Branch: master
Find file History
markmandel Switch to htmltest link checker -- and fix issues
Made switch to [htmltest](https://github.com/wjdp/htmltest) because previous
link checker didn't find lots of missing images, anchors that didn't exist,
external URL checks, etc. These have all been fixed now.

This checker is _brutally_ efficient at finding issues, although because it
checks external URLs, it can not be the fastest. I've integrated the
external URL cache into `make site-test`, which, when populated, runs the
test in several seconds, which makes local testing nice.
Latest commit 7ef08a3 Feb 15, 2019
Permalink
Type Name Latest commit message Commit time
..
Failed to load latest commit information.
build-image Switch to htmltest link checker -- and fix issues Feb 15, 2019
e2e-image Restructure cloudbuild.yaml Jan 28, 2019
gke-test-cluster
grafana Monitoring improvements: Feb 4, 2019
includes Switch to htmltest link checker -- and fix issues Feb 15, 2019
local-includes remove docker run entrypoints and add local-includes in makefile. Feb 10, 2018
Makefile Remove -v from Go testing - becomes too noisy Feb 15, 2019
README.md adding Prometheus+Grafana for metrics and visualizations Jan 14, 2019
boilerplate.go.txt Header insert on gRPC code gen touched too many files Jan 12, 2018
grafana.yaml
helm.yaml Adds OpenCensus metrics integration. Dec 28, 2018
prometheus.yaml Prometheus and grafana improvements based on load testing Jan 25, 2019

README.md

Developing, Testing and Building Agones

Tooling for building and developing against Agones, with only dependencies being Make and Docker

Rather than installing all the dependencies locally, you can test and build Agones using the Docker image that is built from the Dockerfile in this directory. There is an accompanying Makefile for all the common tasks you may wish to accomplish.

Table of Contents

Building on Different Platforms

Linux

  • Install Make, either via apt install make or yum install make depending on platform.
  • Install Docker for your Linux platform.
  • (optional) Minikube will require VirtualBox and will need to be installed if you wish to develop on Minikube

Windows

Building and developing Agones requires you to use the Windows Subsystem for Linux(WSL), as this makes it easy to create a (relatively) cross platform development and build system.

  • Install WSL
  • Install Docker for Windows
  • Within WSL, Install Docker for Ubuntu
  • Follow this guide from "Configure WSL to Connect to Docker for Windows" forward for integrating the Docker on WSL with the Windows Docker installation
    • Note the binding of /c to /mnt/c (or drive of your choice) - this is very important!
  • Agones will need to be cloned somewhere on your /c (or drive of your choice) path, as that is what Docker will support mounts from
  • All interaction with Agones must be on the /c (or drive of your choice) path, otherwise Docker mounts will not work
  • Now the make commands can all be run from within your WSL shell
  • (optional) Minikube is supported via the HyperV driver - the same virtualisation platform as the Docker installation.
  • Note: If you want to dev and test with Minikube, you must run WSL as Administrator, otherwise Minikube can't control HyperV.

macOS

  • Install Make, brew install make, if it's not installed already
  • Install Docker for Mac
  • (optional) Minikube will require VirtualBox and will need to be installed if you wish to develop on Minikube

GOPATH

This project should be cloned to the directory $GOPATH/src/agones.dev/agones for when you are developing locally, and require package resolution in your IDE.

If you have a working Go environment, you can also do this through:

go get -d agones.dev/agones
cd $GOPATH/src/agones.dev/agones

This is not required if you are simply building using the make targets, and do not plan to edit the code in an IDE.

If you are not familiar with GOPATHs, you can read How to Write Go Code.

Testing and Building

Make sure you are in the build directory to start.

First, let's test all the code. To do this, run make test, which will execute all the unit tests for the codebase.

If you haven't run any of the build make targets before then this will also create the Docker based build image, and then run the tests.

Building the build image may take a few minutes to download all the dependencies, so feel free to make cup of tea or coffee at this point. ☕️

Note: If you get build errors and you followed all the instructions so far, consult the Troubleshooting section

The build image is only created the first time one of the make targets is executed, and will only rebuild if the build Dockerfile has changed.

Assuming that the tests all pass, let's go ahead an compile the code and build the Docker images that Agones consists of.

Let's compile and build everything, by running make build, this will:

  • Compile the Agones Kubernetes integration code
  • Create the Docker images that we will later push
  • Build the local development tooling for all supported OS's
  • Compile and archive the SDKs in various languages

You may note that docker images, and tar archives are tagged with a concatenation of the upcoming release number and short git hash for the current commit. This has also been set in the code itself, so that it can be seen in via log statements.

