[evla-sw-discuss] NTP, EVLA MIB's and phase jumps

[Barry Clark]

I don't know - I looked at times first thing this morning.
Ken has nice software for comparing clocks at MIBs, but I
don't.  I cobbled together something in Python that said
MIB clocks were the same as Mchost and Mctest to within
about 20 ms or so.  (Ken's C program is much better.)
I used my VLBA software to find that Mchost, Mctest, and my
machine in Socorro were all within about six or eight ms of
PieTown.

Yes, the symptoms could be explained by a MIB picking the
wrong 52ms tick, but times looked too good for that to be
happening, unless it has been variable.

James Robnett wrote:
> Short answer, we should retry timing tests after 11 or so.
> 
> james
> -------------------------------------
>    The symptoms reported this morning involving phase jumps
> point squarely at the multicast NTP time signals from amos
> and andy.
> 
>   Not only are the symptoms consistent with the phase jumps
> the evidence that it started on Wednesday points directly
> at the OS upgrades on amos and andy.
> 
>   I looked at NTP on both amos and andy and there is what
> appears to be a problem.
> 
>   Both servers point at the stratum-1 GPS clock at the site
> and a stratum-1 system in Green Bank.  They also point
> at each other as peers to try and limit variance when one
> or the other is rebooted.
> 
>   After the upgrade they also pointed at a 3rd peer running
> the same hardware/OS to further limit jitter on reboots.
> 
>   Sadly that systems clock appears to be complete crap.
> It seems to be 100ms off the others and is pulling amos/andy
> backwards.  I've removed references to that system and
> restarted NTP.
> 
>   In about an hour they should be settled enough to retry
> timing tests on the EVLA.
> 
>   On a side note,  was the behavior seen consistent with some
> boards syncing to the trailing 19.2 peak or can you only see
> number of peaks off and not direction ?
> 
> james
> 
> _______________________________________________
> evla-sw-discuss mailing list
> evla-sw-discuss at listmgr.cv.nrao.edu
> http://listmgr.cv.nrao.edu/mailman/listinfo/evla-sw-discuss