Re: [ntp:questions] ntpd behavior with multiple pool definitions
> I tried breaking the hosts up into three equal sized blocks (say, pool1, > pool2, and pool3 for example). and found that if I listed all three pools in > the ntpd directive. i.e.: ... > To my surprise, I got an even distribution of nodes from each of the defined > pools (X is for anonymity): ... > - Is this intended behavior that anyone knows of or did I > just get lucky (i.e. simultaneous resolving of pools to get balanced > answers)? I don't think there is any attempt to balance the usage across multiple pools. It depends on the response from the DNS servers. If fact, I would expect it to work with 2 pools, one for inside a firewall that and another for outside if DNS for the internal pool only worked inside the firewall and the firewall blocked access to the external pool DNS or external NTP servers. > - Can I manually expand the count of active peers so that I get > more than 10? With the pool directives, this gives gives me up to 9 after > startup, but then it somehow drops down to 7 actual sources after a few > minutes. Why do you need so many servers? You can adjust the number of clocks the pool command(s) use with tos maxclock. The default is 10. The count includes slots setup with the "server" command and it looks like it is counting the "pool" slot(s) too. -- These are my opinions. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] NTP 4.2.8p10 released
Harlan said: > We're open to doing an even better job of telling folks about things like > this. I think a message should go out to (almost?) all lists when security fixes are available to the general public. The first mail on questions was David Taylor's announcement of the availability of Windows binaries. There was no mention of a security release. I just checked the archives for announce. Nothing since April 2015. The hackers list has only 3 messages in March. None was an announcement. If the mailing list traffic has moved to other lists and/or venues, then please make an official announcement and disable the old lists. (but please save the archives) -- These are my opinions. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] Monitoring Number of Clients
> Third, in my reading of the source code for the Network Time Protocol Daemon > (NTPD), the program is optimized to handle a high number of requests per > second, not to record who is making the requests. Look at the mrulist. You can see it from ntpq ntpq -c mrulist The default size is 600 slots. You will probably want to make it bigger. Details in miscopt.html if your setup installed it or html/miscopt.html in the source. With a bit of trial and error, you can figure out how big you need to make it to cover the time scale you have in mind and/or how frequently you have to grab a copy of the data to keep the table from overflowing. -- These are my opinions. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] How common is LI=3 - solved.
> By the way, the code I am writing is not part of a NTP algorithm to adjust a > system clock for time. It is for a one-time frequency calibration of an > oscillator. I take a time snapshot at the beginning and at the end of an > approximately six hour period during which I am counting cycles from the > oscillator in question. I hope to achieve a frequency accuracy of 5 PPM. > Once that measurement is made, I store it for subsequent use in my app. > Unless the hardware changes, there is no need to do the calibration again. Why bother with all the packets. Isn't your PC's clock good enough? 5 PPM over 6 hours is 0.108 seconds. Unless you have a crappy network connection, ntpd should keep your clock within a few 10s of ms. If we assume 25 ms measurement error at both start and stop, you only need 3 hours. -- These are my opinions. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
[ntp:questions] PPS on Sure GPS units.
> The Sure device works, but for me it was too difficult to make the PPS work. > I bought two and was unable to make either one output a detectable PPS. The > PPS was there, my scope told me so, but I could not make the computer detect > it. On the other hand, I am very clumsy, and have little experience with > electronics. Are you having troubles with the hardware or the software? Where did you have the scope on the PPS? If software, what OS/distro? -- These are my opinions. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] SHM driver as a PPS source
folk...@vanheusden.com said: > How can I tell NTPd that an SHM time source is actually a PPS and that it > thus should combine it with some other sources? (an '*' + 'o' combination so > to say) I don't think you can do it the way you have described. The API for SHM requires that you have already figured out which second matches your PPS pulse. If you have a real PPS pulse, you can connect it to a modem control pin or whatever your kernel supports and use the ATOM driver. Or maybe use gpsd. If your PPS signal comes in via some strange hardware, you can hack your kernel to support the normal PPS API or you can hack the ATOM driver to talk to your hardware. Or you can make your software that is talking to SHM smart enough to provide the right second. Look at the ATOM driver and see how it decides if a time stamp is good enough. You can probably just use the system time and set minsane to 3 so it will outvote your clock if the system time is way off. I'd have to do some testing to verify that worked right, but it might solve your problem. -- These are my opinions. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] Idea to improve ntpd accuracy
elliott...@comcast.net said: > Thank you very much for the paper reference on temperature compensation. The > paper and its results are excellent. A self-calibration feature (that > worked) so that each user could have it configured for his/her own computer > should not be too hard to implement. I wonder why NTPD never picked it up? You need a temperature sensor mounted right on the crystal. Most people aren't willing to go that far and/or ntpd works well enough without it. -- These are my opinions. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] Idea to improve ntpd accuracy
> I thought of that too but rejected it as way too machine-specific. Because > the clock PLL does not track frequency ramps feed-forward could be very > effective (e.g. order of magnitude). To make this work however, you need to > first discover which one of many potential temperature sensors on the MB > are most closely correlated with frequency, then measure a rough linear > gain from temperature to PPM. Seemed too hardware-instance-specific to me, > but perhaps there's a way... You want to measure the temperature of the crystal. Several/many years ago, Linux used to use the RTC/TOY clock for the main timekeeping. Now, the main CPU clock is used. You want to get a temperate probe on that. Mark Martinec has a wonderful web page on that. NTP temperature compensation https://www.ijs.si/time/temp-compensation/ -- These are my opinions. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] Best NMEA sentence
> Is there a generally accepted NMEA ?best sentence? for use with ntp? For > example, I?ve seen GPRMC, GPGGA and GPZDA all individually recommended in > Google searches, variously at 4800 or 9600 baud. Will whatever sentence I > choose ultimately give the same results? Do you have a specific brand/model in mind? (or chip set?) Simple answer is use GPRMC. Turn off the others if you can. GPGGA doesn't include the date. That might make a difference when booting after power loss if your clock battery is dead. GPZDA doesn't include a valid status marker. The other consideration is that the length of some sentences changes depending on how many satellites are visible and working. If the cluster of sentences starts at the same offset each second and the one you pick is after one that changes the length, the offset will vary by the length and baud rate. So pick one that is early in the cluster. -- These are my opinions. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
[ntp:questions] How to monitor NTP clients local time?
