Re: [chrony-users] Sporadic NTP dropouts (bad maximum delay ratio and maximum delay dev ratio)

[ Thread Index | Date Index | More Archives ]

Am 29.05.2015 um 20:00 schrieb Bill Unruh:

William G. Unruh   |  Canadian Institute for|     Tel: +1(604)822-3273
Physics&Astronomy  |     Advanced Research  |     Fax: +1(604)822-5324
UBC, Vancouver,BC  |   Program in Cosmology | unruh@xxxxxxxxxxxxxx
Canada V6T 1Z1     |      and Gravity       |

On Fri, 29 May 2015, Ulrich Schwesinger wrote:

Thanks for the helpful feedback so far. I contacted the server admin, and he says there should be nothing that makes the server

I am wondering about a couple of things:
* Is there any specific test among these all, or a combination of those which will make the status go to unsynched? * Why is there for example an 8 seconds gap in the log entries? Usually they are < 1 second. It also looks to me that score and root dispersion (sorry, not sure what that exactly is) are kind of "reset" when that happens.
I do not understand what the chrony.conf has in it. The server should be queried something like once ever minute to once an hour, not once every two seconds. What is your minpoll and maxpoll and why have you set
      them so low?
In our setup, we have 8 computers in a local network without global clock reference. One of the computers is acting as NTP server. This server gets its time once at startup from a GPS device (here it might jump once). Bootup order of the computers is not ensured in any way, clients might be up before server. I tried to use initstepslew and makestep directive, to force a quick adoption of the server time after bootup.

chrony.conf looks like this, minpoll is 5, so no clue why it seems to query every two seconds.

server minpoll 5 maxpoll 7 iburst

keyfile /etc/chrony/chrony.keys
commandkey 1
driftfile /rw/.chrony/chrony.drift
log tracking measurements statistics
logdir /rw/logs/chrony/
maxupdateskew 100.0
initstepslew 10
makestep 100 10
dumpdir /rw/.chrony
allow 10/8
allow 192.168/16
allow 172.16/12
logchange 0.5

 *  Why does the stratum test fail? Can you explain what happens?

The server is suddenly reporting its stratum to be higher than yours. The
problems are in the server, not in your system. Your sysadmin of the server is
fobbing you off without looking at his system.
Thanks for clarifying

Am 28.05.2015 um 22:14 schrieb Bill Unruh:
      On Thu, 28 May 2015, Ulrich Schwesinger wrote:

      Tests 1234 abc 5678 are defined in the ntp specifications.

      From RFC 1305

TEst 1 and 2 test to make sure that the timestamps make sense (eg are not the wsame as an old packet and pairs with the last one sent to that peer) test3 is that the originate and receive timestamps are non-zero, Test 4 is taht the delay (round trip time) be reasonable, abc are subsets of that, 5 is authentication, test 6 irequires peer clock by synchronized and that the itnerval since the peer clock was last updated is positive and less than NTP.MAXAGE, test 7 that the host has no lower stratum than the server, and 8 that the header contains reasonable values for rootdelay and rootdispersion.

Ie, if any of the tests are out the server's time is too suspect to use. Your
      server is problematic.

I marked some lines that look suspicious to me. In in the first line, some test seems to fail [110]. In
            the 2nd line, suddenly
            the offset jumps up to -1.9 seconds.
            For the 3rd line, from the documentation I found this:

Leap status: ? means the remote computer is not currently synchronised. 5678: Tests for maximum delay, maximum delay ratio and maximum delay dev ratio, against defined
            parameters, and a test for
            synchronisation loop (1=pass, 0=fail) [1111]

So that would mean that delay ratio and delay dev ratio is bad...not sure what that really means

It means that the round trip time is out of spec. And it looks like your source, whatever that is, went nuts and jumped by 2 sec suddenly. that is why you should be using at least 3 sources. If one goes mad, the others can outvote it. If you only use 1, then its time is by definition right, even if
      it is out by 40 years.

I find this very hard to debug, is there any other thing I could do to find out what's exactly going

Talk to the person who is sysadmin on the server and find out what happened. Use more than one server.

            Thanks for your answers in advance,


To unsubscribe email chrony-users-request@xxxxxxxxxxxxxxxxxxxx with "unsubscribe" in the subject. For help email chrony-users-request@xxxxxxxxxxxxxxxxxxxx with "help" in the subject.
Trouble?  Email listmaster@xxxxxxxxxxxxxxxxxxxx.

Mail converted by MHonArc 2.6.19+