It's important to be able to development and test against a Kubernetes cluster which you can setup and manage locally on your laptop. My personal favorite tool is to use Kind to setup a multi-node Kubernetes cluster locally. This repository contains documentation and resources to help one construct a Kind environment that actually works on their laptop including proper ingress support.
Note, the instructions in this respository assume you are using a Mac for development. Variations may exist if you use Linux or Windows.
There are some tools that you will need to setup prior to configuring your Kind environment.
Kind runs Kubernetes nodes as containers; therefore, you must have Docker Desktop installed.
Follow the instructions here to install Docker Desktop for Mac. Nothing difficult, just follow the instructions.
You will need to adjust the settings of your Docker Desktop to ensure that the kubernetes clusters work well.
Go ahead and start Docker Desktop if you haven't already.
From the menu select Preferences.
Go to the Resources section to adjust the CPU and Memory allocated to containers from your laptop. I would recommend adjusting the values to be at least 4 cores for CPUs and 8GB for Memory. You can go higher if you wish but these values should be sufficient for most local development needs.
Click Apply & Restart button.
You must have the Kubernetes CLI installed. Not absolutely necessary to run Kind but, let's face it, you will be using Kubernetes so you need the CLI installed.
The simplest approach to install the k8s CLI on a mac is via homebrew.
brew install kubectl
Other CLI installation options are available here.
You will obviously need Kind installed as well.
brew install kind
Addtional installation options can be found here.
Lens is a nice graphical tool for accessing, inspecting, and debugging your clusters. Lens exercises kubectl commands under the covers but it allows novice users to easily explore and access information about resources in the cluster. Lens also allows expert users to be more efficient when working with clusters.
You can install Lens directly from the Lens Development site. Simply click on the Download button and select the archive for your operating system. Unzip the archive and execute the install to have Lens installed on your system.
I'm going to walk you through creating a simple three node cluster with a non-HA control node. Kind uses a simple YAML file to describe the cluster. I'm not going to walk you through every setting but I will explain some key points that you will want to use.
First ensure that you clone this repository and cd into the root directory from a terminal session.
git clone [email protected]:dcberg/kind.git
You can kick off the creation now and then read about the key points in the sections that follow.
kind create cluster --name test --config scripts/cluster-3.yaml
This script will create a three node Kube 1.19 cluster. There is a single control node identified by the control-plane role and the version is controlled by the container image that is used. In this case the 1.19 node image is used. You can find the full list of container node images in the release notes.
- role: control-plane
image: kindest/node:v1.19.1@sha256:98cf5288864662e37115e362b23e4369c8c4a408f99cbc06e58ac30ddc721600
The worker nodes are exactly the same as the control-plane node except with the role set to worker.
To allow ingress to work within a local Kind cluster, it is necessary to select one of the nodes to host the ingress controller. On Mac and Windows machines, Docker Desktop does not bridge the container network with the host. Thus you cannot use the node IP directly from the host (i.e., your laptop). In the config yaml above, I have selected the first worker node to be available for ingress. I adjusted the worker node to inject a label (ingress-ready=true) which will be used by a node-selector when deploying the ingress-controller. The other adjustment is to set extraPortMappings to map the container 80 and 443 ports to the 80 and 443 ports on the host (i.e., your laptop). Note, only one node can map to the ports on the host. It is also necessary to set the listenAddress to be 127.0.0.1 for the local. This combination of the host port maps and the localhost listenAddress allows the ingress to be addressed directly from your local machine.
- role: worker
image: kindest/node:v1.19.1@sha256:98cf5288864662e37115e362b23e4369c8c4a408f99cbc06e58ac30ddc721600
kubeadmConfigPatches:
- |
kind: JoinConfiguration
nodeRegistration:
kubeletExtraArgs:
node-labels: "ingress-ready=true"
extraPortMappings:
- containerPort: 80
hostPort: 80
listenAddress: "127.0.0.1"
protocol: TCP
- containerPort: 443
hostPort: 443
listenAddress: "127.0.0.1"
protocol: TCP
You can validate that the cluster was created successfully.
kind get clusters
will list the test cluster that you just created. You can get the list of nodes for the cluster as well.
kind get nodes --name test
test-worker2
test-control-plane
test-worker3
test-worker
After creating a Kind cluster, the kubernetes context is automatically set to the new cluster. You can confirm this by running the following command.
kubectl cluster-info --context kind-test
You should see a result similar to mine.
