We experienced a partial API outage between 09:35 and 12:05 UTC on November 25th.
There was a sudden surge in spam traffic, on top of spam volumes that were already unusually high. One of our data centers began dropping some API calls, and responding slowly to others. Our systems team routed all traffic to another data center, and after a few minutes the load was stabilised. We estimate availability was 50-60% for the first hour, then 70% and up.
Availability has been back to normal since 12:05 UTC, November 25th.
Clients with a re-try feature – including all recent versions of the Akismet plugin for WordPress – will have automatically re-checked any comments that were missed due to the disruption.
Our systems team has identified the components that failed under load, and we’re working on an improved architecture that will allow a quicker response to unusually heavy loads.
Spam levels remain unusually high. The API is handling it comfortably now.