How can I configure kubectl to interact with both minikube and a deployed cluster?

Question

When you use minikube, it automatically creates the local configurations, so it's ready to use. And it appears there is support for multiple clusters in the kubectl command based on the reference for kubectl config.

In the docs for setting up clusters, there's a reference to copying the relevant files to your local machine to access the cluster. I also found an SO Q&A about editing your .kube/config to leverage azure remotely that talked to editing the kube/config file.

It looks like the environment variable $KUBECONFIG can reference multiple locations of these configuration files, with the built-in default being ~/.kube/config (which is what minikube creates).

If I want to be able to use kubectl to invoke commands to multiple clusters, should I download the relevant config file into a new location (for example into ~/gcloud/config, set the KUBECONFIG environment variable to reference both locations?

Or is it better to just explicitly use the --kubeconfig option when invoking kubectl to specify a configuration for the cluster?

I wasn't sure if there was some way of merging the configuration files that would be better, and leverage the kubectl config set-context or kubectl config set-cluster commands instead. The documentation at Kubernetes on "Configure Access to Multiple Clusters" seems to imply a different means of using --kubeconfig along with these kubectl config commands.

In short, what's the best way to interact with multiple separate kubernetes clusters and what are the tradeoffs?

Answer 1

If I want to be able to use kubectl to invoke commands to multiple clusters, should I download the relevant config file into a new location (for example into ~/gcloud/config, set the KUBECONFIG environment variable to reference both locations?

Or is it better to just explicitly use the --kubeconfig option when invoking kubectl to specify a configuration for the cluster?

That would probably depend on the approach you find simpler and more convenient, and if having security and access management concerns in mind is needed.

• From our experience merging various kubeconfig files is very useful for multi-cluster operations, in order to carry out maintenance tasks, and incident management over a group of clusters (contexts & namespaces) simplifying troubleshooting issues based on the possibility to compare configs, manifests, resources and states of K8s services, pods, volumes, namespaces, rs, etc.

• However, when automation and deployment (w/ tools like Jenkins, Spinnaker or Helm) are involved most likely having separate kubeconfig files could be a good idea. A hybrid approach can be merging kubeconfig files based on a division by Service tier -> Using files to partition development landscapes (dev, qa, stg, prod) clusters or for Teams -> Roles and Responsibilities in an Enterprise (teamA, teamB, …, teamN) can be understood also within good alternatives.

• For multi-cluster merged kubeconfig files scenarios consider kubectx + kubens, which are very powerful tools for kubectlt that let you see the current context (cluster) and namespace, likewise to switch between them.

In short, what's the best way to interact with multiple separate kubernetes clusters and what are the trade offs?

• The trade offs should possibly be analyzed considering the most important factors for your project. Having a single merged kubeconfig file seems simpler, even simple if you merge it with ~/.kube/config to be used by default by kubectl and just switching between cluster/namespaces with --context kubectl flag. On the other hand if limiting the scope of the kubeconfig is a must, having them segregated and using --kubeconfig=file1 sounds like the best way to go.

• Probably there is NOT a best way for every case and scenario, knowing how to configure kubeconfig file knowing its precedence will help though.

• In this article -> https://www.nrmitchi.com/2019/01/managing-kubeconfig-files/ you'll find a complementary and valuable opinion:

  • While having all of the contexts you may need in one file is nice, it is difficult to maintain, and seldom the default case. Multiple tools which provide you with access credentials will provide a fresh kubeconfig to use. While you can merge the configs together into ~/.kube/config, it is manual, and makes removing contexts more difficult (having to explicitly remove the context, cluster, and user). There is an open issue in Kubernetes tracking this. However by keeping each provided config file separate, and just loading all of them, removal is much easier (just remove the file). To me, this seems like a much more manageable approach.

  • I prefer to keep all individual config files under ~/.kube/configs, and by taking advantage of the multiple-path aspect of the $KUBECONFIG environment variable option, we can make this happen.

If you’re using kubectl, here’s the preference that takes effect while determining which kubeconfig file is used.

