Cloud Fail

CloudFront edge locations in Europe experienced increased latencies from 4:10 PM PST to 4:20 PM PST. Latencies have returned to normal.

We are investigating increased latencies in Europe.

DNS resolution has recovered. During this issue the service was operating normally.

Some customers on the US West coast may be experiencing problems resolving DNS for the CloudFront control API and console. The CloudFront delivery service continues to operate normally.

Calls to the CloudFront Control APIs have returned to normal. At no time were end-user requests impacted.

We’re investigating elevated error rates to our control API. End-user requests for content from our edge locations are not affected by this issue and are being served normally.

We have identified an issue that occurred when customers made IPv6 DNS requests. We addressed this at 5:54 AM PDT and requests are now being served normally.

We’d like to give you a little more information about the issues that we experienced this morning with the Amazon CloudFront Control APIs. During three separate, but related, periods of time, some requests to the CloudFront Control APIs resulted in failed requests. These periods were from approximately 1:25am PDT – 2:15am PDT, 3:25 am PDT – 4:40am PDT and 8:05am PDT – 9:05am PDT. The failed requests during each of these periods were caused by a recurring issue related to connecting to our control API, which was difficult to detect. We recovered in between these events that led us to prematurely believe the service was operating normally. We now have identified the root cause, and we have implemented a fix to address the issue. At no time this morning were end-user requests for content from our edge locations affected by this issue; content was being delivered normally.

Error rates have recovered. Requests to the Amazon CloudFront control API are working normally. Throughout this issue, requests for objects from Amazon CloudFront edge locations were not impacted.

Error rates are recovering. We continue to monitor the service. Requests for objects served from edge locations were not affected.