Congratulations! You have now successfully tested and built Agones!

Running a Test Google Kubernetes Engine Cluster

This will setup a test GKE cluster on Google Cloud, with firewall rules set for each of the nodes for ports 7000-8000 to be open to UDP traffic.

First step is to create a Google Cloud Project at https://console.cloud.google.com or reuse an existing one.

The build tools (by default) maintain configuration for gcloud within the build folder, so as to keep everything separate (see below for overwriting these config locations). Therefore, once the project has been created, we will need to authenticate out gcloud tooling against it. To do that run make gcloud-init and fill in the prompts as directed.

Once authenticated, to create the test cluster, run make gcloud-test-cluster, which will use the deployment template found in the gke-test-cluster directory.

You can customize GKE cluster via environment variables or by using a local-includes file. See the table below for available customizations :

Parameter Description Default
GCP_CLUSTER_NAME The name of the cluster test-cluster
GCP_CLUSTER_ZONE The name of the Google Compute Engine zone in which the cluster will resides. us-west1-c
GCP_CLUSTER_LEGACYABAC Enables or disables the ABAC authorization mechanism on a cluster. false
GCP_CLUSTER_NODEPOOL_INITIALNODECOUNT The number of nodes to create in this cluster. 3
GCP_CLUSTER_NODEPOOL_MACHINETYPE The name of a Google Compute Engine machine type. n1-standard-4

If you would like to change more settings, feel free to edit the cluster.yml.jinja file before running this command.

This will take several minutes to complete, but once done you can go to the Google Cloud Platform console and see that a cluster is up and running!

To grab the kubectl authentication details for this cluster, run make gcloud-auth-cluster, which will generate the required Kubernetes security credentials for kubectl. This will be stored in ~/.kube/config by default, but can also be overwritten by setting the KUBECONFIG environment variable before running the command.

Great! Now we are setup, let's try out the development shell, and see if our kubectl is working!

Run make shell to enter the development shell. You should see a bash shell that has you as the root user. Enter kubectl get pods and press enter. You should see that you have no resources currently, but otherwise see no errors. Assuming that all works, let's exit the shell by typing exit and hitting enter, and look at building, pushing and installing Agones next.

To prepare building and pushing images, let's set the REGISTRY environment variable to point to our new project. You can choose any registry region but for this example, we'll just use gcr.io.

In your shell, run export REGISTRY=gcr.io/<YOUR-PROJECT-ID> which will overwrite the default registry settings in our Make targets. Then, to rebuild our images for this registry, we run make build again.

Before we can push the images, there is one more small step! So that we can run regular docker push commands (rather than gcloud docker -- push), we have to authenticate against the registry, which will give us a short lived key for our local docker config. To do this, run make gcloud-auth-docker, and now we have the short lived tokens.

To push our images up at this point, is simple make push and that will push up all images you just built to your project's container registry.

Now that the images are pushed, to install the development version (with all imagePolicies set to always download), run make install and Agones will install the image that you just built and pushed on the test cluster you created at the beginning of this section. (if you want to see the resulting installation yaml, you can find it in build/.install.yaml)

Finally to run end-to-end tests against your development version previously installed in your test cluster run make test-e2e, this will validate the whole application flow (from start to finish). If you're curious about how they work head to tests/e2e

Running a Test Minikube cluster

This will setup a Minikube cluster, running on an agones profile,

Because Minikube runs on a virtualisation layer on the host, some of the standard build and development Make targets need to be replaced by Minikube specific targets.

First, install Minikube, which may also require you to install a virtualisation solution, such as VirtualBox as well. Check the Building on Different Platforms above for details on what virtualisation solution to use.

Next we will create the Agones Minikube cluster. Run make minikube-test-cluster to create the agones profile, and a Kubernetes cluster of the supported version under this profile.

This will also install the kubectl authentication credentials in ~/.kube, and set the kubectl context to agones.

Great! Now we are setup, let's try out the development shell, and see if our kubectl is working!

Run make minikube-shell to enter the development shell. You should see a bash shell that has you as the root user. Enter kubectl get pods and press enter. You should see that you have no resources currently, but otherwise see no errors. Assuming that all works, let's exit the shell by typing exit and hitting enter, and look at a building, pushing and installing Agones on Minikube next.

