Leap Second Disrupts Internet Sites

We told you about the one-second adjustment to the world's clocks that brought our measurements in line with the actual rotation of the earth. That was last night, about 8PM EDT. And although it has happened before, and there were warnings, not every website was ready for a temporal disruption that caused older Linux-based systems to crash.

The Pirate Bay was affected.

So was Twitter.

Other sites that went down were Reddit, Gawker media sites, StumbleUpon, Yelp, FourSquare, LinkedIn, and Meetup.

Strangely, Fark was down, but it was a planned outage to move equipment. It just coincided with the leap second crash.

Most of the websites that were affected are back up, or will be soon.

(Image credit: Flickr user Alan Cleaver)

Newest 4
Newest 4 Comments

I strongly doubt the leap second had anything to do with the proliferation of dead sites -- and certainly not Linux. Afterall, Linux, like ever other Unix-derived system, just keeps track of the seconds since midnight, January 1, 1970, UTC, to keep track of time: this is the usual Unix epoch. (Well, Linux keeps track of nanoseconds since the Epoch, but the NTP time daemon that is in charge of keeping coordinated network time is really only suitable for millisecond time resolution -- so the nanoseconds are extremely skewed from machine to machine and only make sense on individual computers. Any finer accuracy than that would require your own atomic clock or using an external radio clock, but that's a digression for another day.)

Perhaps some idiot application developers assumed seconds run from 0 to 59 and did not plan for second 60 -- (much like Y2K; but leap seconds happen all the time, more frequently than leap years) -- but every Unix API that manipulates time is prepared for the 61st second in a minute. (Afterall, it is clear that many of the developers of the affected sites are poor developers -- remember the six-million unsalted and un-iterated password hashes stolen from LinkedIn? Salted and iterated password hashes were already in use on Unix systems in the late seventies, showing that the LinkedIn geniuses were three decades behind everyone else. But that too is a digression for another day.) Reasonable applications would do the same and just keep track of time using the Unix epoch.

No, the more likely cause is probably over-reliance on a single datacenter that did not survive a storm: http://m.wired.com/wiredenterprise/2012/06/real-clouds-crush-amazon/
Abusive comment hidden. (Show it anyway.)
So Mr Coward are you saying that the Pirate Bay, Twitter,
Reddit, Gawker, StumbleUpon, Yelp, FourSquare, LinkedIn, and Meetup all located all their servers in a single data centre that went down at midnight as the result of a storm? Or are you saying that all of these companies a single datacentre each in completely different physical locations which were all hit by storms bad enough to take them down at exactly the same time? Whatever you're saying it doesn't make sense.

Actually TBP went down three hours before midnight, so it wasn't caused by the leap second. It was actually caused by a rack being moved.

Reddit attributed their outage to Casandra not coping with the leap second. Mozilla confirmed that Hadoop had problems with the leap second. It's been confirmed that debian.squeeze servers all over the place died as a result of the leap second, the workaround apparently was to disable ntpd. So it looks like the stories are true.

Why is it that whenever a problem arises with anything to do with Linux there's always some sad fanboi desperately trying to blame something, anything else?
Abusive comment hidden. (Show it anyway.)
Login to comment.




Email This Post to a Friend
"Leap Second Disrupts Internet Sites"

Separate multiple emails with a comma. Limit 5.

 

Success! Your email has been sent!

close window
X

This website uses cookies.

This website uses cookies to improve user experience. By using this website you consent to all cookies in accordance with our Privacy Policy.

I agree
 
Learn More