A Fargate service using the awsvpc network mode cannot launch additional tasks because of InsufficientFreeAddressesInSubnet.
The task CPU and memory are available, and the service must continue scaling out.
Which action is most appropriate?
Because awsvpc tasks consume ENIs and subnet IP addresses, provide larger subnets across multiple Availability Zones and update the ECS service placement.
Detailed explanation
CDN cache capacity does not affect ENI allocation or the available IP addresses in a VPC subnet.
CDN cache capacity does not affect ENI allocation or the available IP addresses in a VPC subnet.
Each awsvpc task needs an ENI and an IP address. Increasing subnet capacity and distributing placement provides launch headroom and availability.
Each awsvpc task needs an ENI and an IP address. Increasing subnet capacity and distributing placement provides launch headroom and availability.
Changing CPU does not create free subnet IP addresses, so task launches fail for the same reason.
Changing CPU does not create free subnet IP addresses, so task launches fail for the same reason.
Fargate tasks use the awsvpc network mode; this change is not available as a way to solve subnet address exhaustion.
Fargate tasks use the awsvpc network mode; this change is not available as a way to solve subnet address exhaustion.
Check official documentation
Compare the following points with the official documentation.
Amazon ECS awsvpc networkingとVPC subnet sizingの公式説明を確認する。
Official references
What to confirm
ECSタスクのIP消費とサブネットCIDRの空き容量を、CPU・メモリの制約と区別して調査できる。
Key points
- awsvpc
- ENI
- Subnet CIDR
- IP address exhaustion
Notes
- Environment: AWS公式SOA-C03試験ガイドとAWS公式ドキュメントの確認
- Confirm current behavior in the official SOA-C03 guide and AWS service documentation.
Foundation review
Capacity boundary
ECS scale-out can be blocked by networking capacity even when CPU and memory remain available.
Multi-AZ placement
Using subnets in multiple Availability Zones adds capacity and reduces a single-subnet dependency.