Choice of time servers - NTP

This is a discussion on Choice of time servers - NTP ; I have two flavors of Stratum 2 servers -- some older Sun servers that are basically out of capacity doing double duty as DNS cache servers, and a set of new Compaq RH Linux servers. The Sun servers are running ...

+ Reply to Thread
Results 1 to 3 of 3

Thread: Choice of time servers

  1. Choice of time servers

    I have two flavors of Stratum 2 servers -- some older Sun servers that are
    basically out of capacity doing double duty as DNS cache servers, and a set
    of new Compaq RH Linux servers.

    The Sun servers are running +/- 0.3 ms offsets to the Stratum 1's with
    default minpoll/maxpoll and other default settings.

    The RH Linux servers are running +/- 1 ms, only after tweeking minpoll and
    maxpoll to 6 (64sec) and using burst. I had one that wouldn't settle down
    even at 4 (16 sec) until we replaced it. There is no load at all on these
    servers.

    Is this an expected result?
    _______________________________________________
    questions mailing list
    questions@lists.ntp.isc.org
    https://lists.ntp.isc.org/mailman/listinfo/questions


  2. Re: Choice of time servers

    Have you seen http://ntp.isc.org/Support/TroubleshootingNTP, especially the
    links to known HW and OS problems?

    H

  3. Re: Choice of time servers

    Bill Myers wrote:

    > I have two flavors of Stratum 2 servers -- some older Sun servers that are
    > basically out of capacity doing double duty as DNS cache servers, and a set
    > of new Compaq RH Linux servers.
    >
    > The Sun servers are running +/- 0.3 ms offsets to the Stratum 1's with
    > default minpoll/maxpoll and other default settings.
    >
    > The RH Linux servers are running +/- 1 ms, only after tweeking minpoll and
    > maxpoll to 6 (64sec) and using burst. I had one that wouldn't settle down
    > even at 4 (16 sec) until we replaced it. There is no load at all on these
    > servers.
    >
    > Is this an expected result?
    > _______________________________________________
    > questions mailing list
    > questions@lists.ntp.isc.org
    > https://lists.ntp.isc.org/mailman/listinfo/questions
    >


    Check your Red Hat systems for the value of the kernel parameter "HZ".
    If it's 100 it's probably not the problem. If it's 250 or 1000 try
    setting it to 100.

    Using "burst" is a very unfriendly thing to do. It places eight times
    the normal load on a server! It was a special purpose hack, never
    intended for general use; unfortunately this fact was never well documented.

+ Reply to Thread