too fast while mine is running slow. It seems to be a software issue since as soon as I reboot, my time will be back to normal (meaning the hardware clock is keeping good time). I have since started using xntpd and it is keeping the time very well. I'd rather not have to rely on a net connection but my DSL connection is usually up all day every day for several months straight (seems to drop about once every 3-4 months). Thanks for the info! Ron On Mon, Feb 18, 2002 at 10:46:50AM +0100, suse-ben wrote:
hey ron, you can get some minutes from my system...
on the titanium g4 667 MHz my suse ppc 7.1 tells me, that we have already march. so i've got 1 month to much- if you take it all you could be happy for the next 39 years. (what the hell i should do with 43.200 minutes?) ;-) btw. both of my apple systems (9.2 and X) show the correct time.
greetz from ben
participants (5)
-
Dave Gomez -
Janus Sandsgaard -
Peter Poeml -
Ron McCall -
suse-ben