All services are online

On 5 November, a routine content update caused our system to refresh cached data for every client account at the same time. This filled the memory of our cache cluster. Early on 6 November a scheduled snapshot of the cache needed more memory than the cluster had left. One part of the cluster then went into recovery, which took about 20 minutes. Calculation requests failed while it recovered.

Maintenance ended

During the routing change on our edge provider our API traffic got moved into a different location which caused a brief outage on our end. The problem is fully resolved now and we're changing the update procedure for this part of the system in order to prevent such accidents in future.