K
Q

Kubernetes - How to edit CoreDNS corefile configmap?

6/19/2019

I have a pretty standard installation of Kubernetes running as a single-node cluster on Ubuntu. I am trying to configure CoreDNS to resolve all internal services within my Kubernetes cluster and SOME external domain names. So far, I have just been experimenting. I started by creating a busybox pod as seen here: https://kubernetes.io/docs/tasks/administer-cluster/dns-debugging-resolution/

Everything works as described in the guide until I make changes to the corefile. I am seeing a couple of issues:

  1. I edited the default corefile using kubectl -n kube-system edit configmap coredns and replaced .:53 with cluster.local:53. After waiting, things look promising. google.com resolution began failing, while kubernetes.default.svc.cluster.local continued to succeed. However, kubernetes.default resolution began failing too. Why is that? There is still a search entry for svc.cluster.local in the busybody pod’s /etc/resolv.conf. All that changed was the corefile.

  2. I tried to add an additional stanza/block to the corefile (again, by editing the config map). I added a simple block :

    .:53{
    log
}

    It seems that the corefile fails to compile or something. The pods seem healthy and don’t report any errors to the logs, but the requests all hang and fail.

I have tried to add the log plugin, but this isn’t working since the plugin is only applied to domains matching the plugin, and either the domain name doesn’t match or the corefile is broken.

For transparency, this is my new corefile :

cluster.local:53 {
    errors
    log
    health
    kubernetes cluster.local in-addr.arpa ip6.arpa {
       pods insecure
       upstream
       fallthrough in-addr.arpa ip6.arpa
    }
    prometheus :9153
    forward . /etc/resolv.conf
    cache 30
    loop
    reload
    loadbalance
}
-- A. Davidson
coredns
dns
kube-dns
kubernetes

Similar Questions

Is it possible to undo kubernetes cluster delete command?
Helm rabbitmq endpoints does not exist
Google Kubernetes Engine monitoring is reset on redeployment
kubernetes Python API Client: execute full yaml file

1 Answer

7/3/2019

It looks like your Corefile got somehow corrupted during editing through "kubectl edit ..." command. Probably it's fault of your default text editor, but it's definitely valid.

I would recommend you to replace your current config map with the following command:

kubectl get -n kube-system cm/coredns --export -o yaml | kubectl replace -n kube-system -f coredns_cm.yaml
#coredns_cm.yaml
apiVersion: v1
data:
  Corefile: |
    cluster.local:53 {
        log
        errors
        health
        kubernetes cluster.local in-addr.arpa ip6.arpa {
           pods insecure
           upstream
           fallthrough in-addr.arpa ip6.arpa
        }
        prometheus :9153
        proxy . /etc/resolv.conf
        cache 30
        loop
        reload
        loadbalance
    }
kind: ConfigMap
metadata:
  creationTimestamp: null
  name: coredns
-- Nepomucen
Source: StackOverflow