> But something is not working as I'm expected that. In the rawstats and > peerstats I can see just the time of the ntp servers from which I'm syncing, > but I don't see the time of clients who are syncing from me. > It's very strange, because according to the docs: rawstats : Each NTP packet > received appends one line to the rawstats file set rawstats only logs packets that are responses in the normal request/response pair. There isn't any simple way to do what you want. You can add a server line for each client you want to monitor. If you include "noselect", it will collect and log all the data and then not use it to pollute your local clock. -- These are my opinions. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] NTP Reach is Zero but tally code remain *
> * I was accepting that once ntp server (1) will have reach value=0, > the tally code (*) will be removed, and orphan mode will be available How long did you wait? It might take 5 or 15 minutes to give up on the old server. -- These are my opinions. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] Making ntpd tweak a TCXO
>> I'm assuming your TCXO is driving your A/D. > Yes. It spits out samples at a rate f_osc / 24 = 60 Hz. My ultimate > task is to timestamp these samples in such a way that there are compressions > and dilations in the time scale, but no discontinuities. 60 Hz is pretty slow. The PPS logic will work at that speed. Can you feed the A/D trigger signal into something that the kernel PPS supports? You don't actually need the A/D trigger signal, just something that is derived from your TCXO. You are building a PLL. They have potential stability problems. -- These are my opinions. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] Making ntpd tweak a TCXO
m...@hau.nz said: > The A/D is a sigma-delta type where you set the output sample rate and it > runs autonomously (locked to the input clock of course). "Sample ready" is > indicated by a falling edge on the SPI MISO line, and I am bitbanging the > serial interface from the PRU ... I'd expect some sort of data ready signal but maybe they don't need one if you are running it with a clock derived from your main CPU clock. > What I am unclear on is how to take that timer snapshot/timestamp, and > convert it into wall-clock time. The latter is only available from inside > the OS (Linux). I do not understand, practically speaking, how these single > instants in time are carried across the boundary from the realtime world > into the nondeterministic world of the OS. Presumably this is what the > Linux PPS code does? The PPS code just grabs a time stamp on an interrupt. There is an RFC on the API. It's running on the main CPU rather than a PPU. It's not cycle accurate, just as good as you can get using interrupts. You can get more accurate results if you have a counter/timer that can capture the counter on the edge that makes the interrupt. I don't know if the ARM PPS code does that. -- These are my opinions. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] Making ntpd tweak a TCXO
> I would like the data stream from my A/D to be long-term accurate. That is, > if the sampling rate is 60 Hz, I would like 60*86400*365 +/- 1 samples in a > [non leap] year. > To that end, I'd like to discipline the TCXO using ntpd. I would like it to > *not* change the virtual clock frequency in software as it usually does, via > adjtime or whatever hooks are provided by the OS. Instead, I want the > frequency corrections to go through my own driver to adjust the hardware > oscillator directly. > Does the NTP software provide any hooks that could be used to accomplish > this? I thought of a crude approach, simply writing a program to examine > the drift file periodically, and using that information to adjust the TCXO. > But that (a) fails to stop ntpd from adjusting the virtual clock frequency > within the OS, and (b) isn't really a proper control loop. I guess it would > work, sort of, as a last resort. I don't think servoing out the drift will do what you want. Suppose your TCXO slows down a bit. After a while, ntpd notices, tweaks the drift, then you notice, and adjust it to go the right speed. But during that time, the clock was running a tiny bit slow so you lost a fraction of a cycle. I'm assuming your TCXO is driving your A/D. I think what you want to do is monitor the times when your D/A gives you data. Suppose you set things up so that each buffer holds 1 second of data. When the buffer fills, grab the time. Compare that with the expected time. ... -- These are my opinions. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] custom NMEA messages
>>http://users.megapathdsl.net/~hmurray/ntp/GPS18LVCx-off.gif > You must have had one of the bad firmware releases: David Taylor's site > shows the improvement he got after upgrading to better firmware releases, in > which I believe you were involved with "Kiwi" Geoff on analysis. > http://www.satsignal.eu/ntp/Garmin-GSP18x-LVC-firmware-issue.htm If you look about halfway down that page, there is a table showing the timing variations of firmware versions 3.70 and 3.90. 3.90 has a peak-peak of 181 ms. That's roughly as bad as whatever firmware I'm running. So I haven't played the new-firmware game. Anybody know if I can update the firmware from Linux? (or *BSD) > If you are looking at NMEA message timing - that's all over the board on > every device, as those serial interface messages get to use only the time > left after SV tracking and GPS message processing is handled, which may not > be a lot on the microcontrollers used. Nav GPS receivers with PPS are good > enough with NTP for most of us who want consistent, correct system time but > do not need lab quality timing. No, some devices do it right. Look at my graph (way above). You can do a whole lot better than most devices do. The issue is not the delay. That can be fudged. The problem is stability. If the problem is CPU cycles to compute the answer, a solution is to wait until the worst case and then send the message. -- These are my opinions. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] custom NMEA messages
> Garmin is a real GPS. please don?t mock me... The timing on the GPS-18x is also horrible. The GPS-18 (no x) was good, but it wasn't very sensitive. http://users.megapathdsl.net/~hmurray/ntp/GPS18LVCx-off.gif -- These are my opinions. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] custom NMEA messages
[SiRF wander.] > Ugh! I would not have expected that much variation over USB - WiFi does > better! It's not USB. It's a firmware "feature". -- These are my opinions. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] set server to a non utc time by hand
> i would like to set up an ntp-server thats does not give back the correct > utc time, but for example utc-1 or another "wrong" time computed before. > Until now the only way I reached that result was to set the local clock to > the wished time and to sychronize then with 127.127.1.1. If you have a local refclock (GPS), you could "fudge" it to have the desired offset. --- This request comes up occasionally. Has anybody looked into adding the fudge offset to internet traffic rather than just refclocks? -- These are my opinions. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] Synchronize distributed PCs with GPS 1 PPS and NTPD for OWD measurements
sandip gangakhedkar said: > Let me clarify that the two nodes *do not* have access to the internet, but > only to each other, over an unreliable wireless link. So I did not consider > the option of choosing NTP sync during the measurements. What sort of distance accuracy are you expecting? What sort of distances will you be operating over? I think your initial message said you wanted time to under 1 millisecond. The speed of light is 1 foot per nanosecond. 1 ms is a million ns, so your error ballpark will be a million feet or 200 miles. Is that interesting? (Did I scramble something?) How unreliable is your link? > No, I want to measure the flight times of UDP packets which are sent from > one node to the other over the direct wireless link. I would ignore NTP and do everything yourself. Do something like ping. That takes 2 packets, but you don't need to know the time on the other end. If both ends need to know the distance, you can make a measurement with 3 packets. Or you can send a dozen packets and use the minimum time assuming the others had delays in the interrupt handler. (and use the spread in the times as an indication of quality) You will probably need to calibrate the response times of the CPUs and the delays through the radios so you can subtract it out. The radio delays may may depend on signal strength which varies with distance, but will also change if you go behind mountains or trees or buildings. Joachim Fabini said: > - Re-compile your kernel for LinuxPPS support, following the instructions on > http://linuxpps.org/wiki/index.php/LinuxPPS_installation . That hasn't been necessary for a long long time. What version of the kernel are you using? Recent kernels need something like: ldattach 18 /dev/ttyS0 which creates /dev/pps0 -- These are my opinions. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] Nice fanless high-perf NTP server: Fitlet!
Re. the Fitlet: With a 3.9 to 4.5 W power budget this box will never get into those ranges, but even handling 1K requests/second with sub-ms jitter and delay would still be a very nice Pool server. A Raspberry PI can do 1500 packets per second. That's a simple measurement with one request in flight at a time. There are lots of reasons that real traffic might go faster or slower. -- These are my opinions. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] NTP Autokey - who is actively using it?
Is there a reasonable HOWTO type document describing how to set things up? Several years ago, Dave announced that one of the machines at UDel was ready for testers. I got as far as discovering that Autokey doesn't work through NAT boxes. Since I'm behind a NAT box, I gave up. I should be able to help testing by setting up a couple of local machines to use Autokey. So the recipe I'm looking for has to cover both client and server in case there are any differences. -- These are my opinions. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] Leap second to be introduced in June
[Context is google-smear.] For distributed logging you have to use the same method for every single node, but that is the case today as well. :-( I.e. with one domain smearing and another stepping, the times between them will be skewed over the entire smearing period. How often do people working with log files from 2 systems care about fractions of a second? -- These are my opinions. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
[ntp:questions] Linux kernel bug in 3:12, low_latency on serial ports is broken
Just a heads-up in case you hit this, or want to avoid it. The symptom is BUG: scheduling while atomic https://bugzilla.redhat.com/show_bug.cgi?id=1065087 The discussion at https://lkml.org/lkml/2014/2/18/77 confirms that it's the low-latency path that is buggy. My setup works fine under low load, but hits the BUG when I do something like compile the ntp package. 3.11 works fine. (for me) There was major work in the serial area for 3.12 I'll send an update when/if it gets fixed. -- These are my opinions. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] kernel pps
In article jf4tou$7rf$1...@dont-email.me, unruh un...@invalid.ca writes: How do I enable kernel pps? I have kernel 2.6.33-7mdb (Mandriva 2010.1) which has a pps_core module. But even if I install that module I get no /dev/pps0 device not a /sys/class/pps/pps0 Somehow I have to tell it to use look at say the DCD line on the first serial port. How do I do that? You need to do something like: ldattach 18 /dev/ttyS0 google should find the details. We should add them to the wiki. -- These are my opinions, not necessarily my employer's. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] Kernel PPS with Debian
In article 95325d10-3ee0-482e-927f-53991e056...@v5g2000vbh.googlegroups.com, jimmyterrence jimmyterre...@gmail.com writes: What do I need to do to get ntp to notice the PPS signal? What am I missing? You need to do something like: ldattach 18 /dev/gps0 That will create /dev/pps0 Beware: If you have CONFIG_PPS_CLIENT_KTIMER turned on, that will gobble up /dev/pps0 and the first ldattach will create /dev/pps1 -- These are my opinions, not necessarily my employer's. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] Trimble Resolution T
In article slrnj9ph83.sve.un...@wormhole.physics.ubc.ca, unruh un...@physics.ubc.ca writes: On 2011-10-17, Hal Murray hal-use...@ip-64-139-1-69.sjc.megapath.net wrote: In article slrnj9oqdj.3i3.un...@wormhole.physics.ubc.ca, unruh un...@physics.ubc.ca writes: In fact it is really hard for a computer to keep to 1us accuracy because of the delays in interrupt processing. It's not the delay that is the problem. It's simple to correct for a fixed delay. (at least in theory) The problem is variations in the delay. Agreed, if only you had some way of knowing what the delay was. But as you point out, if you did, it would not really be a problem (just as the light travel down the cable from the GPS to the computer is not really a problem since it can be calculated easily and thus fixed. Also, it is usually much much smaller than the other delays.) You could patch the interrupt code to flap some bits on the printer port and put a scope on them. You could probably write a hack that ran in user mode and flapped a printer port bit at a specified time. I'm thinking of something like wait until x, spin until x+y, read clock, flap bit, read clock. Then print the two clock times. You know the bit flapped somewhere in between them. With a bit of trial and error you could probably capture a good sample on a digital scope and read off the difference between the PPS and the flap. That would let you compute the clock offset. -- These are my opinions, not necessarily my employer's. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] Trimble Resolution T
In article pine.lnx.4.64.1110161231540.14...@info.physics.ubc.ca, Bill Unruh un...@physics.ubc.ca writes: fine. Otherwise I have never used trimble proprietary language so have no idea what driver to use. There is a driver that covers the Trimble Palisade and Thunderbolt. If you are lucky, it will work with the Resolution. If not, it is probably reasonable to add support for it. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] Trimble Resolution T
In article slrnj9oqdj.3i3.un...@wormhole.physics.ubc.ca, unruh un...@physics.ubc.ca writes: In fact it is really hard for a computer to keep to 1us accuracy because of the delays in interrupt processing. It's not the delay that is the problem. It's simple to correct for a fixed delay. (at least in theory) The problem is variations in the delay. Jitter can easily be caused by cache faults or jitter in finishting the processing of the current instruction, or having interrupts disabled. (There are probably other sources.) -- These are my opinions, not necessarily my employer's. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] Loop Frequency and Offset
In article j5q79m$2u8$1...@dont-email.me, David Woolley david@ex.djwhome.demon.invalid writes: Miguel Gon=E7alves wrote: What is a typical offset of the loop without using special oscillators? Is less than 1 us achievable? Depends on the network or reference clock, and the resolution to which the local clock can be read. It also depends upon the stability of the local oscillator. Inexpensive oscillators (as used in PCs or most servers) are quite temperature sensitive. The temperature in your box will probably change as the load on the system changes or the room temperature changes. In fact, you can use a PC running ntpd as a thermometer. This is good work and a fun read: NTP temperature compensation Mark Martinec, 2001-01-08 http://www.ijs.si/time/temp-compensation/ -- These are my opinions, not necessarily my employer's. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] Single GPS/PPS time source gets marked as a falseticker
$GPZDA ? I tried it on one unit. It didn't help. If anybody has examples of it working please let us know what type of GPS receiver you tested. -- These are my opinions, not necessarily my employer's. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] How do I prevent sudden system time jumps.
