Two nights ago, a leap second was added to system clocks running Linux, causing much-undesired havoc.
On July 1st at 12:00 AM UTC, both of my Amazon EC2 instances fired an alarm indicating high CPU usage. I investigated to find that it was MySQL that was eating all of the CPU. I logged in and ran SHOW PROCESSLIST to find that no queries were running (these servers don't get hit much after business hours). I stopped MySQL, CPU utilization dropped back down to 1-3% (as normal). I restarted MySQL, and it started eating a lot of CPU again. Then, I restarted the server (shutdown -r now), and the problem went away. Both servers had the exact same problem (running Ubuntu 12.04 LTS).
In my particular case, MySQL began eating CPU, even after being restarted. It was a livelock.
The only relevant item I saw in the syslog was:
Jun 30 23:59:59 hostname kernel: [14152976.187987] Clock: inserting leap second 23:59:60 UTC
Oh yeah... leap seconds. Those are super important... so important that we need our servers to freeze up.
http://it.slashdot.org/story/12/07/01/1920217/leap-second-bug-causes-crashes
https://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commit;h=6b43ae8a619d17c4935c3320d2ef9e92bdeed05d
http://stackoverflow.com/questions/11294582/mysql-eating-80-100-cpu-starting-july-1-midnight-a-day-or-so-following-a-ebs-c
https://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commit;h=6b43ae8a619d17c4935c3320d2ef9e92bdeed05d
http://stackoverflow.com/questions/11294582/mysql-eating-80-100-cpu-starting-july-1-midnight-a-day-or-so-following-a-ebs-c
Comments