Quickstart: Deploy Service in Kubernetes

Goal

In this tutorial, we'll show how you can get quickly get a service (in source code form) deployed into Kubernetes, without requiring any server-side infrastructure. We'll then show how you can extend this to a network of services.

If you're a developer, this lets you quickly test your service exactly as it's supposed to be deployed in production without committing your code / waiting on a CD pipeline.

Install / Configure Forge

Prerequisites

Forge has been tested on Mac OS X, Fedora 25, and Ubuntu 16.04. To get started, you're going to need the following installed on your system:

  • Python 2.7
  • Docker (Mac OS X, Ubuntu, Fedora install instructions). If you're using Docker on Mac OS X, we recommend you disable the "securely store docker logins in macOS keychain" option in the preferences section (see issue #28 for details)
  • kubectl

You will also need access to the following services either locally or remotely:

  • a Kubernetes cluster (minikube is fine)
  • a Docker registry a Docker registry, such as Docker Hub or Google Container Registry

Installing Forge

Once you have the prerequisites installed, you can install Forge via curl:

curl https://s3.amazonaws.com/datawire-static-files/forge/$(curl https://s3.amazonaws.com/datawire-static-files/forge/latest.url)/forge -o /tmp/forge
chmod a+x /tmp/forge
sudo mv /tmp/forge /usr/local/bin

Configuration

Create a working directory for Forge and run forge setup to complete the installation. Setup will ask for authentication information to a Docker Registry as part of this process:

mkdir forge-quickstart
cd forge-quickstart
forge setup

Deploy a service

When deploying a service into Kubernetes, you need to provide not just code, but the actual configuration needed to run this code. Forge is a build/deployment system that builds both the code and configuration, together.

  1. We'll show Forge in action with a simple service. Clone our example service:

    git clone https://github.com/datawire/hello-forge.git
    
  2. In the example service, you'll see a service.yaml file. This file contains the basic runtime configuration for the service.

    name: hello-forge
    memory: 0.25G
    cpu: 0.25
    
  3. Normally, if you want to get a service running in Kubernetes, you need to build a Docker image, push the image to a Docker registry, write some Kubernetes YAML, and run kubectl to get the service running.

    With Forge, the deploy command will take care of everything you need to get the service running. Try it now:

    forge deploy
    
  4. Once deploy completes, you can type kubectl get services to get the IP address of the service.

    Note on minikube, use minikube service --url hello-forge instead of kubectl get services

    $ kubectl get services
    NAME         CLUSTER-IP      EXTERNAL-IP       PORT(S)        AGE
    hello-forge  10.91.248.98    XXX.XXX.XXX.XXX   80:30651/TCP   4m
    ...
    
  5. curl to the XXX.XXX.XXX.XXX IP address, and see "Hello, World!".

   $ curl XXX.XXX.XXX.XXX
   Hello World! ...
  1. You can also verify that the limits specified in the service.yaml file are in effect with kubectl describe pod XXX.

Change the service

  1. You've discovered your service is on Hacker News, and you want to bump up the memory and change your greeting. Edit the service.yaml file and change the memory to 0.5G. ProTip: if you don't specify a limit, Kubernetes will default to unlimited ... which will enable an errant service to take down your entire cluster.

    So let's change some source code and redeploy:

    sed -i -e 's/Hello World!/Hello Hacker News!!!/' hello-forge/app.py
    forge deploy
    
  2. Now you can curl and see the new message (Kubernetes may take a few seconds to rollout the new image):

    $ curl XXX.XXX.XXX.XXX
    Hello Hacker News!!! ...
    
  3. You can verify that the service does have more memory with kubectl describe pods, as above.

A network of services

  1. So now we've seen we can easily build and deploy a single service, but microservices are truly useful when you can get a whole bunch of them to work together. Using Forge we can just as easily spin up a whole network of microservices:

    git clone https://github.com/datawire/hello-forge-network.git
    forge deploy
    
  2. You can see Forge has built, pushed, and deployed the entire network of services.

    kubectl get services
    

Next steps

You've seen an example of how Forge can quickly build and deploy services to Kubernetes. Now, try setting up Forge on your own services.

Still have questions? Ask in our Gitter chatroom or file an issue on GitHub.

results matching ""

    No results matching ""