AWS EKS DNS Error(Subnet B succeeds, connection C fails)

Question

I am configuring EKS. I am using US-West-1 and configuring it on V1.29.

The current problem is that when you create a nodeGroup in EKS, add a node, and operate it, the DNS connection fails for the node created in subnet-c. However, the same problem does not occur in nodes created in subnet-b.

Subnet-b and c are using the same Network ACL, and the ACL allows DNS UDP/TCP for in/out bounds.

And the routing table is also structured the same way.

/etc/resolv.conf file

search <name>.svc.cluster.local svc.cluster.local cluster.local us-west-1.compute.internal
nameserver 172.20.0.10
options ndots:5

What else can I check to solve the problem?

For testing purposes, I am upgrading the existing image by installing busybox and testing subnets b and c again.

The strange thing is, when you write the same image link through nodeSelector and deploy by changing only the subnet, you can pull the latest image in B, but in subnet-c, you can only pull the previous image and cannot pull the latest image right away.

Is there a difference in synchronization time between subnets?

Answer 1

Two coreDNSs for EKS appeared only in subnet-b, but were not created in subnet-c, and for this reason, DNS did not work in subnet-c.

After terminating and restarting the coredns pod, the problem was resolved by distributing coredns to each subnet.

Can these problems be prevented in the first place?