The problem is that the adjustment takes to large steps, not that it takes to long time. ntpd will slew the clock at 500 PPM. You may be willing to wait a while for a second or two, but it takes a long time if you have to adjust by several minutes or an hour. That may be OK for your setup, but you should think about it. If you are using the -x flag, be sure to check out the -g flag that will let it do one long jump at startup time. -- These are my opinions, not necessarily my employer's. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] How do I prevent sudden system time jumps.
The following is an example from ntpd.log 14 Apr 07:22:25 ntpd[1048]: synchronized to 10.0.0.5, stratum 1 14 Apr 07:22:25 ntpd[1048]: time reset +0.231004 s 14 Apr 07:23:35 ntpd[1048]: synchronized to 10.0.0.5, stratum 1 14 Apr 07:39:33 ntpd[1048]: time reset +0.318457 s 14 Apr 07:39:33 ntpd[1048]: kernel time sync disabled 0041 Is the clock at 10.0.0.5 good? If all your adjustments are in the same direction, then your system is probably broken. If they alternate signs, then something in your network or remote clocks is probably confused. One thing to try. What's in your drift file? (It's probably in /var/lib/ntp/drift or /etc/ntp/drift ...) If that it a big number like 500, try stopping ntpd, deleting that file, and restarting ntpd. -- These are my opinions, not necessarily my employer's. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] Network problems affecting NTP
In article 51c20190-11b8-4629-a762-cbf95369e...@r18g2000vbs.googlegroups.com, Eugen COCA ec...@eed.usv.ro writes: Due to a network problem on our provider, the packets were routed on a different path than usual yesterday, for several hours. This period, NTP displayed wrong offsets: ntpd assumes the network path is symmetric. If it's not, the time will appear to have an offset. You can turn things around. If you know the time at both sites, you can measure the network delays. If you turn on rawstats, ntpd will write a line to a log file for each packet exchange. (details in the html files) You can graph the offset and delays. If you have a good connection, it's easy to spot network topology changes. -- These are my opinions, not necessarily my employer's. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] Garmin firmware update - GPS 18x 5Hz software version 3.20
In article banlktikc4ydiycbw4no2ay2o-r+dfp9...@mail.gmail.com, steven Sommars stevesommars...@gmail.com writes: You may want to watch for several days. Previous 18x LVC firmware (3.60) drifted on my system from 0.5 to 1.4 seconds over a span of 6 days. I've been running a beta version of 3.70 for several weeks with no problems. How stable is it? I don't care what the offset is as long as it's constant. -- These are my opinions, not necessarily my employer's. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] Ntp not syncing after powerfail of server
In article 35fcd513-e8ea-4feb-a33b-6d47d6556...@mac.com, Chuck Swiger cswi...@mac.com writes: On May 19, 2011, at 10:01 AM, M. Giertzsch wrote: To run into the problem that the client is not syncing to the server I -- stopped the NTP service on both client and server -- changed the time on the client back to january of 1975 -- started the NTP service on the client -- started the NTP service on the server a little later By default, ntpd will not try to correct a clock which is insanely far off. The -g flag can be used to change this; otherwise run ntpdate -b at boot to get the clock close and then run ntpd afterwards to keep the clock in sync. Running ntpdate at boot time won't work if the server isn't ready yet. I think the -g switch to ntpd will do the right thing, that is allow one big jump the first time it sets the clock. -- These are my opinions, not necessarily my employer's. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] Windows time question.
[default pool command uses 10 servers] I agree. It is absurd. It seems to indicate that the ntp folk really really do not trust the pool, and figure that if you get fewer than 10, you have a reasonable chance that a majority will deliver bad time. Ie, they appear to feel that the pool is a pretty useless souce of time. I don't think it's a matter of not-trust. I think it's just using some old code/parameters because the pool code hasn't been working long enough for things like that to get fixed or documented. -- These are my opinions, not necessarily my employer's. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] Using two NTP Server: Bad?
In article slrnir4ti8.j3t.un...@wormhole.physics.ubc.ca, unruh un...@wormhole.physics.ubc.ca writes: On 2011-04-23, Hal Murray hal-use...@ip-64-139-1-69.sjc.megapath.net wrote: All it takes is a firmware bug and to have several NTP servers using the same GPS device with the bug. What good will having more servers do then? If the probablility of 4 bad ones is that high, the probability of 5 out of 9 is also high, and 6 out of 11. Ie, in that case you are basically screwed. Of course everyone could just use the national time standards servers-- they are not liable to be wrong (maybe just overloaded). All I was trying to do is point out that the discussion of how many servers to use gets (much) more complicated if you consider what happens if some of them get the wrong time due to a software bug that is likely to hit all the servers running that software/firmware. -- These are my opinions, not necessarily my employer's. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] Using two NTP Server: Bad?
Consider that failure here encompasses serves bogus time. Suppose you have 7 sources configured and 4 of them cluster nicely around an incorrect time, with 3 clustered around UTC. I suppose it's possible but I'm not going to hold my breath waiting for an instance to appear! All it takes is a firmware bug and to have several NTP servers using the same GPS device with the bug. -- These are my opinions, not necessarily my employer's. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] Windows time question.
