On August 27th, 2020, at 1:00 PM UTC we received multiple reports of degraded performance for our Block Storage service in Fremont. Upon investigating, we were able to determine that the health of the cluster was good but our monitoring showed a slight decline in Block Storage throughput. After an exhaustive investigation, we determined that there was a physical link between a Block Storage aggregation switch and a core router that was reporting a small number of errors. These errors were enough to cause intermittent connection issues to our Block Storage servers which translated to degraded performance on customer volumes. At around 2:32 AM UTC, the problematic link was shut down and we were able to see an improvement in performance. We have scheduled a non-intrusive maintenance to repair and test the faulty link before putting it back into production.