Twitter was unavailable for users worldwide on Tuesday morning, with the site apparently suffering a total outage followed by serious access problems lasting over an hour.
Access to the service began failing over the web, mobile and its API (application programming interface, the system that applications use to speak to the Twitter service) at 8:20am GMT, with error messages warning the network is both “over capacity” and suffering an “internal error”. By 10:00am, the majority of the service had returned to some semblance of normality, with the company’s image handling service and home timelines still suffering.
Twitter’s own status board updated at 9:00, confirming the outage, and the company’s developer-facing monitoring confirmed that four of the five public APIs were down, suffering a “service disruption”. At 8:47, the search API was upgraded to “performance issues”. By 8:55am, a second API was upgraded to “performance issues”, and some users were able to sporadically access the service. But the site’s status yo-yoed, and several services were still reported as down more than two hours after the outage began.
The company initially confirmed the outage by, somehow, tweeting, from its[email protected] account. We were unable to see the tweet, because Twitter was down. Twitter emailed the text of the tweet to the Guardian, which read: “Some users are currently experiencing problems accessing Twitter. We are aware of the issue and are working towards a resolution.”
In the early days of the service, Twitter outages were common enough that the company’s “over capacity” error message gained a nickname: the fail whale.
The service’s architecture prevented the company from easily expanding capacity by simply adding servers to its back end, and so it would frequently collapse under the weight of its users during major events. And “major” is relative: in 2008, the company buckled under the strain of an Apple keynote.