[ntp:questions] MSF leap

Richard Leach rich at richux.plus.com
Sun Jan 1 18:58:27 UTC 2006


Hi, I'm a Brit listening to our longwave radio clock MSF.
I was monitoring over the leapsecond, and missed timestamps from MSF
just before midnight. This was with Johnathan Buzzard's software which, by
my reading, seems not to choke over the extra second.

Here are my MSF stamps. What did anyone else see?

Sat Dec 31 23:53:00 2005
Sat Dec 31 23:54:00 2005
Sat Dec 31 23:55:00 2005
Sat Dec 31 23:56:00 2005
..... do'h!
Sun Jan  1 00:00:00 2006
Sun Jan  1 00:01:00 2006
Sun Jan  1 00:02:00 2006
Sun Jan  1 00:03:00 2006  

I wonder if this was a ploy or a panic by NPL, or just four bad receptions?

It's perhaps worth repeating that, in spite of this, everything else
behaved impeccably. The kernel announced the leap second, the xclock
second hand paused for a second at midnight, ntp_gettime stopped returning
INS and returned OOP at the leap second, then WAIT until NTP had three new
timestamps(?), then OK.

The kernel clock did not advance to 23:59:60 but stayed at 23:59:59 for
two seconds (output of date and ntp_gettime).

NTP remained stable throughout, maintaining sensible offsets, gradually
ignoring network clocks which had not leaped, and taking notice of one
clock which had jumped early.

Debian Linux stable, kernel 2.6.8, ntp 1:4.2.0a with Serge Bets'
nano-HOWTO applied.




More information about the questions mailing list