According to Instructure there was an outage that affected most Canvas customers starting at 6:03 PM and lasting approximately 25 minutes. Canvas experienced a drive failure with a database that provides cluster-wide meta-data and that was further complicated by its interaction with additional database components. Instructure was able to briefly restore Canvas within 10 minutes but found performance was still in question and performed a restart of the affected system within 20 minutes. Within 5 minutes after the restart the system had stabilized and proper operation was restored. No clients at Western reported any abnormalities during this period.
Instructure considers the failure to be the result of a bug and is working to remove dependance on cluster-wide databases over the next two months. The goal of this move is to insure that any outages that do occur will impact only a small subset of customers rather than effect users system-wide.