[time-nuts] Fake leap second ntp servers
zenczykowski at gmail.com
Thu Jan 31 03:10:23 EST 2008
I just joined your email list to inquire about something which some of you
may have some knowledge of...
I'm developing some code (for linux with an ntp server/client running), for
which I would like to verify that it handles positive (and negative)
leap seconds correctly (I need good time stamps for cross-machine
Obviously this is hard to do in practice due to the extreme rarity of these
(even more so, considering my home servers running ntp apparently for some
reason saw leap seconds at the end of 2007...)
I've thought about switching to djb's suggestion of keeping a non-posix time
zone on the machines in question, but this
is not feasible because of existing legacy code...
Thus I am curious to know if someone somewhere has perhaps set up fake ntp
servers which regularly insert and delete
leap seconds (ie. reset their time every couple days and randomly simulate a
positive or negative leap second 2-3 days
down the line).
If not, does anyone know how to go about setting up such a fake ntp server
for testing purposes?
Also, is there a known good list of ntp servers which are known to provide
correct leap second announcements (as mentioned above,
the stratum 1s at least 2 of my machines were synched to apparently
broadcast a leap second at the end of 2008).
I've heard horror stories about time synchronization going haywire the last
time there was a leap second and some upstream
ntp servers announced leap-seconds and others didn't (the time
mis-synchronization brought down the application)...
More information about the time-nuts