Amazon Gets Reboot-Happy
The outage Amazon suffered in April comes to mind immediately, but this is a more controlled situation. The reboots are being scheduled weeks out, with users given advance warning.
What's curious is that there are a lot of reboots. "It's thousands and thousands of instances [virtual machines]. Most of them, most likely," Bias writes in an email to Light Reading.
Amazon isn't giving a specific cause, and a statement sent to the press implies that it's part of the routine upgrades that are always going on. "Some updates require instances to be restarted and we periodically reboot instances in order to apply these updates," a spokeswoman tells Light Reading in an email.
Bias disputes that. "If it were routine, they would have rolled it out for all new instances, waited a month to see what got the upgrade naturally, then scheduled tranches of reboots," he writes.
This is another reminder of the uncertainties of relying on the cloud, but it's got an upside. "I'm not panning AWS on this. I actually think it's great they can roll out a security update to 100,000+ servers," Bias noted on Twitter. (The popular theory is that Amazon is patching a security hole and requires reboots to do it.)
There's a lesson for developers, too. Applications intended for the cloud should be "designed to be rebooted in a rolling fashion," so that they'll survive a mass reboot in the case of emergency, Bias wrote.
— Craig Matsumoto, West Coast Editor, Light Reading