1. use --kubeconfig flag, if specified
2. use KUBECONFIG environment variable, if specified
3. use $HOME/.kube/config file

With this, you can easily override kubeconfig file you use per the kubectl command:

#
# using --kubeconfig flag
#
kubectl get pods --kubeconfig=file1
kubectl get pods --kubeconfig=file2

#
# or 
# using `KUBECONFIG` environment variable
#
KUBECONFIG=file1 kubectl get pods
KUBECONFIG=file2 kubectl get pods

#
# or 
# merging your kubeconfig file w/ $HOME/.kube/config (w/ cp backup)
#
cp $HOME/.kube/config $HOME/.kube/config.backup.$(date +%Y-%m-%d.%H:%M:%S)
KUBECONFIG= $HOME/.kube/config:file2:file3 kubectl config view --merge --flatten > \
~/.kube/merged_kubeconfig && mv ~/.kube/merged_kubeconfig ~/.kube/config
kubectl get pods --context=cluster-1
kubectl get pods --context=cluster-2

NOTE: The --minify flag allows us to extract only info about that context, and the --flatten flag allows us to keep the credentials unredacted.

---

Bonus (extra points!)

Using multiple kubeconfigs at once

You can save AKS (Azure Container Service), or AWS EKS (Elastic Container Service for K8s) or GKE (Google Container Engine) cluster contexts to separate files and set the KUBECONFIG env var to reference both file locations.

For instance, when you create a GKE cluster (or retrieve its credentials) through the gcloud command, it normally modifies your default ~/.kube/config file. However, you can set $KUBECONFIG for gcloud to save cluster credentials to a file:

KUBECONFIG=c1.yaml gcloud container clusters get-credentials "cluster-1"

Then as we mentioned before using multiple kubeconfigs at once can be very useful to work with multiple contexts at the same time.

To do that, you need a “merged” kubeconfig file. In the section "Merging kubeconfig files" below, we explain how you can merge the kubeconfigs into a single file, but you can also merge them in-memory.

By specifying multiple files in KUBECONFIG environment variable, you can temporarily stitch kubeconfig files together and use them all in kubectl .

#
# Kubeconfig in-memory merge
#
export KUBECONFIG=file1:file2
kubectl get pods --context=cluster-1
kubectl get pods --context=cluster-2

#
# For your example
# merging your kubeconfig file w/ $HOME/.kube/config (w/ cp backup)
#
cp $HOME/.kube/config $HOME/.kube/config.backup.$(date +%Y-%m-%d.%H:%M:%S)
KUBECONFIG= $HOME/.kube/config:file2: kubectl config view --merge --flatten > \
~/.kube/merged_kubeconfig && mv ~/.kube/merged_kubeconfig ~/.kube/config
kubectl get pods --context=cluster-1
kubectl get pods --context=cluster-2

Merging kubeconfig files

Since kubeconfig files are structured YAML files, you can’t just append them to get one big kubeconfig file, but kubectl can help you merge these files:

#
# Merging your kubeconfig file w/ $HOME/.kube/config (w/ cp backup)
#
cp $HOME/.kube/config $HOME/.kube/config.backup.$(date +%Y-%m-%d.%H:%M:%S)
KUBECONFIG=$HOME/.kube/config:file2:file3 kubectl config view --merge --flatten > \
~/.kube/merged_kubeconfig && mv ~/.kube/merged_kubeconfig ~/.kube/config
kubectl get pods --context=cluster-1
kubectl get pods --context=cluster-2

• ref article 1: https://ahmet.im/blog/mastering-kubeconfig/
• ref article 2: https://github.com/kubernetes/kubernetes/issues/46381

Answer 2

I have a series of shell functions that boil down to kubectl --context=$CTX --namespace=$NS, allowing me to contextualize each shell [1]. But if you are cool with that approach, then rather than rolling your own, https://github.com/Comcast/k8sh will likely interest you. I just wish it was shell functions instead of a sub-shell

But otherwise, yes, I keep all the config values in the one ~/.kube/config

footnote 1: if you weren't already aware, one can also change the title of terminal windows via title() { printf '\033]0;%s\007' "$*"; } which I do in order to remind me which cluster/namespace/etc is in effect for that tab/window