AWS recently added a new Availability Zone, eu-west-2d, to the London region. The change was not announced until after it was already available, and I've heard reports that some customers experienced problems afterward. Has anyone seen deployment failures, networking issues, resource recreation, or service-impacting changes related to the new Availability Zone?
4 Answers
The new zone may not have supported every service immediately. In particular, missing or delayed support for features such as PrivateLink and Transit Gateway could cause problems for customers whose automation assumes those services are available in every zone.
Adding another zone can expose assumptions in existing infrastructure, but it also gives AWS more capacity. Customers should avoid hard-coding a fixed number of zones, verify which services are supported in each zone, and test subnet and Terraform logic whenever the regional topology changes.
The London region still being the only UK region is a separate resilience concern. For customers that must keep data in the UK, there isn’t another UK region to use for full regional redundancy.
A common failure mode is code that assumes there are exactly three Availability Zones. Once a fourth zone appears, scripts, Terraform modules, or deployment logic that iterate over the available zones can behave differently and may recreate or redistribute resources. It’s also possible that older subnet automation only created subnets in the original zones.
This can happen when an old CloudFormation stack or provisioning script has been left untouched for years and quietly controls the subnet layout.
Some organizations had to update Terraform configurations because deployments started failing midway. There were also reports of an outage affecting a major UK financial services organization, although that may not have been caused solely by the new zone. AWS posted a service health notification about EC2 impact in eu-west-2 on August 19, which has since been resolved.
That seems more like an availability and rollout issue than a general AWS failure—the zone appeared before the full set of expected capabilities was ready.