This document describes various installation options for Kubernetes and extensions which we are using in the pattern examples. All examples has been tested with Minikube thoroughly, but should also work without change on the other installation options.
Content:
minikube is recommended not only for trying out theses examples but also for an easy and simple to use Kubernetes setup in general.
You can download Minikube from …. Essentially, Minikube is a CLI tool for starting a virtual machine with a Docker daemon and a single node Kubernetes cluster.
To start Minikube you just call
minikube start|
Note
|
When running on Mac with an M1 or M2 chip, for the best experience it is recommended to use Docker as the underlying virtualization platform instead of the default of qemu2. It allows for better networking and tunneling for the examples that involved networking. You can chose the Docker driver with the option --driver docker or set it permanently with minikube config set driver docker. Of course, for this to work, you need to have Docker Desktop installed locally.
|
If you do this for the first time Minikube will download all required files and starts a VM whose default runtime depends on your operating system and installation options.
Some of the examples use an Ingress object to export a service to the outside.
For Minikube you have to enable ingress support with:
minikube addons enable ingressOtherwise, the pure Kubernetes examples should work out of the box with the default options of Minikube.
If you want to optimise Minikube, use minikube start --help for all options available for starting Minikube.
The most important options are
--memory-
Set the heap memory of the VM. The default of 2048 is sometimes too less, especially when using more advanced deployments. Use 4096 (or even 8192) to be on the safe side.
--vm-driver-
Use this option if you want to change the VM to use, because the default choice doesn’t works for you. Check
minikube start --help | grep vm-driverto find out the possible options. --disk-size-
Size of the VM’s hard drive. The default of 20 GB might be too small if you use many images or too large if you only want to use Minikube for our examples.
You can easily access the Docker daemon running within the Minikube VM by using:
eval $(minikube docker-env)Now you can just use the Docker CLI docker as usual.
There are many more features to explore (like ssh-ing into the VM, exposing Services via NodePort, port-forwarding to Pods e.g. for remote debugging and more) Please consult the Minikube documentation for details.
An alternative to minikube is kind, which can also be used to simulate a multi-node cluster on your Desktop.
kind (Kubernetes in Docker) is a tool for running local Kubernetes clusters using Docker container nodes. It’s often used for development and testing purposes. You can easily create a multinode cluster with kind by defining a configuration file that specifies the desired number of control-plane and worker nodes.
Here’s a step-by-step guide to creating a multinode cluster with kind:
- Install kind and Docker
-
First, make sure you have Docker installed. You can follow the installation instructions for your platform from the official Docker website: https://docs.docker.com/get-docker/. Next, install kind. You can follow the installation instructions from the kind GitHub repository: https://github.com/kubernetes-sigs/kind#installation-and-usage
- Create a configuration file
-
Create a new YAML file, for example,
kind-multinode.yaml, and configure the desired number of control-plane and worker nodes. The following example creates a cluster with one control-plane node and two worker nodes:
kind: Cluster
apiVersion: kind.x-k8s.io/v1alpha4
nodes:
- role: control-plane
- role: worker
- role: worker- Create the cluster
-
Use the
kind create clustercommand with the--configflag to create the cluster using your configuration file:
kind create cluster --config kind-multinode.yaml- Check the cluster status
-
After the cluster is created, you can check its status using
kubectl:
kubectl cluster-infoWhen you’re done with the cluster, you can delete it using the kind delete cluster command:
kind delete clusterThere are many options you can get your hands on a managed Kubernetes cluster provided by a cloud Provider. Most of the offerings also have a time-limited tier to try things out. This would be an ideal playground for the more advanced patterns which involve multiple nodes and can not be easily tested on a single node cluster like Minikube.
Here’s a list of Kubernetes cloud offerings, which is very likely not complete [1]:
- Google Kubernetes Engine (GKE)
-
For new users Google provides a free budget for a certain time period so that you can easily try things out.
- Azure Kubernetes Service (AKS)
-
Managed Kubernetes Cluster by Microsoft
- Elastic Container Service (EKS)
-
Kubernetes on Amazon Web Services (AWS)
- Kubernetes on Digital Ocean
-
Managed Kubernetes cluster provided by Digital Ocean
ALternatively you can also try out these examples on these playgrounds:
- Kubernetes Playground
-
Fully pre-installed Kubernetes playground provided by Katacoda (you can also use
kinstead ofkubectlin the terminal). - Play with Kubernetes
-
Playground for setting up a Kubernetes cluster, provided by Tutorius. More about how to learn installing Kubernetes, but then can be used for our examples, too.
