CoreDNS ConfigMap auto revert to default

[Erik Ilavsky]

After changing configuration of corefile "kubectl edit configmap coredns -n kube-system", new valid configuration works only few minutes, then is reverted back to default configuration.

Bug is similar like https://github.com/Azure/AKS/issues/811

I try to change in deployment image form rocks.canonical.com:443/cdk/coredns/coredns:1.6.7 to coredns/coredns:1.8.0, after few minutes something killed corendns:1.8.0 and revert back deployment/coredns:1.6.7

kubectl logs -f coredns-7bb4d77796-6mv6z -n kube-system
.:53
[INFO] plugin/reload: Running configuration MD5 = e489106e2ec809a2688af7714fc2421e
CoreDNS-1.6.7
linux/amd64, go1.13.6, da7f65b
[INFO] Reloading
[INFO] plugin/health: Going into lameduck mode for 5s
[INFO] plugin/reload: Running configuration MD5 = 378fe9d7068a8941a2477f27b5655c63
[INFO] Reloading complete
[INFO] Reloading
[INFO] plugin/health: Going into lameduck mode for 5s
[INFO] plugin/reload: Running configuration MD5 = e489106e2ec809a2688af7714fc2421e
[INFO] Reloading complete

kubectl logs -f coredns-5f467455fd-7vwzk -n kube-system
.:53
somedomain.org.:53
CoreDNS-1.8.0
linux/amd64, go1.15.3, 054c9ae
[INFO] SIGTERM: Shutting down servers then terminating
[INFO] plugin/health: Going into lameduck mode for 5s

I think it's uncorrect behavior because by charm coredns isn't installed (have corefile config option), but in this case plugin or bundle don't have corefile option, therefore it cann't modified.

[George Kraft (cynerva)]

This is expected behavior when using addons deployed by cdk-addons. If you would like more control over the configuration of CoreDNS, you will need to use the CoreDNS charm as documented here: https://ubuntu.com/kubernetes/docs/cdk-addons#coredns