Amazon S3 service experiences increased error rates

Courtesy Amazon Web Services

This Tuesday, an Amazon storage center in northern Virginia effectively shut down for about four hours. This affected access to many websites across the east coast and, to a lesser extent, the world, including IFTTT, Imgur, SoundCloud, DownDetector, and, perhaps most notably for Stevens students, Canvas. Aside from causing much frustration on the part of the students and faculty, it does not seem that any permanent side effects have been sustained from Canvas’ downtime. The errors first appeared around 9:37 a.m. PST, and what Amazon officially called “increased error rates” continued through 1:57 p.m. PST.

The Amazon S3 service, the service affected by the error, is used by approximately 148,213 websites. In effect, many of these websites were rendered either wholly or partially inaccessible to users on the east coast. Websites are not the only things that use the Amazon S3 service, however. Many connected Internet-of-Things devices, such as smart thermostats, light bulbs, security systems, and even remotes, also experienced issues during the downtime because their systems are built using S3.

This downtime comes on the heels of last week’s CloudBleed bug, an issue with internet security firm CloudFlare’s infrastructure. This bug caused random, potentially sensitive, data to leak to insecure servers for months (If you haven’t already changed all your passwords and enabled two-factor authentication, do so now).

Amazon has offered no explanation as to the root cause of the S3 issues and has thus far offered no public plan to ensure that similar issues don’t happen again. Early Tuesday morning, Amazon’s mobile API also experienced elevated error rates for about five hours.