I would have hoped that the pool folk kept their documentation up-to-date, so I'll post that on their list. You didn't mention how to control the number of servers uses - is it just multiple pool entries? You only need one pool command. pool server-name will use as many addresses from server-name as it needs. If it doesn't get enough with the first DNS lookup, it will try again in a minute or so, and keep trying until it gets enough. The default is to collect 10 servers. If you don't like that, you need to add something like: tos maxclock 5 Using more than one pool command might be interesting if the names are in different administrative domains. If one is broken, the other might work. If you use more than one pool command, ntpd takes the answers from whichever DNS server answers first. I think it will drop servers that don't work and get new ones to replace them. -- These are my opinions, not necessarily my employer's. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] ntpq 4.2.7p153 only shows 5 peers
In article iof9dn$alt$1...@speranza.aioe.org, Edward T. Mischanko etm1...@hotmail.com writes: ntpq 4.2.7p153 only shows 5 peers when many more are configured? server 127.127.20.0 minpoll 3 maxpoll 3 prefer fudge 127.127.20.0 time2 0.800 refid PPS flag1 1 flag3 1 server tick.cerias.purdue.edu burst server tock.cerias.purdue.edu burst It's a bug/feature that was introduced a while ago when the DNS stuff was cleaned up. I forget the term, but servers that get added via DNS don't show up to ntpq until they get enough responses. -- These are my opinions, not necessarily my employer's. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] new driver development
In article hiUjp.6071$yl2.1...@newsfe07.iad, Bruce Lilly bruce.li...@gmail.com writes: Mutexes can work across processes; look for pthread_mutexattr_setpshared. Thanks. The first party who needs the mutex (if not already initialized) initializes it. Destruction is another issue. How do I decide who is first? There are a lot of complications (e.g. the mutex (and its attributes) typically are mapped to different addresses in different processes). What happens to a mutex in shared memory if all the processes using that memory go away? Does that shared memory stay around, or would the mutex have to be reinitialized if one of the programs using it gets restarted? Is there any way to test to see if a mutex has been initialized? -- These are my opinions, not necessarily my employer's. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] new driver development
In article 4da0c50f.3050...@comcast.net, Richard B. Gilbert rgilber...@comcast.net writes: On 4/9/2011 12:55 PM, Hal Murray wrote: Is there any way to test to see if a mutex has been initialized? You initialize your mutex BEFORE it is needed. Right. But consider the case of ntpd and gpsd. Who initializes it? Even if you say You must start X before Y (which looks pretty ugly to me) you need something to cover the case where you start X, it initializes the mutex, you start Y, and now you want to restart X. This time, it can't initialize the mutex because Y may be using it. -- These are my opinions, not necessarily my employer's. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] Help getting IRIG working
The clock puts out all of its signals on BNC. I'm not clear on the nomral assumptions with such wiring; for example, does one normally connect everything with BNC-T's, with the first and last T terminated with a 50ohm resister, or should I just connect direct through? It's audio you don't need any terminations to avoid reflections. (Well maybe if you have a few spools of coax in your setup.) This is where its perhaps most problematic for me...how do I know what's wrong with the input? How do I know what the input is supposed to do? It seems, for example, level adjustment is just guess-and-test with no way to know if I'm making it better or worse..how should one normally do this? The driver adjusts the gain. That value shows up in clockstats. Find it and watch it as you experiment. It should go to 255 when you disconnect the signal. As Dave Mills said, you will probably need an external attenuator to reduce the volume to a reasonable range. I used clip leads and resistors until I found something that worked. -- These are my opinions, not necessarily my employer's. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] Secure NTP
In article slrnionrs2.teh.koste...@stasis.kostecke.net, Steve Kostecke koste...@ntp.org writes: On 2011-03-24, Hal Murray hal-use...@ip-64-139-1-69.sjc.megapath.net wrote: Yes. The encryption also verifies that you are talking to the server you think you are talking to rather than an imposter. NTP Authentication adds signatures to the packets. There is no encryption. Thanks for the correction. -- These are my opinions, not necessarily my employer's. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] Secure NTP
In article ghps58-1a@mail.specsol.com, j...@specsol.spam.sux.com writes: When I see questions like this my first response is Why all the bother?. There is nothing secret or proprietary about the time of day. Since all NTP servers provide UTC, the service reveals nothing about the machine other than the fact that the clock is correct. If you don't want your resources utilized by outsiders, you just block access to the NTP port for everyone but your own clients as a blocked port uses less resources than denying an unsucessful authorization does. Am I missing something?? Yes. The encryption also verifies that you are talking to the server you think you are talking to rather than an imposter. -- These are my opinions, not necessarily my employer's. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] ntpd -q and driftfile
oh, come off it. Your reaction time is nowhere near what ntp -q would give you. Using ntp -q run once every hour, and assuming say a 20PPM drift for the crystal, his clock would be out by less than a 100 ms due to the drifting, and your reaction time with your watch ( and wyour watch) are nowhere near that accurate. Actually, it is on the borderline of the achievable. You phase lock your finger bounces to the second ticks and just go that little further on the exact time. It's certainly good to 200ms and possibly good to 100ms. That would be a fun project for a science fair. Build a setup. Collect a lot of data. How much does it vary? For one person, between people, ... How close can you get if you are trying to set a clock? Can you calibrate the person and subtract off a constant? What if you get multiple samples? Many years ago, my boss did that sort of stuff. I think he said that normal reaction time from light-on to button-press was 250 ms. It's faster if there is no penalties for false presses. -- These are my opinions, not necessarily my employer's. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] ntpd -q and driftfile
You can probably do better than that. That is for the reaction to something that you do not expect. Ie, you have decide that the event has occured and then send the messages to your finger to press. However for the timing, you know exactly when it is going to occur. You can get yourself into a rhythm on each second, and then finally press the button. So I suspect you can do about .1 sec if you try hard. That would be another good science fair type experiment. How long does it take to respond to a predictible event vs to an unpredictable stimulus. What is the distribution? -- These are my opinions, not necessarily my employer's. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] new driver development
In article 2wlgp.34776$d46.31...@newsfe07.iad, Bruce Lilly bruce.li...@gmail.com writes: o POSIX mutex for synchronized access to shared memory for updates -- obviates mode 0 / mode 1 / OLDWAY I'm far from a POSIX wizard. When I google for POSIX mutex I get a bunch of hits that all are part of pthreads. Does that stuff work across processes rather than threads? The mutex needs to be in shared memory so both processes can get at it. Right? Who initializes it? -- These are my opinions, not necessarily my employer's. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] Large offset and lots of time resets
In article Xns9E95DA199C4695r68mtnbtdvsdr@194.109.133.242, Maarten Deen zq...@kf4nyy.ay writes: ... 23 Feb 20:42:53 ntpd[27664]: time reset +2.421296 s 23 Feb 20:52:05 ntpd[27664]: synchronized to 217.67.231.25, stratum 2 23 Feb 20:58:31 ntpd[27664]: time reset +2.356457 s 23 Feb 21:04:48 ntpd[27664]: synchronized to 217.67.231.25, stratum 2 23 Feb 21:14:26 ntpd[27664]: time reset +2.400477 s This goes on and on and on. And regularly ntp stops having connection, reach goes to 1 and builds up again. Never that it is a poll that's been missed, it is always back to 1 as if ntpd has started itself up again. I'm using ntp 4.2.4p4 on a Ubuntu box. Linux kernel 2.6.24. What's wrong in this setup? My guess is that the clock on your system is broken. Are you using some sort of power saving that is changing the CPU frequency? Have you tried a different clocksource? (google for linux clocksource) -- These are my opinions, not necessarily my employer's. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] Getting PPS to work with Oncore ref clock
In article ia4v28-a6c@p4x2400c.home.lordynet.org, David Lord sn...@lordynet.org writes: or even a LED + diode + resistor will be enough to show if there is a pulse. Radioclkd2 can be used in debug mode to confirm the single DCD pulse and width. Some devices put out a narrow (10 microsecond) pulse. I don't know what the Oncore does. I have one PC that works fine with a narrow pulse and another PC that doesn't see it. I kludged together a pulse stretcher (diode, R, C) and it started working. Well, I thought it was working. It turns out that it only sees some of the pulses. -- These are my opinions, not necessarily my employer's. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] Detecting bufferbloat via ntp?
In article slrnilckj1.3pd.nom...@xs8.xs4all.nl, Rob nom...@example.com writes: Hal Murray hal-use...@ip-64-139-1-69.sjc.megapath.net wrote: It would be good if someone (or someones) has actually been collecting rawstats for a long period, to serve as a baseline. Bufferbloat is a relatively new phenomenon. I'll bet it's been there for a long time. (meaning at least several years) I see delays up to 3 seconds, but only when I'm downloading a big file. I assume it's all queued up at the other end of my DSL link, but I can't prove that. When downloading a single big file causes you 3 second delays, you simply have set an insanely large TCP window. I didn't tweak anything. It's possible that my 3 second delay was while using Firefox rather than a simple download. -- These are my opinions, not necessarily my employer's. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] Detecting bufferbloat via ntp?
In article ijhd7d$au6$1...@news.eternal-september.org, E-Mail Sent to this address will be added to the BlackLists Null@BlackList.Anitech-Systems.invalid writes: As far as I can tell the products have all the knobs they need to be properly configured to deal with the issue. The ISPs / NOC staff / IT departments / ... just aren't taking the time, to properly / fully configure the products, rather than just the minimal configuration necessary to make it work. The defaults have to work reasonably well. The support staff doesn't have enough time to help each user setup their system correctly. -- These are my opinions, not necessarily my employer's. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] Getting PPS to work with Oncore ref clock
In article aanlktimlss1bjtrde77av92-wwfyr_g7ovm7zd0yr...@mail.gmail.com, Chris Albertson albertson.ch...@gmail.com writes: (It's a 75 foot serial cable.) I suspect a software configuration problem or maybe polarity. If the polarity is wrong, you will just use the other edge. It will be off in timing by the pulse width. It will appear to work, just get the wrong answer. -- These are my opinions, not necessarily my employer's. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] Detecting bufferbloat via ntp?
In article 87lj1q823k@cruithne.co.teklibre.org, d...@taht.net (Dave =?utf-8?Q?T=C3=A4ht?=) writes: Presently! It sounds like a lightly modified ntp server - or set thereof - could actually attempt to keep track of this information. I don't think ntpd needs any modifications. Do ntp clients use an ephemeral udp src address for a query? The server listens on port 123 and sends the answers back to the source port they came from. ntpd is both a client and server. It sends from port 123. -- These are my opinions, not necessarily my employer's. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] Detecting bufferbloat via ntp?
