Once upon a time, Marco Marongiu <[email protected]> said: >As far as I remember, in 2006 we found out that the system clock stepped >back one second, the database engine got mad, and crashed. Back then we >were using NTPv3 because xntpd on some Solaris host didn't know about v4.
Another problem I saw in 2009 was with older versions of Linux kernels (such as that found in Red Hat Enterprise Linux 4). The kernel handled inserting the leap second okay, but when it tried to print an informative message about it, the system would deadlock if it was under load. That particular bug was easy to reproduce on a test system. I had a script that set the time to a few seconds before midnight UTC, set the leap flag with adjtimex(), count until a couple of seconds after midnight, and loop. It would be fine as long as the system was idle, but as soon as it was busy, the kernel deadlocked. A similar setup might be used to test other problem software. -- Chris Adams <[email protected]> Systems and Network Administrator - HiWAAY Internet Services I don't speak for anybody but myself - that's enough trouble. _______________________________________________ questions mailing list [email protected] http://lists.ntp.org/listinfo/questions
