any reason why NTP client would receive -10secdiff? - NTP

This is a discussion on any reason why NTP client would receive -10secdiff? - NTP ; Hi all, I'm a newbie timekeeper and I tried searching the lists and ntp docs for an answer to this problem but haven't found a clue so far. This morning a client phoned me telling me that their server was ...

+ Reply to Thread
Results 1 to 3 of 3

Thread: any reason why NTP client would receive -10secdiff?

  1. any reason why NTP client would receive -10secdiff?

    Hi all,
    I'm a newbie timekeeper and I tried searching the lists and ntp docs for an answer to this problem but haven't found a clue so far.

    This morning a client phoned me telling me that their server was being told it's clock was in advance of our ntp time by +10sec or else was being asked to step it's time backwards 10sec (i'm not exactly sure what the manifestation of the problem was because I was only given a rough verbal description of the trouble). In any case this is unheard of in my limited time as an NTP admin!! Naturally their ntpd wasn't going to accept a jump by -10sec, so subsequently their server started polled our tier-one server every second for a while before giving up, after which their time had drifted out of sync by about a minute. They run a Windows network and it was critical that their time was in good sync (I cannot divulge the client for confidentiality reasons, but believe me it is a critical problem). They've started using another tier two server and apparently have had no further problems so far. I can't see any obvious problems with our servers, but maybe I'm missing something?
    Their software using NTP was something called "Netware", which I'm not familiar with, so I was unable to help them. One potential cause of problems at our end was a lightening strike at our facility on Friday 7th Nov, which affected other systems in the same building but not, as far as I know, the clocks or NTP servers. The client started reporting this problem on that Friday evening, but that could be just a coincidence.

    So my question is why might this client problem occur? Any suggestions would be appreciated. I'm a bit in the dark because they won't give me access to their system logs and they are instead suggesting there is a problem with our servers.

    - Blair Smith

  2. Re: any reason why NTP client would receive -10sec diff?

    b.smith@irl.cri.nz (Blair Smith) writes:

    >Hi all,
    >I'm a newbie timekeeper and I tried searching the lists and ntp docs for an answer to this problem but haven't found a clue so far.


    >This morning a client phoned me telling me that their server was being told it's clock was in advance of our ntp time by +10sec or else was being asked to step it's time backwards 10sec (i'm not exactly sure what the manifestation of the problem was because I was only given a rough verbal description of the trouble). In any case this is unheard of in my limited time as an NTP admin!! Naturally their ntpd wasn't going to accept a jump by -10sec, so subsequently their server started polled our tier-one server every second for a while before giving up, after which their time had drifted out of sync by about a minute. They run a Windows network and it was critical that their time was in good sync (I cannot divulge the client for confidentiality reasons, but believe me it is a critical problem). They've started using another tier two server and apparently have had no further problems so far. I can't see any obvious problems with our servers, but maybe I'm missing something?
    > Their software using NTP was something called "Netware", which I'm not familiar with, so I was unable to help them. One potential cause of problems at our end was a lightening strike at our facility on Friday 7th Nov, which affected other systems in the same building but not, as far as I know, the clocks or NTP servers. The client started reporting this problem on that Friday evening, but that could be just a coincidence.


    >So my question is why might this client problem occur? Any suggestions would be appreciated. I'm a bit in the dark because they won't give me access to their system logs and they are instead suggesting there is a problem with our servers.


    What do your system logs say? Why not persuade them to run the reference
    version of ntp, rather than some partially bastardized version.
    The fact that it started polling you at 1 sec intervals shows it is a very
    poor implimentation.

    It is possible that your lightning strike screwed up your serverBut you
    shoudl be able to see that from the logs on your system (peerstats,
    loopstats...)

    Who is your ntp server?



    >- Blair Smith


  3. Re: any reason why NTP client would receive -10sec diff?

    Blair Smith wrote:
    >


    + limited time as an NTP admin!! Naturally their ntpd wasn't going to
    + accept a jump by -10sec, so subsequently their server started polled our

    A normal, out of the box, ntpd would step, if all sources agreed on the
    error. However, it is most unlikely that this is ntpd.

    + tier-one server every second for a while before giving up, after which

    A real ntpd cannot be configured to poll that fast, and such a fast
    polling rate would be considered abusive. Do you,by any chance, have
    kiss of death enabled on your server, as that polling rate would trigger
    it. I believe one of the tactics it uses to try and dissuade abusive
    clients is to report a false time to them. A non-standard (or
    non-recent) implementation might not recognize the other signals that it
    had received the kiss.

    > Their software using NTP was something called "Netware", which I'm

    + not familiar with, so I was unable to help them. One potential cause of

    I didn't realise people still used this. I thought Microsoft CIFS/SMB
    networking now completely dominated the market. When I last encountered
    it, Netware was a file server that ran as a DOS extender; it used to be
    the dominant corporate file server software in the days of MS-DOS, and
    the early days of Windows. Although clients can time synchronize to
    this, it is only the server that I would imagine would try to use NTP.

    + problems at our end was a lightening strike at our facility on Friday
    + 7th Nov, which affected other systems in the same building but not, as
    + far as I know, the clocks or NTP servers. The client started reporting
    + this problem on that Friday evening, but that could be just a coincidence.
    >


+ Reply to Thread