Kubernetes control plane is running at https://127.0.0.1:53759
KubeDNS is running at https://127.0.0.1:53759/api/v1/namespaces/kube-system/services/kube-dns:dns/proxy
To further debug and diagnose cluster problems, use 'kubectl cluster-info dump'.
You can now run kubectl commands. Try getting the nodes.
kubectl get nodes
NAME STATUS ROLES AGE VERSION
test-control-plane Ready master 9m11s v1.19.1
test-worker Ready <none> 8m40s v1.19.1
test-worker2 Ready <none> 8m40s v1.19.1
test-worker3 Ready <none> 8m40s v1.19.1
Launch Lens if you haven't already. Lens will open with no clusters connected.
Click on the + button to configure a new connection to your kind cluster.
By default Lens will read your system kubeconfig file and automatically detect the new kind-test cluster that you created. If you have additional clusters on your system, you may need to select the kind-test context from the drop down menu.
Lens will now connect to your cluster and you can explore a bit to ensure that resource information is properly being read (e.g., click on Nodes to see the three worker nodes).
If you select on the Cluster section in the left menu you will see messages indicating that metrics are not available because Prometheus has not been setup. You could manually setup Prometheus yourself or you can let Lens do it for you. Having metrics available directly in Lens is helpful for debugging memory leaks as well as setting up proper resource requests and limits for your containers.
Right click on the image of your cluster in the far left navigation and select Settings
Scroll to the bottom of the Settings page to the Features section and click on the Install button for the Metrics Stack. This will setup a prometheus instance within your cluster.
You will need to wait a few minutes while the install is completing and for metrics to be gathered for the cluster. Close the settings page. You should now be on the main Cluster page and you should start to see cluster wide metrics.
If you navigate to the Pods section under Workloads you will see metrics per pod as well.
If you plan to test a horizontal pod autoscaler or to check the resource consumption of a pod or node, you will need the Kubernetes metrics server. The server is not installed by default. You can use the following command to install the metrics server.
kubectl apply -f scripts/metrics-server-components.yaml
You now have a basic three node cluster with metrics gathering. Now we want to be able to deploy and access services. We will do that through an ingress controller in the cluster. We will be using the kubernetes community nginx ingress controller.
From your terminal session, execute the following command to install the nginx ingress controller into your cluster.
kubectl apply --filename https://raw.githubusercontent.com/kubernetes/ingress-nginx/master/deploy/static/provider/kind/deploy.yaml
kubectl wait --namespace ingress-nginx \
--for=condition=ready pod \
--selector=app.kubernetes.io/component=controller \
--timeout=90s
The config file is specifically designed for kind clusters to ensure that the ingress controller is deployed onto the node that you have labelled as ingress-ready=true. This is important within the kind environment and Docker Desktop so that we can access the services from your local host.
As a side note, Lens will update showing you the deployments that are taking place. For example you can see the pending deployments on the main Workloads page.
Let's deploy a simple hello-world service into the cluster for testing purposes.
kubectl run hello \
--expose \
--image nginxdemos/hello:plain-text \
--port 80
We will now configure an ingress resource that will allow us to access the hello service using a URL and our nginx ingress controller.
kubectl apply -f - <<EOF
apiVersion: networking.k8s.io/v1
kind: Ingress
metadata:
name: hello
spec:
rules:
- host: hello.test.com
http:
paths:
- pathType: ImplementationSpecific
backend:
service:
name: hello
port:
number: 80
EOF
I have defined a host value of hello.test.com. Obviously this domain name is not resolvable on the internet so we need to adjust our local system to allow this domain to be resolved locally. To do this you will need to edit your local /etc/hosts file and add the following line.
127.0.0.1 hello.test.com
Note, we are using the localhost address which is the same as the listenAddress that we included in the kind cluster config for worker node that is hosting the ingress controller. Hopefully this is all making sense now.
From your terminal (or web browser), you should be able to access the service.
curl hello.test.com
Server address: 10.244.1.5:80
Server name: hello
Date: 10/Feb/2021:16:47:27 +0000
URI: /
Request ID: 50eec66a0f425819257a1190a1cb1ab7
The previous sections describe how to setup a Kind environment using container images that you have built and deployed. This is useful for testing a build or other packaged solutions within a local Kubernetes environment. However, there are times that you would like to have a highly dynamic development environment where you change code and see the changes immediately appear within your Kubernetes environment. The following instructions will show you how to setup a dynamic local development environment using a Kind cluster.
