Lack of pod connectivity on EKS

0

I have an EKS cluster(1.24) launched with Terraform's AWS EKS module. System generated security groups. Private or public subnets.

Cluster Inbound: cluster inbound

Node Inbound: node inbound

Node Outbound: node outbound

The cluster has 1 node and the backplane. It comes online the node connects and is in a ready state. Although coredns is mega sad as it can't fetch the version from the Cluster API.

When it comes online there is no sign of issue but it wont connect to the control plane API via the local loop back. And it has no internet connectivity.


./kubectl run -i --tty --rm debug --image=busybox -- sh
If you don't see a command prompt, try pressing enter.
/ #  wget $KUBERNETES_PORT_443_TCP_ADDR:$KUBERNETES_PORT_443_TCP_PORT/version
Connecting to 172.20.0.1:443 (172.20.0.1:443)
wget: can't connect to remote host (172.20.0.1): Connection timed out

/ # nc gooogle.com
nc: bad address 'gooogle.com'

I've checked the NACL against another account where its working and it looks okay to me. I've tried it with and without the VPC CNI. NACL Inbound: naclinbound NACL Outbound: nacloutbound

At node level, the node can connect to the internet, and pull images. What am I missing at pod level, why can't it connect to the API and why can't it connect to the internet? I don't have admin rights for the VPC and Subnetting but I have checked it against another account where the same stuff seems to work.

已提問 1 年前檢視次數 904 次
2 個答案
0

Someone who comes up with our custom AMI images decided to set sysctl net.ipv4.ip_forward to 0 which blocked all network traffic

已回答 1 年前
0

For AWS level networking configuration (i.e. the VPC), I recommend testing your connectivity with the VPC Reachability Analyzer tool. If that doesn't clarify the problem then it is likely at the Kubernetes or Operating System levels.

AWS
已回答 1 年前

您尚未登入。 登入 去張貼答案。

一個好的回答可以清楚地回答問題並提供建設性的意見回饋,同時有助於提問者的專業成長。

回答問題指南