You may remember in the first part of this walkthrough, we ran make build, which created all the images and binaries we needed to work with Agones locally. We can push these images them straight into Minikube very easily!

Run make minikube-push which will send all of Agones's docker images from your local Docker into the Agones Minikube instance.

Now that the images are pushed, to install the development version, run make minikube-install and Agones will install the images that you built and pushed to the Agones Minikube instance (if you want to see the resulting installation yaml, you can find it in build/.install.yaml).

It's worth noting that Minikube does let you reuse its Docker daemon, and build directly on Minikube, but in this case this approach is far simpler, and makes cross-platform support for the build system much easier.

If you find you also want to push your own images into Minikube, the convenience make target make minikube-transfer-image can be run with the TAG argument specifying the tag of the Docker image you wish to transfer into Minikube.

For example:

$ make minikube-transfer-image TAG=myimage:0.1

Running end-to-end tests on minikube is done via the make minikube-test-e2e target. This target use the same make test-e2e but also setup some prerequisites for use with a minikube cluster.

Running a Test Kind cluster

This will setup a Kubernetes IN Docker cluster named agones by default.

Because Kind runs on a docker on the host, some of the standard build and development Make targets need to be replaced by kind specific targets.

If you have go and docker installed go get sigs.k8s.io/kind is all you need !

Next we will create the Agones Kind cluster. Run make kind-test-cluster to create the agones Kubernetes cluster.

This will also setup helm and a kubeconfig, the kubeconfig location can found using the following command kind get kubeconfig-path --name=agones assuming you're using the default KIND_PROFILE.

You can verify that your new cluster information by running (if you don't have kubectl you can skip to the shell section):

KUBECONFIG=$(kind get kubeconfig-path --name=agones) kubectl cluster-info

Great! Now we are setup, we also provide a development shell with a handful set of tools like kubectl and helm.

Run make kind-shell to enter the development shell. You should see a bash shell that has you as the root user. Enter kubectl get pods and press enter. You should see that you have no resources currently, but otherwise see no errors. Assuming that all works, let's exit the shell by typing exit and hitting enter, and look at a building, pushing and installing Agones on Kind next.

You may remember in the first part of this walkthrough, we ran make build, which created all the images and binaries we needed to work with Agones locally. We can push these images them straight into kind very easily!

Run make kind-push which will send all of Agones's docker images from your local Docker into the Agones Kind container.

Now that the images are pushed, to install the development version, run make kind-install and Agones will install the images that you built and pushed to the Agones Kind cluster.

Running end-to-end tests on Kind is done via the make kind-test-e2e target. This target use the same make test-e2e but also setup some prerequisites for use with a Kind cluster.

If you are having performance issues, check out these docs here

Running a Custom Test Environment

This section is addressed to developers using a custom Kubernetes provider, a custom image repository and/or multiple test clusters.

Prerequisites:

  • a(some) running k8s cluster(s)
  • Have kubeconfig file(s) ready
  • docker must be logged into the image repository you're going to use

To begin, you need to set up the following environment variables:

  • KUBECONFIG should point to the kubeconfig file used to access the cluster; if unset, it defaults to ~/.kube/config
  • REGISTRY should point to your image repository of your choice (i.e. gcr.io/)
  • IMAGE_PULL_SECRET must contain the name of the secret required to pull the Agones images, in case you're using a custom repository; if unset, no pull secret will be used
  • IMAGE_PULL_SECRET_FILE must be initialized to the full path of the file containing the secret for pulling the Agones images, in case of a custom image repository; if set, make install will install this secret in both the agones-system (for pulling the controller image) and default (for pulling the sdk image) repositories

The second step is to prepare your cluster for the Agones deployments. Run make setup-custom-test-cluster to install helm in it.

Now you're ready to begin the development/test cycle:

  • make build will build Agones
  • make test will run local tests
  • make push will push the Agones images to your image repository
  • make test-e2e will run end-to-end tests in your cluster
  • make install will install/upgrade Agones into your cluster

You can combine some of the above steps into a single one, for example make build push install or make build push test-e2e.

If you need to clean-up your cluster, you can use make uninstall to remove Agones and make clean-custom-test-cluster to reset helm.

Next Steps

Have a look in the examples folder to see examples of running Game Servers on Agones.

Make Variable Reference

VERSION

The version of this build. Version defaults to the short hash of the latest commit

REGISTRY

The registry that is being used to store docker images. Defaults to gcr.io/agones-images - the release + CI registry.