It would be good if someone (or someones) has actually been collecting rawstats for a long period, to serve as a baseline. Bufferbloat is a relatively new phenomenon. I'll bet it's been there for a long time. (meaning at least several years) I see delays up to 3 seconds, but only when I'm downloading a big file. I assume it's all queued up at the other end of my DSL link, but I can't prove that. ntp tries to filter out the packets with long (queueing) delays so you will see the bad stuff in rawstats that you won't see in peerstats or ntpq -peers I was afraid of that. What qualifies as a long queueing delay? Normally, ntpd uses the lowest delay of the last 8 packets. There is a huffpuff filter. (details in miscopt.html) -- These are my opinions, not necessarily my employer's. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] Detecting bufferbloat via ntp?
In article iithjg$83s$1...@news.eternal-september.org, David Woolley david@ex.djwhome.demon.invalid writes: Dave T=E4ht wrote: A well managed router would prioritised NTP traffic and isolate it from=20 the buffering, which should only really affect bulk TCP traffic. How many routers to that? -- These are my opinions, not necessarily my employer's. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] Detecting bufferbloat via ntp?
In article 87aai6d0t6@cruithne.co.teklibre.org, d...@taht.net (Dave =?utf-8?Q?T=C3=A4ht?=) writes: I've been racking my brain trying to come up with a good way of semi-passively detecting bufferbloat at the datacenter. What would wild swings in latency on the order of seconds from a ntp client register on a ntp server as? Are you trying to detect it in real time, or collect long term data? Turn on rawstats. (info in monopt.html) That will log 4 time stamps for each ntp packet received. If the time on both systems is accurate, that will give you a pair of one-way times. ntp tries to filter out the packets with long (queueing) delays so you will see the bad stuff in rawstats that you won't see in peerstats or ntpq -peers -- These are my opinions, not necessarily my employer's. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] Odd results with Oncore UT+ ref clock
In article AANLkTi=ju-fv8e-q4rh3p7rpdf3cqncn7dpec9zrc...@mail.gmail.com, Chris Albertson albertson.ch...@gmail.com writes: What I'm looking for here is advice on what to check. Is there anything interesting in the ntp log file? (defaults to syslog which is usually /var/log/messages) Is there anything interesting in clockstats? Do you have clockstats enabled? -- These are my opinions, not necessarily my employer's. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] Peridoc anomalies in loopstats file ...
In article g7do08-bma@gateway.py.meinberg.de, Heiko Gerstung heiko.removethistext.gerst...@meinberg.de writes: Mmmh, the polling cycle I use is 8s (patched ntpd), and if you look at this sample sequence ...: Does the problem happen without that patch? -- These are my opinions, not necessarily my employer's. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] GPX18x LVC 3.50 firmware - high serial delay problem workround
In article AANLkTi=E0_9LzCDg9esXx7yZp_NJGmSU+cuS=FY9=8...@mail.gmail.com, Chris Albertson albertson.ch...@gmail.com writes: Could this be automated? Maybe, to some degree. The reference clock driver would need to have a survey mode setting where it would run for many hours and compare it's own time to others. NTP does this already, almost, what it lacks is a way to capture the measured offset and fold it back to a config file. If you run the to-be-calibrated server in noselect mode, it won't pollute your local clock. If you turn on peerstats, you can get lots of data about how far off that clock is. That assumes your local clock is correct. If you believe that the PPS is correct, you only have to get the NMEA text close-enough. You can easily get there using typical network connections. -- These are my opinions, not necessarily my employer's. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] How to keep Linux server in Chicago and Mumbai in sync to within 5 microseconds
In article 20110112162440.ab9f81c...@ptavv.es.net, Kevin Oberman ober...@es.net writes: GPS satellites broadcast time in TAI, not UTC. Currently the offset between TAI and UTC is 15 seconds. Actually, they broadcast GPS time which is offset from TAI by the number of leap seconds that had already happened when they started using GPS time. They also include the offset from GPS time to UTC so you can get UTC from GPS satellites. (They also include leap second warnings so you can do the right thing when leap seconds happen.) From http://tycho.usno.navy.mil/leapsec.html As of 1 January 2006, TAI is ahead of UTC by 33 seconds. TAI is ahead of GPS by 19 seconds. GPS is ahead of UTC by 14 seconds. We've had one more leap second since 2006, so that would be 34 and 15. http://maia.usno.navy.mil/ser7/tai-utc.dat 1999 JAN 1 =JD 2451179.5 TAI-UTC= 32.0 S + (MJD - 41317.) X 0.0 S 2006 JAN 1 =JD 2453736.5 TAI-UTC= 33.0 S + (MJD - 41317.) X 0.0 S 2009 JAN 1 =JD 2454832.5 TAI-UTC= 34.0 S + (MJD - 41317.) X 0.0 S -- These are my opinions, not necessarily my employer's. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] How to keep Linux server in Chicago and Mumbai in sync to within 5 microseconds
In article aanlktincze9exucs_6ioz3geoexmzt3etoj7bpl6b...@mail.gmail.com, Chris Albertson albertson.ch...@gmail.com writes: On Wed, Jan 12, 2011 at 11:25 AM, unruh un...@wormhole.physics.ubc.ca wrote: ... There is nothing in CDMA that would require very accurate absolute time I presume. Thus there is no real reason why they should deliver absolute time with high precision. ... Tower triangulation based on time of flight delay is used to determine location of a cell phone. There is a requirement to attach location to 911 calls and triangulation is a kind of last resort method for non gps enabled phones. It's not a last-resort. It's an ecnomoic decision. Do you replace all the cell phones with new ones that have GPS chips or do you upgrade all your towers with gear to do the triangulation? Note that being able to do triangulation doesn't require that the time at the towers be accurate, just that they be in sync. Triangulation would all work if all the clocks were off by 1.37 seconds. (as long as they all used the same value of 1.37) -- These are my opinions, not necessarily my employer's. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] Problem with Trimble Placer 450 - gpsd - ntpd
In article 201101011238.p01ccalc009...@heartache.daihls.com, Bruce Dale bd...@stny.rr.com writes: ... Do you have contacts at gpsd with which you work out problems? If not, I would be glad to provide the email addresses of the people who were helping me. I can provide info about the problem, my solution, and a test bed for a fix. There haven't been enough interesting problems to worry about it. The SHM api is pretty simple. -- These are my opinions, not necessarily my employer's. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] What version of ntpd am I really running??
