#eks
Amazon EKS in production: pod IP exhaustion, scheduling failures, autoscaling choices, and the networking decisions that drive most of the cluster cost.
4 articles tagged #eks, newest first.
100.64.0.0/10 in AWS: Using CGNAT Space to Fix EKS IP Exhaustion
When an EKS cluster runs out of pod IPs and the VPC has no RFC 1918 space left, a 100.64.0.0/10 secondary CIDR is the fix. Why AWS allows it, and what breaks.
AWS `failed to assign an IP address to container`: What It Means and How to Fix It
failed to assign an IP address to container is the VPC CNI out of addresses: a full subnet, or a node at its ENI limit. Ranked causes and fixes.
AWS `0/N nodes are available: Insufficient cpu`: What It Means and How to Fix It
0/N nodes are available: Insufficient cpu is the scheduler saying no node has room — measured against requests, not usage. Causes and their fixes.
The VPC DNS Limit: 1,024 Packets per Second per ENI
The VPC resolver caps every network interface at 1,024 DNS packets per second. It is not adjustable, has no CloudWatch metric, and looks like random SERVFAIL.