For a quick setup experience we recommend Katacoda’s Kubernetes Playground.
It provides a two node cluster, so we can try out 'real' cluster examples here which are not possible with Minikube.
If you try out the samples with this playground, we recommend to checkout these examples with git clone https://github.com/k8spatterns/examples.git before starting.
Kubectl is the CLI client used to access any Kubernetes cluster.
Installation instructions for kubectl can be found on the Kubernetes site.
Pick the variant matching your operating system and put it into your execution path.
Most of the time you don’t have to update kubectl when you update your cluster as older kubectl version typically also work nicely with newer clusters.
We recommend to set a shell alias like k to shorten kubectl in the command line because you have to type it quite often.
Other tools which are used in the examples:
Some other handy aliases and functions:
# Shorten kubectl to k
alias k=kubectl
# Get a pod by only providing a partial name
function pod {
kubectl get pod -o name | grep -v "Terminating" | grep $1 | sed -e "s/^pods\///"
}
# Get the nodePort of first port mapping of a service
function svc_node_port {
kubectl get svc $1 -o jsonpath={.spec.ports[0].nodePort}
}
# Change the current context (e.g. 'k8s_context minikube')
function k8s_context {
kubectl config use-context $1
}
# Watch periodically all pods
function k8s_watch {
watch kubectl get pods
}
# Add command line completion for your shell. Replace 'zsh' with your shell.
source <(kubectl completion zsh)Minishift is the OpenShift equivalent to Minikube.
It can be downloaded from the [Minishift release page]
Minishift is very similar to Minikube and shares a similar set of commands.
For example, to startup Minishift just use minishift start, too.
Nearly every Minikube option is available for Minishift, too.
There are some additional commands which you can leverage:
- minishift oc-env
-
This command can be evale to set the PATH to an
ocbinary, which is OpenShift equivalent tokubectl - minishift console
-
Open the OpenShift console
- minishift openshift service
-
The equivalent to
minikube service, i.e. list services and their exposed URLs
For installing Knative on your own, you have several options. These are described directly on the Knative GitHub repository.
For a minikube based installation, just follow these instructions.
For our example of Knative build we need these installations steps:
-
Startup minikube with the appropriate options.
-
Install Istio
-
Install Knative serving
-
Install Knative build (as described [here])
In short the following commands will setup your installation. Please wait after each step that all new pods has been settled. In case of any problems, please refer to the installation instructions for Knative serving and [Knative build].
# Install minikube
minikube start --memory=8192 --cpus=4 \
--kubernetes-version=v1.11.5 \
--vm-driver=hyperkit \
--disk-size=30g \
--extra-config=apiserver.enable-admission-plugins="LimitRanger,NamespaceExists,NamespaceLifecycle,ResourceQuota,ServiceAccount,DefaultStorageClass,MutatingAdmissionWebhook"
# Install istio
# (note that there are some errors like "unable to recognize ..." at the end of the output,
# but this seems to be harmless)
curl -L https://github.com/knative/serving/releases/download/v0.3.0/istio.yaml \
| sed 's/LoadBalancer/NodePort/' \
| kubectl apply --filename -
# Label the default namespace with istio-injection=enabled.
kubectl label namespace default istio-injection=enabled
# Wait until all istio pods are up
sleep 60
kubectl get pods --namespace istio-system
# Install Knative serving
# (with similar warnings as for the Istio installations, potentially harmless)
curl -L https://github.com/knative/serving/releases/download/v0.3.0/serving.yaml \
| sed 's/LoadBalancer/NodePort/' \
| kubectl apply --filename -
# Wait until knative-serving pods are up
sleep 60
kubectl get pods --namespace knative-serving
# Install Knative build
kubectl apply --filename https://storage.googleapis.com/knative-releases/build/latest/release.yaml
# Wait until knative-build is up and running
sleep 60
kubectl get pods --namespace knative-buildAs an alternative to a Minikube installation you can also use the free offering from Instruqt for experimenting with Knative with their Knative tutorials.