In article ig0ajc$sp...@speranza.aioe.org, Edward T. Mischanko etm1...@hotmail.com writes: What is the prper way to determine the current version of ntpd that you have installed or are running? One way to find out what you are (or were) running is to scan /var/log/messages or wherever your syslog goes. There should be something like this: Jan 4 10:53:39 shuksan ntpd[24383]: ntpd 4.2.7p...@1.2416-o Tue Jan 4 18:52:49 UTC 2011 (1) (The date on the end is when the system was compiled.) -- These are my opinions, not necessarily my employer's. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] Problem with Trimble Placer 450 - gpsd - ntpd
In article 201012291506.obtf6xa5021...@heartache.daihls.com, Bruce Dale bd...@stny.rr.com writes: ... I have a Trimble Placer 450 GPS receiver generating a PPS signal and NMEA sentences to gpsd (2.95), which passes that information to ntpd (4.2.6p2) over the shared memory interface (reference clock driver type 28). When the Placer 450 has a consistent 3D fix on several satellites, ntpd selects SHM(1), the PPS signal, as the syspeer. No problem. The problems occur when the Placer 450 loses satellite fix. In my environment, this happens multiple times a day. ... The time server is a system consisting of the Trimble GPS receiver, gpsd and ntpd. The gpsd folks are happy to implement a fix to the problems with the Placer 450, but how are problems between gpsd and ntpd resolved? Specifically, how should a GPS receiver report a loss of satellite fix to ntpd over the shared memory interface? Should ntpd ignore the PPS during the interval of no fix? And what is the correct behavior of ntpd when this happens? If the NMEA signal says no-good, gpsd should ignore the PPS pulses. It should probably wait for several seconds of good NMEA messages before turning on PPS. There isn't any way for gpsd to say clock-died via the SHM interface. All it can say is here-is-data. ntpd should timeout if gpsd stops feeding it data. It may take 2 minutes. If that's not good enough we can look into tweaking the SHM interface. -- These are my opinions, not necessarily my employer's. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] signal_no_reset: signal 18 had flags 20000
In article 67578d3d-adff-469a-a57b-a9ff54460...@39g2000yqa.googlegroups.com, bombjack bombjac...@gmail.com writes: I have STFI like crazy, but can't find an explanation for the error code signal_no_reset: signal 18 had flags 2 Could someone explain what this mean? Or even better, give me a hint where I can find info about all these error codes etc for ntp Most ntpd error/syslog messages come from a call to msyslog which has printf type arguments. If you have the sources handy, you can grep for part of the text from the error message. In this case, grep signal_no_reset: signal . -r gets one hit in ./libntp/syssignal.c Note that it helps to have the sources that correspond to the code you are running. You can get that either by tracking down the sources for the code you are running, or grabbing a known source and rebuilding and installing. http://support.ntp.org/bin/view/Main/SoftwareDownloads -- These are my opinions, not necessarily my employer's. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] ntp# pid 1012 (ntpd), uid 0: exited on signal 11
In article idpmi8$ll...@speranza.aioe.org, Edward T. Mischanko etm1...@hotmail.com writes: ntpd not working on a Soekris NET4501 with a nanoBSD install. Any ideas where to start? ntpdate is working OK. signal 11 is segmentation fault (illegal address) so something is really screwed up. Lots of people run ntpd on NET4501s so it's probably something simple. What version of ntpd are you running? Was it compiled for your OS? Have you looked in /var/log/messages and such? There is a -d switch to ntpd. http://www.eecis.udel.edu/~mills/ntp/html/ntpd.html I think it disables switching to daemon mode so the debugging printout goes to your terminal. -- These are my opinions, not necessarily my employer's. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] Question on ntpd -g -q option and Step Threshold
In article 0a89d263-0355-4dee-8c9c-b4912076d...@w38g2000pri.googlegroups.com, SteveW qsw...@email.mot.com writes: I wanted to confirm that my understanding of the step threshold and how it applies to ntpd used with -g -q options (similar to ntpdate) is correct. With the default step threshold of 128 ms, if a machine's time is within 128 ms of the ntp server's time, when it syncs with the server, will it exit from ntpd -g -q without stepping the clock? This appears to be the behavior I am seeing. I tried setting the clock on my machine off by 110 msec from the ntp server and it didn't step the clock. When I set it off by 220 msec it did step it. There are two ways to adjust the time. One is to step it by just smashing the desired time into the clock register. The other is to slew it by slightly adjusting the rate the local clock ticks until it has gained/lost the right amount of time. ntpd normally steps for changes over 128 ms and slews for changes under. In your 110 ms test, I expect it to slew the clock rather than step it. It may exit before the slew has finished. You can check by waiting a few minutes and then running it again. -- These are my opinions, not necessarily my employer's. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] -8.625 ms offset between ref clock (NMEA PPS) and nearest stratum 1 peer
In article ibm99p$ls...@speranza.aioe.org, Edward T. Mischanko etm1...@hotmail.com writes: I am running a Garmin 18x LVC with PPS enabled; it is very stable and I beleave it to be accurate. My nearest back-up server is consistantly around -8.625ms away. This is so much that other clients on my network see me as a false ticker even with a prefer setting. I am on a farly fast ADSL connection to the internet, but of course, it is non-symetrical. Other users I have conversed with don't seem to have this wide of an offset, so something is wrong somewhere. Yes, I can fudge the offset away on the ref clock, but it doesn't seem right to tell a ref clock to lie about what time it is just to keep everyone else happy? Have you tried a few other servers on the internet? How far away is that server? It's not crazy to be off that far due to asymmetrical routing on the internet. (That's even if you don't have an A in your ADSL connection.) -- These are my opinions, not necessarily my employer's. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] help needed for ntpd ipv6 setup
In article ibdgsl$40...@news.eternal-september.org, David Woolley da...@ex.djwhome.demon.invalid writes: joan lee wrote: Wireshark runs on PC1 shows NTP client messages from 2002::c0a9:0102 + are received properly, but instead of replying with NTP server messages, + PC1 is sending ICMPV6 unreachable (administratively prohibited) messages. Could anybody gives me some hint on why NTPD ignores the IPv6 NTP + client messages? Administratively prohibited doesn't sound like a condition than an application program can generate. I would look at your firewall. First, I'd check the simpler solution of does the version of ntpd he is running support IPv6. What does netstat -ul say? -- These are my opinions, not necessarily my employer's. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] Local clock - sync issue
I've found that bad things happen at much above 50ppm. I have lots of systems that work fine at well over 50 ppm. With a good PPS, the drift will track temperature and you can use the system as a thermometer. -- These are my opinions, not necessarily my employer's. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] Test ntpd performance
The most I've seen a single netperf *single-byte* burst-mode test do is on the order of 350K transactions per second, and that is when netperf is bound to a core other than the one taking interrupts from the NIC to get some additional parallelism. In a netperf TCP_RR test, netperf does virutally nothing but a send()/recv() pair and a couple conditionals and adds. Is that sending 1 byte of data per packet, or batching up several/many bytes per packet? (and mostly measuring the CPU time for a one byte send/recv pair) Even if you have a test program that blasts lots of packets, that won't mimmic real traffic. It's bypassing all the setup of ARP and router slots. The only really useful data I've seen is the numbers from the people running heavily loaded servers. -- These are my opinions, not necessarily my employer's. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] Linux clocksource
[Linux TSC calibration troubles.] YOu do not say which kernel you are using. I believe that the problem with tsc has been fixed in later kernel versions. Does anybody know which version it was fixed in? -- These are my opinions, not necessarily my employer's. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] Decommission NTP server
In article qgofk7-dsb2@ntp.tmsw.no, Terje Mathisen terje.mathisen at tmsw.no writes: John Kennedy wrote: We have an NTP server that we want to decommission. Before we do this we need to know which servers are using this server to their time. Is there a way to log which servers are syncing off of this server? I have looked at the logconfig directive and can not find enough info on it. I have tried logconfig=all but that does not help. 'ntpdc -c monlist' is your friend, it will list the last 600-700 clients. If you have a lot more clients than this, then you'll have to either install a new/dev ntpd version or insert WireShark or a similar sniffer on a mirror of the server switch port. Or fix the clients you can see, then check again to see who is still using the server. -- These are my opinions, not necessarily my employer's. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] Maximum time2 fudge value for NMEA refclock?
I created a symlink called gps0 to my serial port ttyS0, because it told me that in the logs when I tried to start up ntpd without it. Can somebody who uses the NMEA driver in linux tell me if there is something else that I should have compiled in? gpsd works fine, but every time I switch to the NMEA type 20 driver, it doesn't work. It doesn't seem to write anything to the system logs, or clockstats either. You also need a bit more magic. There should be good documentation someplace, but I don't know where it is. Try man ldattach I'm using this in /etc/sysconfig/ntpd on Fedora if [ ! -e /dev/gps12 ]; then # GPS 18x LVC ln -s /dev/ttyS0 /dev/gps12 setserial /dev/ttyS0 low_latency ldattach 18 /dev/gps12 ln -s /dev/pps0 /dev/gpspps12 fi -- These are my opinions, not necessarily my employer's. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
[ntp:questions] Batch of 4 NTP RFCs are out: 5905-5908
RFC 5905 Title: Network Time Protocol Version 4: Protocol and Algorithms Specification Pages: 110 Obsoletes: RFC1305, RFC4330 URL:http://www.rfc-editor.org/rfc/rfc5905.txt RFC 5906 Title: Network Time Protocol Version 4: Autokey Specification Pages: 58 URL:http://www.rfc-editor.org/rfc/rfc5906.txt RFC 5907 Title: Definitions of Managed Objects for Network Time Protocol Version 4 (NTPv4) Pages: 26 URL:http://www.rfc-editor.org/rfc/rfc5907.txt RFC 5908 Title: Network Time Protocol (NTP) Server Option for DHCPv6 Pages: 9 URL:http://www.rfc-editor.org/rfc/rfc5908.txt -- These are my opinions, not necessarily my employer's. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] SNTP with 1ms of precision?