We will be using a very simple Hello World example to demonstrate how to setup and use a local kubernetes environment that is directly mounting a development directory.
First ensure that you have deleted any existing clusters. You can check for existing clusters using the get command.
kind get clusters
If you have a cluster from previously, delete it using the following delete command.
kind delete cluster --name test
We will create a new cluster that leverages the extraMounts feature in Kind to mount a directory on your local machine. Before we can do this we need to ensure that Docker Desktop is configured with the proper rights to mount your local drive.
Open the Docker Desktop Preferences.
Select Resources > FILE SHARING. These are the directories that Docker has the rights to mount. When we create the kind cluster, we will mount a directory that must be either listed directly here or a subdirectory of a directory listed here. In my example I have added /Users/dberg/development/github which is the root directory where I clone git repositories. This allows me to mount any directory under the github root directory. You will need to do something similar on your local system.
Note, if you are not using Docker Desktop then you will likely need to configure a similar preference using the Docker Engine directly.
With Docker Desktop configured, you can now create new cluster for our Hello World example. Create a new local cluster starting from the root of of the kind repository.
kind create cluster --name hello --config scripts/cluster-3-local.yaml
Let's take a closer look at what is happening behind the scenes. The configuration file is creating the same three node cluster with the exception that we are adding a mount to a directory on the local host machine. This is done by adding an extraMounts section to each of the worker nodes. You need to ensure that you add this section to each worker to ensure that when a pod is scheduled to any of the worker nodes that it has access to this local host directory.
extraMounts:
# Set the hostPath to a path on your local system (can be relative).
- hostPath: hello
# This is the path to be exposed within the container nodes.
containerPath: /kind/apps/hello
The hostPath is the directory on your local machine where you have your code that you want to make changes and run within the containers.
The containerPath is the path that will be mounted within worker node container. From the point of view of the kubernetes pods that you are deploying, this path will be the host path on the node.
Execute the following command to install the nginx ingress controller.
kubectl apply --filename https://raw.githubusercontent.com/kubernetes/ingress-nginx/master/deploy/static/provider/kind/deploy.yaml
kubectl wait --namespace ingress-nginx \
--for=condition=ready pod \
--selector=app.kubernetes.io/component=controller \
--timeout=90s
Next we will build the image that will be used for testing the hello world application.
docker build -t kind/hello-py:latest ./hello
If you look into the Dockerfile, you will see on line 4 that the working directory is switched to /hello. This is where we will have the python code for our application. Line 9 has been commented out because it is not necessary to copy the hello-world.py file into the image since we will be mounting the directory with the code (i.e., we are not copying the code directly into this dev image).
Once the image is built, you need to load the image into the kind cluster so it can be deployed.
kind load docker-image kind/hello-py:latest --name hello
You can now deploy the hello world example within your local cluster. From the kind directory, execute the following command:
kubectl apply -f hello/hello-svc.yaml
Let's exam the kubernetes yaml file to see how everything is setup to make the magic work.
Open thge hello/hello-svc.yaml file.
Line 19 captures the image kind/hello-py:latest (just built and loaded into the cluster) that will be deployed in the cluster.
Line 23 shows where we are mounting the directory where the source code resides. The mountPath must match the WORKDIR used in the Dockerfile.
volumeMounts:
- mountPath: /hello
name: hello-volume
The volumeMount must match a volume entry defined within the cluster namespace by the same name (hello-volume). Line 26 defines the hello-volume volume details. Note that the hostPath of the volume must match the containerPath defined in the kind cluster configuratiomn for the extraMounts. In this case the value is /kind/apps/hello as this is the path that will be on the worker node.
volumes:
- name: hello-volume
hostPath:
# Directory path from the containerPath in the Kind extraMounts section
path: /kind/apps/hello
That's the secret sauce for getting the proper volume mounts needed to access source code files directly on your local machine from within the container in the kind cluster.
If you still have hello.test.com in your hosts file then you can curl the end point to see the message from your deployment.
curl hello.test.com
Hello, World!
Now open /hello/hello-world.py in a code editor.
On line 8, change the message to something different such as Hello Everyone!.
Save your editor and run the curl command again.
curl hello.test.com
Hello Everyonbe!
Since the container deployment is mounting your local drive, you are able to make changes directly to your source code without building and redeploying your container(s).
You have successfully setup and configured a local kubernetes environment. You can now explore, deploy, and configure additional resources in our cluster to your heart's content.