In the generated version of this file at /contrib/kube-prometheus/manifests/grafana/grafana-deployment.yaml the version was bumped to 4.5.2 but seems it was overlooked in this templated version of the file.

https://github.com/kubernetes/kubernetes/issues/29542 has been fixed,
use apply instead of create to create third party resources in custom service
monitoring example.

Starting with Kubernetes 1.6 `kubectl apply` works for TPRs so we can
start using it, therefore using multiple files is no issue.

Without this, a race condition may occur, preventing grafana-watcher
from logging in to grafana.

These manifests should live in kube-system; otherwise the label selector
won't work.

Datasource links were duplicated in the grafana dashboads. This now also
allows exporting grafana dashboards from the UI and just dropping them
into the assets directory and they will be wrapped by the manifest
generation script.

For `--dry-run` to work with kubectl a Kubernetes cluster's apiserver is
actually used, which is unnecessary for generating these manifests. This
approach also allows further customization, such as adding labels to the
generated manifests.

git-subtree-dir: contrib/kube-prometheus
git-subtree-mainline: 050ca21276696c8603375c699513ec487301ed62
git-subtree-split: 81c0d2f4