In article aanlktil2jbpj9p5eqfykgnjev8_ejr6ic_0cuodgf...@mail.gmail.com, Marcelo Pimenta marcelopiment...@gmail.com writes: I thought that SNTP was used in a kind of Intranet with no routers and no internet, so the low latency there is no need of so much calculations to adjust the clock. I haven't been able to figure out what you are really trying to ask. The packets on the wire are the same for NTP and SNTP. The results you will get will depend upon the hardware and software you are using. If you have a good/reference server on your local LAN, it should be reasonable to get the clock on a client to within 1 ms most of the time. I would expect occasional quirks. How many would depend mostly on the software you are using. (If somebody is going to get hurt or killed if your clock is off by more than 1 ms, you should do something else.) You should probably setup a test bed and see how well it actually works. -- These are my opinions, not necessarily my employer's. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] proprietary hardware clock as NTP reference source
In article 40aa2d27-9826-4948-a405-e5b311358...@z8g2000yqz.googlegroups.com, apobrien apobriens...@gmail.com writes: Hello list, I have a set of proprietary hardware timing cards (Symmetricom bc635PCIe) which synchronize their clocks using a dedicated interconnect. As you might imagine the timing card conditioned time drifts from that of the hosts they're installed in. What I'd like to do is make the master timing card's time into a NTP reference clock then use NTP to distribute that time to the other hosts in the (private) network. I've looked at Orphan mode and undisciplined local clocks but they only refer to the host's software clock if I'm not mistaken. I've also searched through the archive but I'm afraid I lack the appropriate terminology to get meaningful results. Can someone point me toward (some google words maybe) creating an arbitrary NTP reference source (under Linux)? I think I'm just missing something very basic. I've been looking at the LinuxPPS project as these cards output a PPS that I might use to condition the host clock using Linux PPS but I don't have a 8250 serial port on these new fangled PCs. TIA! Andy I think you have two choices. One is to write a stand alone program that talks to your hardware and puts the info into shared memory where the shared-memory driver (driver 28) can get it. gpsd works this way. The other is to write your own refclock-driver. The usual approach is to find a driver that is as close as you can get and modify it. If your changes are small and don't break the old moce, you may be able to convince the ntp project to merge them into the main source package. It may be tricky to get started this way. The best documentation is to read the code. There is some overview here, but it probably won't help much until you look at the code: http://www.cis.udel.edu/~mills/ntp/html/howto.html -- These are my opinions, not necessarily my employer's. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] Restrict vs DNS lookup
In article 4c0bccfe.30...@ntp.org, Danny Mayer ma...@ntp.org writes: On 6/6/2010 3:24 AM, Hal Murray wrote: https://bugs.ntp.org/show_bug.cgi?id=1568 Dave Hart points out that ntp-dev has a server option to the restrict command. Description here: http://www.eecis.udel.edu/~mills/ntp/html/accopt.html Would somebody who uses restrict please check to see if this does what you want. Hal, If this is about my suggestion to add a server option for restrict lines to allow easier control of packets from servers defined in the various server/pool, etc. lines then neither of these references describe that. Both mention restrict server Yes, the part in accopt.html is hidden in the fine print. The goal is to allow through packets from the servers you list even though there may be other restrict lines. I think restrict server will do that. I hope somebody more familiar with restrict will double check. I'm not sure I understand the intention of your note. Danny There have been occasional discussion here about the interactions of DNS with restrict. There was one recently. I entered the bug to collect thoughts and keep it from falling through the cracks. It's possible that some work on the documentation will make me happy and help others avoid confusion. I think it's simple after you understand it, but it took me a while to figure that out and I'm not really sure I've got it right. I think part of my confusion is that there are two things you might want to do with restrict and DNS. One is the case you mention, let through packets from servers that are looked up via DNS when your restrict line would otherwise block them. I think the current code will do that. The other possibility it to block servers from a CIDR block, even if you get one from DNS. This isn't interesting if you trust the people running the servers you are using and if you don't trust them, why are you using their servers? But you might want to skip servers in XXX (pick your favorite bad guy) even if they make it into the pool. I think I'd be happy as documentating the latter case as not working. -- These are my opinions, not necessarily my employer's. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
[ntp:questions] Restrict vs DNS lookup
https://bugs.ntp.org/show_bug.cgi?id=1568 Dave Hart points out that ntp-dev has a server option to the restrict command. Description here: http://www.eecis.udel.edu/~mills/ntp/html/accopt.html Would somebody who uses restrict please check to see if this does what you want. -- These are my opinions, not necessarily my employer's. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] Log client connections to ntpd server
In article cca1921f-cdee-43f8-bc28-31a2a0982...@a27g2000prj.googlegroups.com, s a n ihsa...@gmail.com writes: Thanks guys.. for the monlist and the tcpdump script info. Was exactly what I was looking for. Dave Hart recently cleaned this area up so the replacement for monlist in ntp-dev now works for lots of clients. Try ntpq -c mrulist It would be nice if more people helped test it. -- These are my opinions, not necessarily my employer's. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] Is this normal?
In article c430da47d51d0c4b940f6928bae21f4626a3f16...@sbkmxsmb06.win.dowjones.net , Russell, David david.russ...@dowjones.com writes: The device is a piece of networking equipment and so I doubt that the crystal is temperature controlled but since it is in a data center the temperature and power demand is steady. If you could see the graph you would see that it seems to oscillate around -3.05 us/s over a 24 hour timeframe. The graph has a pretty nice looking sine wave with a 12 hour period. I was expecting that if the true drift rate were -3.05 then NTP would converge upon that value. Are you saying that this oscillation is typical for non-temperature (cheap) crystals or should the drift look more random? The drift follows temperature. Do you have any way to measure the temperature? Any thoughts on validating a black box NTP process? Setup NTP on another box and tell that box to use the black-box as a server. Use noselect if you want to make sure it won't try to sync to that system. Then turn on loopstats and peerstats. That will give you lots of data to look at. I had not realized until I sent out the email that no attachments are allowed. Is there a way to share a graph? Put it on the web someplace and send a URL. -- These are my opinions, not necessarily my employer's. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] Help with Resolution SMT
In article 4beb53a2.4050...@signaturealpha.com, Marc Leclerc marc-lecl...@signaturealpha.com writes: Hi, Seems the generic parser does't work with NTP with resolution SMT. Has anyone modified some source to make it work?. NTP reports answer as bad Data, but trimble studio hooked directly to the chip reports good data. What exactly have you tried? Have you tried the Palisade driver (29) with the Thunderbolt flag? Have you tried using the Trimble software to switch it to NMEA mode and then using the NMEA driver? -- These are my opinions, not necessarily my employer's. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] Trimble Resolution SMT
In article 4be025aa.4050...@signaturealpha.com, Marc Leclerc marc-lecl...@signaturealpha.com writes: I am trying to have NTP use the trimble resolution SMT gps module, I have tried other trimble clock driver without success so I assume that one specific to the module has to be used. Unfortunately there does not seems to be any way to search the mailing list to see if this was discuss before. I would appreciate if anyone with information could get back to me. I am trying to make this work on an embedded linux platform, I already have the linuxpps driver installed and have the /dev/pps0 node available. the module is wired to /dev/ttyS1. Using the clock driver for the palisade lead to wrong answer format messages. The Palisade driver supports several different variations of Trimble products. Which one(s) did you try? The data sheet says it speaks TISP and NMEA. I don't know which one is the default. You might try running some helper code before starting ntpd to switch it to NMEA mode, and then telling ntpd that it's a NMEA device. -- These are my opinions, not necessarily my employer's. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] How to debug PPS?
GPGGA has one added benefit: If you turn on clockstats you get a log of the current position every second, this is perfect input for a statistical averaging of the current antenna position. :-) It doesn't have the date. GPRMC is the only one I know of with both time and date. -- These are my opinions, not necessarily my employer's. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] Cheap ($29.90) GPS receiver
The serial chip does not do that. It will send the interrupt at the time it has assembled a full character. Some of the serial chips have large FIFOs and wait until the FIFO has several characters and/or there is no activity for a while. The idea is to save CPU cycles by batching interrupts. -- These are my opinions, not necessarily my employer's. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] Cheap ($29.90) GPS receiver
How precise can a 4800 baud serial port be in determining the start of a start bit? About 1/4800s i.e. about 200 microseconds? Don't the chips sample at 8 or 16 times baud rate, so couldn't the start bit be determined to ~25-30 microseconds? And if the baud rate was lot higher (the unit claims 57600 baud)? You would be using the unit in a slightly different mode to an existing GPS - just detecting the start of an RS-232 burst for the PPS, and then reading a message for the actual value What you describe could be made to work quite well, but I've never seen one that works that way. The data sheet says 1 microsecond. It also says SiRF chip set. All the SiRF units that I've tested are horrible. They aren't even trying for good timing. My guess is that the 1 microsecond was copied from another data sheet that did have a PPS signal. If anybody tests one of these, please let us know how well it works. -- These are my opinions, not necessarily my employer's. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] Cheap ($29.90) GPS receiver
I think that depends on how you set up the chip. I think you can set it up so it will interrupt on each byte. This is not terribly efficient for a serial port driver, but if it is timing you want, then that is probably what you want. On linux, I use: setserial /dev/ttyS0 low_latency Well, actually you probably want the interrupt to occur at the beginning of the first start bit of the first character, which could in theory on a 4800bd give you sub ms resolution ( and maybe much better) . That's just a constant offset. You can correct for that. -- These are my opinions, not necessarily my employer's. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] Quick sync between two computers not connected to the internet
In article 0f5de21d-2a20-40fa-ab68-a03cba5ae...@u9g2000yqb.googlegroups.com, Evandro Menezes evan...@mailinator.com writes: On Mar 23, 11:38=A0am, Chuck Swiger cswi...@mac.com wrote: If I really had to solve the latter problem, I would likely connect the m= achines to a valid NTP timesource long enough to calibrate each machines' i= ntrinsic drift from realtime, and then run time in standalone mode against = their local clock. Useless. Clocks may drift wildly due to temperature changes, be it ambient or inside the box. The temperature drifts that I've seen are much smaller than the basic calibration. On the other hand recent Linux kernels have screwed up the basic calibration on startup so rebooting a machine is much worse than a temperature change. (If they have fixed it recently, I haven't seen any announcement. You can fix it by hacking a constant into your kernel.) -- These are my opinions, not necessarily my employer's. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] Which version of Linux works best?
