Image processingDelayed response for 2% of processing requests
At 3.43pm UTC on 1 July 2019, a hardware failure on a processing server combined with heavy processing load caused some new image processing requests to take longer than usual to process. This degraded performance affected approximately 2% of new image processing requests over a 3 hour period (taking more than 10 seconds to process). The faulty hardware was replaced by 5.51pm UTC and average image processing time had been reduced back to normal (150ms) by 6.59pm UTC. All other services continued as normal throughout the incident. The slow requests made up less than 0.2% of requests during the process - the other 99.8% of requests were returned rapidly as normal.