I'm currently setting up an application platform on a new cloud provider, OpenTelekomCloud, which is based on Huawei Cloud and OpenStack. I've previously worked on AWS EKS and clusters set up with kubeadm, where the kube-dns service in the kube-system namespace was always present and pointed to a CoreDNS deployment. I took that for granted. Now, however, I'm noticing that the kube-dns service is missing in this new environment, and there's only the CoreDNS deployment available. For most workloads, using the provided /etc/resolv.conf works fine, but I've run into an issue with the Grafana Loki helm chart that relies on the kube-dns service for nginx configuration. I created the service myself to point to the CoreDNS pods, and it seems to work, but now I'm confused about who should be responsible for the missing service. Is OpenTelekomCloud at fault for not including it, or should I blame Grafana/Loki for assuming kube-dns is available? Any insights or documentation would be appreciated!
1 Answer
During the 1.1x.y releases, Kubernetes transitioned from kube-dns to CoreDNS. If your helm chart is still relying on kube-dns, you might want to check for a newer version. You’ve got a workaround, which is cool, but it could be worth looking into updates for the chart to align it with what most current setups use.
True, AWS EKS and kubeadm still show the kube-dns service even if they use CoreDNS, so many might assume it's a standard. But since the Loki helm chart is up to date, maybe it's more about how Helm values are set for your setup.