A core piece of infrastructure on our Cloud platform, the database, is being updated. Our team has been performing many tests on the new version over the last few weeks, all of which have been successful.
Yesterday, at 09:00 BST, our team deployed a new read-only slave to a small subset of customers. This new slave began to fail at around 13:40 BST, causing a number of HTTP endpoints to fail. Our team removed the failing slave from the cluster at 14:01 BST to restore normal service. We are still investigating the nature of the failure on the slave to ensure it cannot happen again in the same way.