elsen@imec.be (Marc Elsen) (06/06/91)
Recently we switched from timed to ntpd on our DECstations : we have 21 DECstations (3100) booting from a DECsystem 5400 (server). Since we have no Internet connection I set up the 5400 as being a "reference clock". The stations are all ntp clients and the 5400 is always configured as the (only time server). When setting up ntpd on the 5400 and on the clients I did nothing special , I followed the instructions in the manual exactly (ntpd(8)). The following has happened 2 times since : on 2 DIFFERENT decstation the clock seemingly halted !! That is subsequent date commands always showed the same frozen time !!! The clock could be restarted by setting it to the current time. Is this due to a bug in ntpd. You can imagine that this problem is very annoying for end users. Marc (elsen@imec.be)
jah@margo.ots.utexas.edu (Jeff Hayward) (06/07/91)
In article <980@imec.UUCP> elsen@imec.be (Marc Elsen) writes: > [...] > The stations are all ntp clients and the 5400 is always configured > as the (only time server). > [...] > The following has happened 2 times since : on 2 DIFFERENT decstation > the clock seemingly halted !! > That is subsequent date commands always showed the same frozen time !!! I've seen this behavior many times on my DS3100's running ultrix V4.0 and the DEC-supplied ntpd. I went to xntp, same results. If anyone knows more about this I'd appreciate hearing about it. -- Jeff Hayward The University of Texas System +1 512 471 2444 Office of Telecommunication Services jeff@nic.the.net
harry@atmos.washington.edu (Harry Edmon) (06/07/91)
I have been told by some people on campus that this is due to a race condition between ntp and Ultrix, and that it is fixed in 4.2. However, I have not verified this. -- Harry Edmon INTERNET: harry@atmos.washington.edu (206) 543-0547 UUCP: uw-beaver!atmos.washington.edu!harry Dept of Atmospheric Sciences, AK-40 University of Washington
trost@reed.edu (06/08/91)
In article <50136@ut-emx.uucp>, jah@margo.ots.utexas.edu (Jeff Hayward) writes: In article <980@imec.UUCP> elsen@imec.be (Marc Elsen) writes: > The following has happened 2 times since : on 2 DIFFERENT decstation > the clock seemingly halted !! > That is subsequent date commands always showed the same frozen time !!! I've seen this behavior many times on my DS3100's running ultrix V4.0 and the DEC-supplied ntpd. I went to xntp, same results. If anyone knows more about this I'd appreciate hearing about it. Could this have anything to do with the mysterious clock-stopping problem people were reporting for their DECstations? (This is a little disturbing, since our central campus server is a DS3100 as well...) In any event, I suspect that a suitable workaround for the moment is to add the line /usr/etc/ntp -s -f ...hosts named in /etc/ntp.conf... to rc.local before ntpd gets started.
mogul@pa.dec.com (Jeffrey Mogul) (06/11/91)
In article <980@imec.UUCP> elsen@imec.be (Marc Elsen) writes: > > Recently we switched from timed to ntpd on our DECstations : > > The following has happened 2 times since : on 2 DIFFERENT decstation > the clock seemingly halted !! > That is subsequent date commands always showed the same frozen time !!! > > The clock could be restarted by setting it to the current time. > > Is this due to a bug in ntpd. > You can imagine that this problem is very annoying for end users. This is a bug in the Ultrix 4.1 kernel (and perhaps 4.0); I believe it only affects MIPS-based systems (not Vaxen). Anyway, it's fixed in Ultrix 4.2. I don't know if there's a patch available for Ultrix 4.1, but if there is one it would be a new kern_clock.o module. -Jeff