KUBECONFIG

The Kubernetes config file used to access the cluster. Defaults to ~/.kube/config - the file used by default by kubectl.

CLUSTER_NAME

The (gcloud) test cluster that is being worked against. Defaults to test-cluster

IMAGE_PULL_SECRET

The name of the secret required to pull the Agones images, if needed. If unset, no pull secret will be used.

IMAGE_PULL_SECRET_FILE

The full path of the file containing the secret for pulling the Agones images, in case it's needed.

If set, make install will install this secret in both the agones-system (for pulling the controller image) and default (for pulling the sdk image) repositories.

Make Target Reference

All targets will create the build image if it is not present.

Development Targets

Targets for developing with the build image

make build

Build all the images required for Agones, as well as the SDKs

make build-images

Build all the images required for Agones

make build-sdks

Build all the sdks required for Agones

make build-sdk-cpp

Build the cpp sdk static and dynamic libraries (linux libraries only)

make test

Run the linter and tests

make push

Pushes all built images up to the $(REGISTRY)

make install

Installs the current development version of Agones into the Kubernetes cluster

make uninstall

Removes Agones from the Kubernetes cluster

make test-e2e

Runs end-to-end tests on the previously installed version of Agones. These tests validate Agones flow from start to finish.

It uses the KUBECONFIG to target a Kubernetes cluster.

See make minikube-test-e2e to run end-to-end tests on minikube.

make setup-prometheus

Install Prometheus server using stable/prometheus chart into the current cluster.

By default all exporters and alertmanager is disabled.

You can use this to collect Agones Metrics.

See make minikube-setup-prometheus and make kind-setup-prometheus to run the installation on Minikube or Kind.

make setup-grafana

Install Gafrana server using stable/grafana chart into the current cluster and setup Agones dashboards with Prometheus datasource.

You can set your own password using the PASSWORD environement variable.

See make minikube-setup-grafana and make kind-setup-grafana to run the installation on Minikube or Kind.

make prometheus-portforward

Sets up port forwarding to the Prometheus deployment (port 9090 Prometheus UI).

On Windows and MacOS you will need to have installed kubectl.

See make minikube-prometheus-portforward and make kind-prometheus-portforward to run on Minikube or Kind.

make grafana-portforward

Sets up port forwarding to the grafana deployment (port 3000 UI).

On Windows and MacOS you will need to have installed kubectl.

See make minikube-grafana-portforward and make kind-grafana-portforward to run on Minikube or Kind.

make controller-portforward

Sets up port forwarding to a specified PORT var (defaults to 8080 for controller metrics) to the controller deployment.

make pprof-web

Start the web interface for pprof.

make shell

Run a bash shell with the developer tools (go tooling, kubectl, etc) and source code in it.

make godoc

Run a container with godoc (search index enabled)

make build-controller-image

Compile the gameserver controller and then build the docker image

make build-agones-sdk-image

Compile the gameserver sidecar and then build the docker image

make build-ping-image

Compile the ping binary and then build the docker image

make gen-install

Generate the /install/yaml/install.yaml from the Helm template

make gen-crd-client

Generate the Custom Resource Definition client(s)

make gen-gameservers-sdk-grpc

Generate the SDK gRPC server and client code

Build Image Targets

Targets for building the build image

make clean-config

Cleans the kubernetes and gcloud configurations

make clean-build-image

Deletes the local build docker image

make build-build-image

Creates the build docker image

Google Cloud Platform

A set of utilities for setting up a Kubernetes Engine cluster on Google Cloud Platform, since it's an easy way to get a test cluster working with Kubernetes.

make gcloud-init

Initialise the gcloud login and project configuration, if you are working with GCP

make gcloud-test-cluster

Creates and authenticates a small, 3 node GKE cluster to work against

make gcloud-auth-cluster

Pulls down authentication information for kubectl against a cluster, name can be specified through CLUSTER_NAME (defaults to 'test-cluster')

make gcloud-auth-docker

Creates a short lived access to Google Cloud container repositories, so that you are able to call docker push directly. Useful when used in combination with make push command.

Minikube

A set of utilities for setting up and running a Minikube instance, for local development.

Since Minikube runs locally, there are some targets that need to be used instead of the standard ones above.

make minikube-test-cluster

Switches to an "agones" profile, and starts a kubernetes cluster of the right version.

