get_service_ip is not honoring the kubectl config file

[Cory Johns]

As of 1.18-beta.2, kubectl no longer falls back to connecting to localhost:8080 when no config file can be found and instead generates a fatal error. This leads to the following error:

unit-kubernetes-master-0: 14:09:15 INFO unit.kubernetes-master/0.juju-log kube-control:6: Invoking reactive handler: reactive/kubernetes_master.py:993:send_cluster_dns_detail
unit-kubernetes-master-0: 14:09:15 DEBUG unit.kubernetes-master/0.kube-control-relation-changed error: Missing or incomplete configuration info. Please point to an existing, complete config file:
unit-kubernetes-master-0: 14:09:15 DEBUG unit.kubernetes-master/0.kube-control-relation-changed
unit-kubernetes-master-0: 14:09:15 DEBUG unit.kubernetes-master/0.kube-control-relation-changed 1. Via the command-line flag --kubeconfig
unit-kubernetes-master-0: 14:09:15 DEBUG unit.kubernetes-master/0.kube-control-relation-changed 2. Via the KUBECONFIG environment variable
unit-kubernetes-master-0: 14:09:15 DEBUG unit.kubernetes-master/0.kube-control-relation-changed 3. In your home directory as ~/.kube/config
unit-kubernetes-master-0: 14:09:15 DEBUG unit.kubernetes-master/0.kube-control-relation-changed
unit-kubernetes-master-0: 14:09:15 DEBUG unit.kubernetes-master/0.kube-control-relation-changed To view or setup config directly use the 'config' command.
unit-kubernetes-master-0: 14:09:15 INFO unit.kubernetes-master/0.juju-log kube-control:6: DNS addon service not ready yet

The get_service_ip in kubernetes-common needs to be fixed to explicitly point to the config file since HOME is not set in a Juju hook context.

[George Kraft (cynerva)]

PR: https://github.com/charmed-kubernetes/layer-kubernetes-common/pull/11