Intermittent timed out accessing nodelocaldns about dns HOT 2 CLOSED

I'm noticing a similar issue here when working with GitLab Runners. Was curious if there has been any movement on this one. I notice it on both self-host Kubernetes cluster (RKE2) and AWS EKS clusters.

from dns.

I managed to solve it.
I identified initially the problem as related with localdns but the root cause was an incompatible issue with DockerInDocker container of the self hosted runner. If the container is deployed as a side-car container, then the networking at the pod was not working correctly.
The initial issue was identified as a problem with nodelocaldns but it was not working many other things like:

• Ping to external nodes. Executing this command ping 8.8.8.8 -c 5 resulted in a timeout and host not reachable.
• Telnet to nodelocaldns telnet 10.1.0.10 53
• Telnet to coredns telnet 10.1.0.3 53
• Ping to other pod running in the same node: ping 10.2.0.134
• Ping from other pod running in the same node to the github runner didn't work either

So, in order to remove the side-container it was needed to set this attribute:

dockerEnabled: false

And then in the Github action workflow, specify to use the Kubernetes driver for the docker Build X action.

      - name: Set up Docker Buildx
        uses: docker/setup-buildx-action@v2
        with:
          driver: kubernetes

I suppose that this step will be different in Gitlab.

Finally, check the MTU from the container and the Kubernetes cluster because I had to customised it with this attribute:

dockerMTU: 1400

from dns.