Amazon Web Services’ US-EAST-1 region, which last week caused massive disruption to online services, is having another bad day as internal dependencies again prove problematic.

At 3:36 PM PDT on October 28 (10:36PM UTC), the cloud colossus advised customers that “Earlier today some EC2 launches within the use1-az2 Availability Zone (AZ) experienced increased latencies for EC2 instance launches.”

Amazon throttled some requests for EC2 resources, but said retrying a request should resolve the issue.

Another impact of the incident created “task launch failure rates for [Elastic Container Service] ECS tasks for both EC2 and Fargate for a subset of customers in the US-EAST-1 Region.”

Amazon’s status page advises that ECS operates cells in the US-EAST-1 Region “and a small number of these cells are currently experiencing elevated error rates launching new tasks and existing tasks may stop unexpectedly.” The cloudy concern also warned that customers “may also see their container instances disconnect from ECS which can cause tasks to stop in some circumstances” but advised it had identified the problem and was working to fix it.

The incident also impacted EMR Serverless services – the elastic map reduce service Amazon offers to run big data tools like Hadoop and Spark.

Here we go again

At 5:31 PM PDT AWS updated its advice to reveal that “EMR Serverless maintains a warm pool of ECS clusters to support customer requests, and some of these clusters are operating in the impacted ECS cells.”

Amazon said it was “actively working on refreshing these warm pools with healthy clusters” and that it had made progress “on recovering impacted ECS cells, but progress is not visible externally.”

“ECS has stopped new launches and tasks on the affected clusters. Some services (such as Glue) are observing recovery for error rates, but may still be experiencing increased latency,” AWS’s status page advises, before stating a “current best estimate of an ETA is 2-3 hours away.”

AWS has not posted any info about the cause of the incident, but whatever caused it is bad news, as the reason for last week’s incident was that many AWS services relied on the operation of another – the DynamoDB database.

In this incident, the problems with EMR serverless are related to issues with ECS – again showing that internal dependencies make the Amazonian cloud fragile.

AWS lists ten services impacted by this incident – App Runner, Batch, CodeBuild, Fargate, Glue, EMR Serverless, EC2, ECS and the Elastic Kubernetes Service – but at the time of writing The Register isn’t seeing reports of service disruptions. That may be because US-EAST-1 is home to six availability zones, meaning plenty of AWS resources remain available if customers have chosen to use them. This incident is also not a full outage, meaning AWS may well have spare resources in the impacted availability zone that customers can use instead of the broken or throttled bits. ®