7 comments

[ 3.4 ms ] story [ 25.3 ms ] thread
Talking about a bad week. I guess their availability percentages are gonna look quite bad (when and if they switch from the May ones).
It's somewhat heartening to see that this report doesn't talk about the control rods for their dyno manifolds creating unintended flux in the tachyon emitter.

It's not great, but at least it's easy to understand what went wrong without finding a marketing/english translator.

ha! Good point on their "naming". Who knows what those things actually do.
This is one of the worst parts of running database on EC2, IO can basically go to crap and cause drives to eat themselves. Databases don't like this much.

If you use a database on top of EC2, make sure it's replicated. This saves you a ton of headache. We (MongoHQ) spent all night getting single server databases back up and functional (though we didn't have to restore any from backups), while our replicated plans just hummed along throughout the outage.

Ouch, Heroku seems to be going through a rough patch.