Traefik Mesh not resolving in new AWS EKS deployment

Traefik Mesh
#1

AWS: eks.4
CoreDNS: v1.8.4-eksbuild.1
Mesh: helm-4.0.2

Values:
kubedns: false

No logs from controller;

.:53
maesh.:53
traefik.mesh.:53
[INFO] plugin/reload: Running configuration MD5 = ff9bb101e62cade152ba36d623f72657
CoreDNS-1.8.4
linux/amd64, go1.13.15, 053c4d5c
[INFO] Reloading
[INFO] plugin/reload: Running configuration MD5 = 47d57903c0f0ba4ee0626a17181e5d94
[INFO] Reloading complete

Issue: svc.ns.svc.cluster.local resolves fine; svc.ns.traefik.mesh does not resolve

resolve.conf doesn't show any new pointers in the search

Silly issue: it worked for 30 seconds, no changes, I had four successful curl -Lv svc.ns.traefik.mesh and then it stopped resolving. Controller/proxy contained NO LOGS!

#2

Resolved.

eksctl delete addon --cluster my_cluster --region us-east-1 --name coredns --preserve

coredns configmap was being reverted by eks managed addon.

#3

Sorry for responding to the very old issue, but curious how you figured out it was the addon that was reverting the configmap? I'm having a very similar issue on k3s where DNS resolves for about a minute after deploying the Traefik Mesh helm chart, then something seems to break. Also not getting any helpful logs from the controller/proxy, so any thoughts on where to look would be super appreciated!

#4

I realized that upstream was breaking because the mesh.traefik route would fail to resolve. I simply checked the coredns config after installing and saw the patch, then I checked after it broke and saw it was reverted.

Since coredns is managed, I checked the managed fields:

kubectl get deployment/coredns -n kube-system -o yaml

This led me to realize the configmap patch was overwriting managed fields which would be then overwritten by the coredns deployment.