Twitter was down for three-quarters of an hour this morning (meaning dedicated Twitter users had nowhere to complain about Twitter being unavailable). Twitter’s explanation for what happened is a reminder that no deployment is immune from unexpected problems — and also a nifty example of how to understate those issues.
Here’s how Twitter explained the issue on its status page
During a planned deploy in one of our core services, we experienced unexpected complications that made Twitter unavailable for many users starting at 11:01am. We rolled back the change as soon as we identified the issue and began a controlled recovery to ensure stability of other parts of the service. The site was fully recovered by 11:47am PDT.
There’s lots of understatement and reassurance here. We’re told it was a “planned deploy” immediately, lest we worry that Twitter has no idea what is going on. (We’d suggest “planned deployment” as more accurate, by the way.)
“For many users” ignores what seemed to be the reality: Twitter was entirely unavailable. “As soon as we identified the issue” doesn’t tell us whether that took two minutes of the 46-minute outage, or most of the time.
A 46-minute outage still means Twitter won’t have more than 99.89 per cent uptime this month. As we’ve pointed out before, the length of outage often matters more than the percentage.