Modern Linux kernels don't support PPS in the sense of RFC-whateveritis. There is support for an ioctl that says wake me up when a modem signal changes. gpsd uses that to provide PPS support. I don't have any data. I believe but am not sure, that that uses an interrupt. I think so. But the point is that with the PPS support, the kernel grabs a timestamp in the interrupt routine. The ioctl stuff just wakes up the user program so it can grab the timestamp. On a lightly loaded system, that will probably work OK. But if the system gets busy, there will be more noise in the data. -- These are my opinions, not necessarily my employer's. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] GPS_NEMA, but high offset and jitter?
In article 4b8580e5$0$1616$742ec...@news.sonic.net, Kelsey Cummings k...@sonic.net writes: I'm not sure why it lost sync but I went ahead and upgraded to 4.2.7p5 from the ntp-dev port. I'm still seeing high offset and variable jitter. == xGPS_NMEA(1) .GPS. 0 l 45 64 3770.000 -311.36 75.859 -time.sr207.200.81.. 2 u 348 512 3770.6430.977 2.984 *clock.sjc.he .CDMA. 1 u 362 512 3776.348 -2.584 0.289 The NMEA driver is a bit strange. It processes two sources of time. One is the text on the serial port. The other is the PPS signal. I think it has to get close enough on the text mode before the PPS processing kicks in. How about turning off the PPS fudge flag so you know you are using the text mode. Then watch it for a while and add a fudge time2 to get it reasonably close. When that works, turn the PPS back on. -- These are my opinions, not necessarily my employer's. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] NTP servers redundancy
In article 1245298544.863191263816061789.javamail.r...@spooler9-g27.priv.proxad.net, j...@free.fr writes: Hello everybody, for a ship project, we've got a LAN with two NTP servers synchronised by a masterclock, itself synchronized by GPS. The idea was to have redundancy between the 2 servers, to ensure continuous synchronization to the clients in case one of the server went down. What sort of thing is your master clock? Does it speak NTP? If not, how does the timing info get from the master to the NTP servers? How many systems are going to be using your NTP servers? What sort of accuracy do you need? What sort of reliability? Suppose you only had one server. How long would it take to fix it if something went wrong? How expensive would that be? (In terms of lost data rather than just the cost of fixing it.) I wonder now if the dual NTP source architecture is the good one : Maybe, but maybe not. In general, 2 is a nasty number for ntp. How do you tell which one is better? One server is simpler and easier to understand. Is the clock hopping resulting of this dual architecture a real problem ?? Probably not. That's the sort of thing you have to try. You should also try seeing what happens if you only have one server and it gets unplugged. In general, if a system has been up for a while (day or two) it will coast if it can't contact any servers. That coasting may be good enough. It will drift more if the temperature changes. Does the NTP algorithm enable that when a server go down (or is no more synchronized), its clients switch easily to the second one, without any visible synchronization interruption ?? Will it be transparent for the clients when the first server will be back ? If one dies, the clients will switch to the other one. Maybe they won't do it as fast as you would like. There may be hiccups. It depends on how accurate you need the time to be. How shall be the clients configured for this architecture ?? server host1 server host2 Thanks a lot for your help. Jean -- These are my opinions, not necessarily my employer's. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] NTP servers redundancy
The fundamental problem with two servers is this: which one do you believe when the two differ? You know that at least one of the two must be wrong but it's impossible to determine WHICH one! You get more than time from a packet exchange. You also get a good idea of the accuracy. Two clocks can give (slightly) different answers without either being wrong. If that happens you can use either one. The problem is that IF they give wildly different answers (perhaps because one of them is broken) THEN you don't know which one is good. -- These are my opinions, not necessarily my employer's. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] Is dispersion jitter in all situations
In article slrnhk7469.kji.un...@wormhole.physics.ubc.ca, unruh un...@wormhole.physics.ubc.ca writes: As I said, it is possible that all outgoing ntp requests go via a 1Gb ethernet, and all return packets go via a 300baud modem. In that case the estimate of 1/2 the round trip would be a good estimate of the systematic error. Unfortunately there is absolutely nothing ntp can do on its own to figure that out. A more likely cause for asymmetry is queueing delays on home DSL links. Nasty. -- These are my opinions, not necessarily my employer's. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] ntpd goes into oscillation
In article slrnhjkta6.4e4.un...@wormhole.physics.ubc.ca, unruh un...@wormhole.physics.ubc.ca writes: ntpd has suddenly broken out into oscillations. it is fed by a Garmin 18 LVC PPS via shm. The oscillation has a period of just under an hour ( about 50 min) and an amplitude of about 10usec. in the offsets ( amplitude of about .005PPM in the rate). Since this is acting as the clock for a number of other machines, they are also showing the oscillation especially in the rate. While I suppose this could be something in the GPS itself, it looks more like an oscillation in ntpd. Nothing changed when the oscillations started. Ntpd had been started on Dec 14, and this change began on Dec24. You can see the graph on www.theory.physics.ubc.ca/chrony/chrony.html , the graph for the machine called string. Has anyone else ever seen this kind of thing? What OS? (version?) What is the temperature like? If you feed a sawtooth into a PLL, the offset will be the derivative, a square wave. The amplitude of the square wave is smaller with higher gain. A sawtooth with a 1 hour period is possible from air conditioners. I've seen oscillations on boxes using the pool, or at least stuff that looks like oscillations to my eyeball. That's on Linux. (They have been fixing the timekeeping code. I wouldn't be surprised by anything.) -- These are my opinions, not necessarily my employer's. I hate spam. ___ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions
Re: [ntp:questions] Tobit LAN!Time DCF77 receiver not working
Over the years I've come to the conclusion that it's easier to rewrite from scratch than to reverse engineer someone else's code. Especially so when no effort is made to document what the code is doing, the algorithms used, what the variables represent, etc. Far too many people code in just this way! Yes there is a lot of shitty code out there, but sometimes it's the only documentation you can get. If nothing else, you may want to scan it to extract some key ideas needed to write your own version. Personally, I don't want comments that tell me stuff I can figure out by glancing at the code. What I want is clues about the strange things like feature X doesn't work on OS Y, or watch out for overflows. -- These are my opinions, not necessarily my employer's. I hate spam. ___ questions mailing list questions@lists.ntp.org https://lists.ntp.org/mailman/listinfo/questions
Re: [ntp:questions] Tobit LAN!Time DCF77 receiver not working
Does the software include any documentation explaining what this output means? It certainly does not seem to be in any way useful unless there is some explanation of the output. If there is none, I'd drop it in the garbage and use something else! Or look at the source code. -- These are my opinions, not necessarily my employer's. I hate spam. ___ questions mailing list questions@lists.ntp.org https://lists.ntp.org/mailman/listinfo/questions
Re: [ntp:questions] Busybox and ntp
but have hit some snags where certain commands are not supported in busybox (service start/stop/reset, chkconfig). service is just a script/program to run one of the start/stop scritps. It may be a redhat/fedora/linux thing. google will find the man page. chkconfig is a utility to setup the links for starting and stopping services. Again, there is a man page. If you aren't familiar with how servers get started/stopped, look in /etc/rc.d and friends. -- These are my opinions, not necessarily my employer's. I hate spam. ___ questions mailing list questions@lists.ntp.org https://lists.ntp.org/mailman/listinfo/questions
Re: [ntp:questions] Monitor of NTP service
I have several AIX Unix 5.2 serves that point to a Cisco router as the NTP time source. Is there any way I could monitor the time source and against the servers to ensure that they are in sync? Or how are others performing time source/server monitoring? I'm assuming you want to do the monitoring from another system. Setup ntpd on that system. Turn on some of the statistics collection. details are in monopt.html loopstats will tell you what ntpd thinks of it's local clock. peerstats will tell you what the other systems look like. I use filegen xxx file xxx type day link to break things up into a file per day. Then add server xxx for each system you want to monitor. you can adjust minpoll and maxpoll to control how often they get polled you can turn on noselect if you think the monitoring system might get confused. That will collect a bunch of data. Then you can graph it or whatever you want. -- These are my opinions, not necessarily my employer's. I hate spam. ___ questions mailing list questions@lists.ntp.org https://lists.ntp.org/mailman/listinfo/questions
Re: [ntp:questions] mtp srver load figures?
I'm preparing to build a few ntp servers for a somewhat large population of NTP-misbehaving boxes (they top out at a 120 sec interval). So, to size this right - what kind of request load can I safely throw at ntpd on a current Xeon server? Is 200 requests per second a problem? What do you calculate? Don't calculate, measure. I just measured over 5000 packets per second on a 1 GHz Via. You didn't specify what OS you are using. That may change things by a factor of two or more. -- These are my opinions, not necessarily my employer's. I hate spam. ___ questions mailing list questions@lists.ntp.org https://lists.ntp.org/mailman/listinfo/questions
