This will cover the steps need to be done in order to build your local development environment for Kubeless.
As Kubeless project is mainly developed in the Go Programming Language, the first thing you should do is guarantee that Go is installed and all environment variables are properly set.
In this example we will use Ubuntu Linux 16.04.2 LTS as the target host on where the project will be built.
- Visit https://golang.org/dl/
- Download the most recent Go version (here we used 1.9) and unpack the file
- Check the installation process on https://golang.org/doc/install
- Set the Go environment variables
export GOROOT=/GoDir/go
export GOPATH=/GoDir/go/bin
export PATH=$GOPATH:$PATH
export KUBELESS_WORKING_DIR=$GOROOT/src/github.com/kubeless/
mkdir -p $KUBELESS_WORKING_DIR
- Visit the repo: https://github.com/kubeless/kubeless
- Click
Fork
button (top right) to establish a cloud-based fork.
cd $KUBELESS_WORKING_DIR
git clone https://github.com/<YOUR FORK>
cd $KUBELESS_WORKING_DIR/kubeless
git remote add upstream https://github.com/kubeless/kubeless.git
# Never push to upstream master
git remote set-url --push upstream no_push
# Checking your remote set correctly
git remote -v
To get all the needed tools to build and test, run:
cd $KUBELESS_WORKING_DIR/kubeless
make bootstrap
Or if you want to use a containerized environment you can use minikube. If you already have minikube use the following script to set it up:
cd $KUBELESS_WORKING_DIR/kubeless
./script/start-test-environment.sh
This will start a new minikube virtual machine and will open a bash shell in which you can build any local binary or execute the tests. Note that the Kubeless code will be mounted from outside so you can still edit your files with your favourite text editor.
To make the binaries for your platform, run:
cd $KUBELESS_WORKING_DIR/kubeless
make binary
make function-controller
This will instruct "make" to run the scripts to build the kubeless client and the kubeless controller image.
You can build kubeless for multiple platforms with:
make binary-cross
The binaries accordingly located at bundles/kubeless_$OS_$arch
folder.
Each Kubeless trigger controller is being developed on its own repository. You can find more information about those controllers in their repositories:
To regenerate the most updated k8s manifests file, run:
Note that you will need the
kubecfg
in yourPATH
in order to generate the Kubeless manifests.
cd $KUBELESS_WORKING_DIR
export KUBECFG_JPATH=$PWD/ksonnet-lib
git clone --depth=1 https://github.com/ksonnet/ksonnet-lib.git
cd $KUBELESS_WORKING_DIR/kubeless
make all-yaml
If everything is ok, you'll have generated manifests file under the $KUBELESS_WORKING_DIR
root directory:
kubeless-openshift.yaml
kubeless-non-rbac.yaml
kubeless.yaml
You can also generate them separated using the following commands:
make kubeless-openshift.yaml
make kubeless-non-rbac.yaml
make kubeless.yaml
Usually you will need to upload your controller image to a repository so you can make it available for your Kubernetes cluster, whenever it is running.
To do so, run the commands:
docker login -u=<dockerhubuser> -e=<e-mail>
docker tag kubeless-controller-manager <your-docker-hub-repo>/kubeless-test:latest
docker push <your-docker-hub-repo>/kubeless-test:latest
Make sure your image repository is correctly referenced in the "containers" session on the yaml file.
containers:
- image: fabriciosanchez/kubeless-test:latest
imagePullPolicy: Always
name: kubeless-controller
serviceAccountName: controller-acct
Hint: take a look at the imagePullPolicy
configuration if you are sending images with tags (e. g. "latest") to the Kubernetes cluster. This option controls the image caching mechanism for Kubernetes and you may encounter problems if new images enters the cluster with the same name. They might not be properly pulled for example.
In order to upload your kubeless controller image to Kubernetes, you should use kubectl as follows, informing the yaml file with the required descriptions of your deployment.
kubectl create ns kubeless
kubectl create -f <path-to-yaml-file>/kubeless.yaml
Branch from it:
git checkout -b myfeature
Then start working on your myfeature
branch.
# While on your myfeature branch
git fetch upstream
git rebase upstream/master
git commit
Likely you go back and edit/build/test some more then commit --amend
in a few cycles.
git push origin myfeature
There are several files that are automatically generated by Kubernetes code-generator based on the API specification in the repository.
These include:
- Clientset
- Listers
- Shared informers
- Deepcopy functions
If you make any changes to API specification, you will need to run make update
to regenerate clientset, informers, lister and deepcopy functions.
The simplest way to try kubeless is deploying it with minikube
You can start working with the local minikube VM and test your changes building the controller image and running your tests. Once you are happy with the result and you are ready to send a pull request you should run the unit and end-to-end tests (to spot possible issues with your changes):
make validation
make test
make build_and_test
Note that for running the end-to-end tests you need to provide a clean profile of minikube (you can create a specific profile for the tests with minikube profile tests
).
Any new feature/bug fix made to the code should be accompanied by a unit or end to end test.
- Visit your fork at https://github.com/$your_github_username/kubeless.
- Click the
Compare & pull request
button next to yourmyfeature
branch. - Make sure you fill up clearly the description, point out the particular issue your PR is mitigating, and ask for code review.
Example of shell script to setup a local environment, build the kubeless binaries and make it available on kubernetes.
#!/bin/bash
# Please set GOROOT and GOPATH appropriately before running!
#rm -rf $GOROOT/src/github.com
#export GOROOT=
#export GOPATH=
#export PATH=$GOPATH:$PATH
#KUBELESS_WORKING_DIR=$GOPATH/src/github.com/kubeless/
#mkdir -p $KUBELESS_WORKING_DIR
#cd $KUBELESS_WORKING_DIR
#git clone https://github.com/<INCLUDE HERE YOUR FORK AND UNCOMMENT>
#cd $KUBELESS_WORKING_DIR/kubeless
#git remote add upstream https://github.com/DXBrazil/kubeless
#git remote set-url --push upstream no_push
#git remote -v
# git checkout <INCLUDE HERE YOUR BRANCH AND UNCOMMENT>
#git fetch
#make binary
#make controller-image
#docker login -u=<your docker hub user> -e=<your e-mail>
#docker tag kubeless-controller <yourrepo>/<your-image>
#docker push <your repo>/<your-image>
#kubectl delete -f <path-to-yaml>
#kubectl delete namespace kubeless
#a=Terminating
#while [ $a == Terminating ]
#do
#a=`kubectl get ns | grep Termina | awk '{print $2}'`
#sleep 5
#done
#kubectl create namespace kubeless
#kubectl create -f <path-to-yaml>
We use dep to vendor the dependencies. Take a quick look at the README to understand how it works. Packages that Kubeless relies on are listed at Gopkg.toml.
Happy hacking!