[ntp:questions] Re: [VMS] config or clock?

Richard B. Gilbert rgilbert88 at comcast.net
Sun Feb 1 05:58:09 UTC 2004


I'm afraid your local clock is beyond any help that NTP can provide!  
500ppm is the maximum that NTP can cope with and you are there or beyond.

It may be possible to fix the clock by patching the O/S to give the 
clock a permanent "slew" that will cause it to tick every 10ms within, 
say, 100ppm..  I'd suggest inquiring on comp.os.vms if this is 
feasible.  I know that in past years, there were programs that would 
allow you to make the change between standard and daylight time by 
running the clock faster in the spring and slower in the fall  for a 
period of an hour or two.  This was for people who, for one reason or 
another, could not simply jump the clock forward or backward by an 
hour.  With more modern versions of VMS far fewer people have need of 
this software; since VMS V7.0 (I think!) VMS has kept UTC internally and 
simply displayed local time as required.

It may also be possible to fix the hardware to bring the 10ms tick 
within spec.   I'm running  NTP on six Alphas  with VMS V6.2-1H3 with 
UCX 4.2 ECO 5 or VMS V7.2-1 with TCPIP V5.1 ECO 4 and not seeing a 
problem with any of them.  Depending on the quality of the time supplied 
by servers, they all manage to synch within 10ms and usually less than that.

Alder wrote:

> I've just started paying some attention to the system time on my 
> AlphaServer 1000A 4/266.  It's running VMS 7.3 and TCP/IP Services 5.4.
>
> After visiting www.ntp.org and perusing the docs, I tested the 
> stability and accuracy of my system clock by stopping the NTP service 
> and setting my clock with ntpdate against the 'closest' time server to 
> me.  I found this:
>
> $ ntpdate "-o" 4 "-p" 4 ntp.cpsc.ucalgary.ca
> Selected server is ntp.cpsc.ucalgary.ca (136.159.2.3) offset +0.4557905
>
> ... 8 hours later:
>
> $ ntpdate "-o" 4 "-p" 4 ntp.cpsc.ucalgary.ca
> Selected server is ntp.cpsc.ucalgary.ca (136.159.2.3) offset +34.9372245
>
> That looked to me like my clock was losing time rapidly.  I then 
> started the NTP service and monitored the drift and log files for 
> awhile.  THe drift file moved from a value of 0.0 at service startup 
> to 500.0 within an hour or two.
>
> My log file over this period showed an ever increasing 'slew' value as 
> NTP tried correcting my clock:
>
> $ search tcpip$ntp_run.log "time slew"
> 31 Jan 09:44:58 ntp[216]: time slew 0.390476 s
> 31 Jan 09:58:53 ntp[216]: time slew 1.085532 s
> 31 Jan 10:18:18 ntp[216]: time slew 1.682264 s
> 31 Jan 10:36:10 ntp[216]: time slew 2.123613 s
> 31 Jan 10:52:20 ntp[216]: time slew 2.604056 s
> 31 Jan 11:07:54 ntp[216]: time slew 3.144392 s
> 31 Jan 11:31:09 ntp[216]: time slew 3.654378 s
> . . .
> 31 Jan 15:48:20 ntp[216]: time slew 14.203758 s
> $
>
> My config file is pretty basic, so I'm wondering if the 'symptoms' I 
> reported above are a function of my clock, or are they the result of 
> bad configuration?
>
> $ type tcpip$ntp.conf
> driftfile SYS$SPECIFIC:[TCPIP$NTP]TCPIP$NTP.DRIFT
> server time.nrc.ca version 4
> server ntp1.cmc.ec.gc.ca version 4
> server tick.utoronto.ca version 3
> server time.chu.nrc.ca version 3
> server ntp.cpsc.ucalgary.ca version 4
> server timelord.uregina.ca version 3
> statistics loopstats
> statsdir SYS$SPECIFIC:[TCPIP$NTP.STATS]
> $
>
> Any help most appreciated...!
>
> Terry
>




More information about the questions mailing list