Use MINIKUBE_DRIVER variable to change the VM driver (defaults virtualbox for Linux and macOS, hyperv for windows) if you so desire.

make minikube-push

Push the local Agones Docker images that have already been built via make build or make build-images into the "agones" minikube instance.

make minikube-install

Installs the current development version of Agones into the Kubernetes cluster. Use this instead of make install, as it disables PullAlways on the install.yaml

make minikube-setup-prometheus

Installs prometheus metric backend into the Kubernetes cluster. Use this instead of make setup-prometheus, as it disables Persistent Volume Claim.

make minikube-setup-grafana

Installs grafana into the Kubernetes cluster. Use this instead of make setup-grafana, as it disables Persistent Volume Claim.

make minikube-prometheus-portforward

The minikube version of make prometheus-portforward to setup port forwarding to the prometheus deployment.

make minikube-grafana-portforward

The minikube version of make grafana-portforward to setup port forwarding to the grafana deployment.

make minikube-test-e2e

Runs end-to-end tests on the previously installed version of Agones. These tests validate Agones flow from start to finish.

make minikube-shell

Connecting to Minikube requires so enhanced permissions, so use this target instead of make shell to start an interactive shell for development on Minikube.

make minikube-transfer-image

Convenience target for transferring images into minikube. Use TAG to specify the image to transfer into minikube

make minikube-controller-portforward

The minikube version of make controller-portforward to setup port forwarding to the controller deployment.

Kind

Kind - kubernetes in docker is a tool for running local Kubernetes clusters using Docker container "nodes". Kind is primarily designed for testing Kubernetes 1.11+, initially targeting the conformance tests.

Since Kind runs locally, there are some targets that need to be used instead of the standard ones above.

make kind-test-cluster

Starts a local kubernetes cluster, you can delete it with make kind-delete-cluster

Use KIND_PROFILE variable to change the name of the cluster.

make kind-push

Push the local Agones Docker images that have already been built via make build or make build-images into the "agones" Kind cluster.

make kind-install

Installs the current development version of Agones into the Kubernetes cluster. Use this instead of make install, as it disables PullAlways on the install.yaml

make kind-setup-prometheus

Installs prometheus metric backend into the Kubernetes cluster. Use this instead of make setup-prometheus, as it disables Persistent Volume Claim.

make kind-setup-grafana

Installs grafana into the Kubernetes cluster. Use this instead of make setup-grafana, as it disables Persistent Volume Claim.

make kind-prometheus-portforward

The minikube version of make prometheus-portforward to setup port forwarding to the prometheus deployment.

make kind-grafana-portforward

The minikube version of make grafana-portforward to setup port forwarding to the grafana deployment.

make kind-test-e2e

Runs end-to-end tests on the previously installed version of Agones. These tests validate Agones flow from start to finish.

make kind-shell

Connecting to Kind requires so enhanced permissions, so use this target instead of make shell to start an interactive shell for development on Kind.

make kind-controller-portforward

The Kind version of make controller-portforward to setup port forwarding to the controller deployment.

Custom Environment

make setup-custom-test-cluster

Initializes your custom cluster for working with Agones, by installing Helm/Tiller.

make clean-custom-test-cluster

Cleans up your custom cluster by reseting Helm.

Dependencies

This project uses the dep as a dependency manager. You can see the list of dependencies here.

Troubleshooting

Frequent issues and possible solutions

$GOPATH/$GOROOT error when building in WSL

If you get this error when building Agones in WSL (make build, make test or any other related target):

       /usr/local/go/src/agones.dev/agones/cmd/controller (from $GOROOT)
       /go/src/agones.dev/agones/cmd/controller (from $GOPATH)
  • Are your project files on a different folder than C? If yes, then you should either move them on drive C or set up Docker for Windows to share your project drive as well
  • Did you set up the volume mount for Docker correctly? By default, drive C is mapped by WSL as /mnt/c, but Docker expects it as /c. You can test by executing ls /c in your linux shell. If you get an error, then follow the instructions for setting up volume mount for Docker

I want to use pprof to profile the controller.

Run make build-images GO_BUILD_TAGS=profile and this will build images with pprof enabled in the controller, which you can then push and install on your cluster.

To get the pprof ui working, run make controller-portforward (or minikube-controller-portforward if you are on minikube), which will setup the port forwarding to the pprof http endpoint.

Run make pprof-web, which will start the web interface. It may take a few minutes to start up, but it can be opened on http://localhost:6060/ui.