lmjm@doc.ic.ac.uk (Lee McLoughlin) (10/11/90)
I am trying to setup xntp locally. As we are in the UK we are far from any usable internet clock and little chance of getting a decent radio clock to link to one of the GMT servers over here. So I decided to, as an initial test, to run xntp between three machines: Processor OS Name --------- -- ---- Sparc 4/380 Sunos 4.1 (stork) Sparc 1+ Sunos 4.1 (crane) Gould Powernode 9080 UTX/32 2.1A (gould) The gould has a useless clock so I just run ntpdate on it every ten minutes querying stork and crane. The system clock on stork is the master clock. Stork has a config file of: server 127.127.1.1 driftfile /var/adm/ntp.drift Crane has a config file of: server 129.31.103.3 # stork peer 129.31.103.1 # crane driftfile /var/adm/ntp.drift I currently have two problems. Firstly xntpd doesn't seem to issue time till after running for about 8.5 minutes. According to xntpdc for the first 8.5 mins after startup neither stork nor crane has selected peers. Is this what it should be doing?? I've tried reading the NTP (Version 2) document but it is a bit heavy going and I've not found anything which suggests this is the way it should work. My second problem is that the once both crane and stork's xntpd's are talking with each other the times do not seem to stay as much in sync as I had hoped. Ignoring the gould whose time bounces about all over the place. I expected crane's clock to stay within 20-30 ms of stork. However when the crane gets busy (like during a nightly dump or when the students are about) it can be as far as 3-4 seconds out (according to timedc'd clockdiff command). I couldn't find a revision number in the xntp sources but the compressed tar file is identical to that held at toronto.edu (491419 bytes). If some kindly guru could suggest what might be causing the above problems I'd really appreciate it. Lee -- -- Lee McLoughlin phone: 071 589 5111 X 5037 fax: 071 581 8024 Department of Computing, Imperial College, 180 Queens Gate, London SW7 2BZ, UK Janet: lmjm@uk.ac.ic.doc Uucp: lmjm@icdoc.UUCP (or ..!ukc!icdoc!lmjm) DARPA: lmjm@doc.ic.ac.uk (or lmjm%uk.ac.ic.doc@nsfnet-relay.ac.uk)
J.Crowcroft@ucl-cs.UUCP (10/19/90)
From: Jon Crowcroft <J.Crowcroft@uk.ac.ucl.cs> Lee, if you can stand the memory loss, i can give you NTP on ROS (using ISODE - runs on a sun, takes about 1Mbyte), and you could peer over JANET with one of our machines (we then also talk NTP to US, brunel, nottingham and cambridge) - or if you get someone to buy some ethernet-remote bridges, you could come over the 2Mbps livenet digital path to us, and thence direct to the US with no interference from the JNT... jon