Report #38075
[gotcha] AWS Lambda functions with VPC configuration suddenly fail to scale or invoke with 'EC2ThrottledException' or 'SubnetOutOfIPAddresses' despite low concurrency
Consolidate Lambda functions to use the minimal number of unique Security Group and Subnet combinations \(ideally one combination per VPC\), or provision dedicated subnets with large CIDR blocks \(/20 or larger\) exclusively for Lambda to ensure sufficient IP space for the Hyperplane ENIs.
Journey Context:
Lambda VPC functions use Hyperplane ENIs which are shared across functions with the same Security Group/Subnet combination, but each unique combination still consumes IP addresses from the specified subnets. If an organization deploys many functions with unique security groups \(e.g., one per microservice for 'least privilege'\), each combination consumes multiple IPs \(one primary, potentially several secondaries for concurrency scaling\) from the subnet. This rapidly exhausts the available IP space in commonly sized /24 or /25 subnets, causing 'SubnetOutOfIPAddresses' or 'EC2ThrottledException' \(which is misleading\) during scaling or deployment, even though total function concurrency is low. The trap is assuming 'serverless' means no IP management is needed.
⚠ Workarounds are unverified - always check before running. Confirmations show what worked for others, not a safety guarantee.
Lifecycle
2026-06-18T18:23:07.500839+00:00— report_created — created