DNS Resolver : The name to resolve is too specific.

[treussart]

Describe the bug
Into the file:

loki/production/ksonnet/loki/config.libsonnet

Line 96 in 9cef86b

dns_resolver: 'kube-dns.kube-system.svc.cluster.local',

The pod is crashloopbackoff because it cannot resolve the name: kube-dns.kube-system.svc.cluster.local

On my k8s cluster the name is coredns instead of kube-dns.

For resolve this issue, I created a new service with the same spec than coredns, but I think we should be able to change it in the Helm chart config.

To Reproduce
Steps to reproduce the behavior:

1. Rename the service to something other than kube-dns
2. Install the chart grafana/loki-simple-scalable

Expected behavior
A crashloopbackoff of the pods loki-simple-scalable-*

Environment:

• Infrastructure: Kubernetes
• Deployment tool: helm

[nktl]

Same problem here on RedHat OpenShift - it uses CoreDNS and kube-dns.kube-system.svc.cluster.local does not exist there.

[irizzant]

This is configurable through the chart options

[erSitzt]

Just for anyone who might stumble upon this..

RKE2 Cluster

global:
  clusterDomain: cluster.local
  dnsNamespace: kube-system
  dnsService: rke2-coredns-rke2-coredns