NTP has been around a *looong* time. Handling propagation delay was one of the design criteria. 28k dial-up will be fine.
|
Configuring ntpd on Slackware
Yes, I read that it (NTP) is one of the oldest protocols still in use, and apparently maintained by its 'inventor' still.
Anyhow, since my last post, I decided to go ahead and configure my ntp daemon. Luckily for me, others have taken this path before me, as the man pages are about the most vague I have seen to date, and most info on the net is based on them. A search of LQ brought me To This Thread which is where I got much of the stuff with which I filled in the blanks in my ntp.conf file. As of this moment, I think I am astep with the OP of that thread, if not ahead; when I finally got ntpd running correctly, it eventually told me that 11160 seconds was too big a jump, and to set the time manually. Well, 11160s is about 3 hours, which is precisely the amount that any one of my clocks is out-of-whack at any given time. So, I went to the BIOS and set the system clock to the CORRECT time UTC, and rebooted. I imagine the daemon is now running, but I have been posting this thread and haven't checked the log yet. However, I would like to post the file.conf and log output, so that someone who knows what they're talking about can maybe tell me two or three things: #1 - Is this output indicative of a properly functioning ntpd? and #2 - in the interest of not abusing the NTP server network, can anyone tell me what units of time are being used in the broadcastdelay setting in ntp.conf, and does this setting do what I think it does? The initial value in my file was 0.0008 but I have changed it to 10.0 . So, what does it mean, and what are the units? Further, how DO I tell it how often to check/sync the clock? Here's my file right now, with most comments removed: Code:
# Configuration file for ntpd. Hmmm.. Code:
29 Sep 01:42:18 ntpd[2369]: logging to file /var/log/ntp.log And I can see that it synchronized a second time, to a different server, and checking the log again before posting this, I see it is syncing at weird intervals. It's working, but is it working right? Well hey, my desktop clock is correct, despite that rotten 'Halifax' under it-- I'll locate that config file sometime, Raskin-- and I'll hafta wait till tomorrow to see when the crons run, but I suspect they'll be ontime. Meanwhile, feedback welcome! PS - 'scuse my typing, my pet snake's sleeping in/on my left hand :) |
Code:
29 Sep 02:38:07 ntpd[2369]: system event 'event_peer/strat_chg' (0x04) status 'leap_none, sync_ntp, 15 events, event_sync_chg' (0x6f3) more log output.. What's this 'ppm error' ? |
Well, from documentation follows that you do not need broadcast delay - you use unicast NTP to server anyway. It is needed when one server sends data to multiple clients, and for some reason doesn't want to exchange with them one-by-one. For frequency add 'minpoll <k> maxpoll <l>' after each server; ntpd will poll this server no more frequent than 2^k seconds, but not less frequent than 2^l seconds. It should sync at weird intervals - they are random to reduce chances of worst-case scenarios for servers. I would check status messages with Wireshark - maybe unreach means packet loss, for example, sniffer will show everything. And what desktop environment are you using? I know that KDE doesn't display TZ. Neither does IceWM.
PPM error is probably 'pulses per minute' - frequency error that can be calculated from multiple time updates. |
Wow, good stuff! Thanks Raskin :)
For the reocrd, I am using KDE, and it does display my timezone, but I shrunk the font so small that it now poses no problem.. I'd like to use a lesser WM, but to date I haven't gotten around to allocating the time it might take to get a lighter WM set up so I'd be happy with it. I'll have to look into Wireshark-- I have no experience with it, and don't even know what it is. The PPM info you suggest sounds good enough to be correct, so that works for me. I will add the minpoll and maxpoll, and hope it accepts them. Where did you get those arguments from? Thanks again! |
I went to www.ntp.org and read documentation there. That's probably the only place where you can be sure to get all options... Wireshark is just a program to show you all the network traffic on your computer. By the way, do you use plain or digital clock in KDE? I checked digital with no additional information checked in 'configure clock'->appearance.
|
I am using digital clock, with 'frame' and 'flashing dots'. I used analog clock until recently, and it was the same. I still have 'Halifax' under it, but no big deal.
I didn't realize, and didn't see, that ntp.org would have the ntpd configuration options. I'll read them. Since I'm here, I did this in my config file: Code:
# Canada pool servers: Code:
: offset 0.069469 sec freq 9.926 ppm error 0.033880 poll 6 :) |
Config options for NTPD
|
Well, minpoll is greater than maxpoll, it can be expected to fail.
PS. 'Ctrl-+' allows you not to think about font size PPS. Editing ~/.kde/share/config/clock*rc and setting Initial_TZ=0 (and RemoteZones= ) can help. |
minpoll & maxpoll
Quote:
Seemed logical, in the absence of specific details, but.. It was the other way 'round. All fixed up now :) and thanks again *again* for the config editing tips -- I'll check into that too! EDIT -changing the clock*rc didn't work. Soon as I save it, the next time I open it it's back the way it was. Anyway, as I mentioned, the font is so small that it doesn't bother me any more. :) Саша |
I told that the measurement unit is second, as it is stated on ntp.org..
And clock*rc should be edited without KDE running. |
Quote:
Thanks again ;) |
All times are GMT -5. The time now is 01:33 AM. |