Deploying with Kubernetes
Purpose
Deploying your Ampersand application to production can be done fast and frequent. For this purpose we are working on deployment on the Kubernetes platform. This allows you to deploy easily on your laptop (for testing), in your own data center (The minimum you need is a virtual machiine that runs kubernetes), or in the cloud (to benefit from the reliability and security provided by hosting providers). We use Kubernetes because it is open, free, well known, widely used, and is continuously being impoved by a large community. It lets you deploy fast and lets you upgrade the application without taking it offline.
Way of working
- A Kubernetes cluster is needed as the platform from which to launch and maintain the application. If it is there, that's fine. Otherwise you get one from a provider, from a data center, or you create one yourself (e.g. on your laptop). The cluster is built on top of virtual or physical machines and hosts and exposes the services that constitute the application.
- On the platform I need a registry from which Kubernetes can pull the image(s).
I tried it out for myself. I stole my inspiration from Kevin Smets' instruction. Thank you Kevin! I ran the following from my terminal (i.e. MacOS cli), just for the purpose of getting some hands-on experience...
Requirements
To make a Kubernetes cluster, I used Minikube. Minikube requires that VT-x/AMD-v virtualization is enabled in BIOS. To check that this is enabled on OSX / macOS, I ran:
sysctl -a | grep machdep.cpu.features | grep VMX
If there's output, it works!
Prerequisites
- kubectl
- docker (for Mac)
- minikube
- virtualbox
brew update && brew install kubectl && brew cask install docker minikube virtualbox
Verify
docker --version # Docker version 17.09.0-ce, build afdb6d4
docker-compose --version # docker-compose version 1.16.1, build 6d1ac21
docker-machine --version # docker-machine version 0.12.2, build 9371605
minikube version # minikube version: v0.22.3
kubectl version --client # Client Version: version.Info{Major:"1", Minor:"8", GitVersion:"v1.8.1", GitCommit:"f38e43b221d08850172a9a4ea785a86a3ffa3b3a", GitTreeState:"clean", BuildDate:"2017-10-12T00:45:05Z", GoVersion:"go1.9.1", Compiler:"gc", Platform:"darwin/amd64"}
Start
minikube start
This can take a while, expected output:
Starting local Kubernetes cluster...
Kubectl is now configured to use the cluster.
Great! You now have a running Kubernetes cluster locally. Minikube started a virtual machine for you, and a Kubernetes cluster is now running in that VM.
Check k8s
kubectl get nodes
Should output something like:
NAME STATUS ROLES AGE VERSION
minikube Ready <none> 40s v1.7.5
Within minikube, a docker platform is running. However, on my Mac I have another docker daemon running. So which docker do we want to talk to? In this case that is definitely the docker daemon that is inside minikube...
Use minikube's built-in docker daemon
eval $(minikube docker-env)
(You might add this line to .bash_profile or .zshrc or ... to use minikube's daemon by default. Or if you do not want to set this every time you open a new terminal).
If I want to talk to the docker daemon on my Mac, I have to revert back to it by running:
eval $(docker-machine env -u)
When running docker ps, it showed the following output:
CONTAINER ID IMAGE COMMAND CREATED STATUS PORTS NAMES
3835176d93bb k8s.gcr.io/k8s-dns-sidecar-amd64 "/sidecar --v=2 --lo…" 30 minutes ago Up 30 minutes k8s_sidecar_kube-dns-86f4d74b45-4b48w_kube-system_91346644-e424-11e8-bae2-080027501972_0
3c803544be8a k8s.gcr.io/kubernetes-dashboard-amd64 "/dashboard --insecu…" 30 minutes ago Up 30 minutes k8s_kubernetes-dashboard_kubernetes-dashboard-6f4cfc5d87-mvtgw_kube-system_962829f4-e424-11e8-bae2-080027501972_0
1d74c735c1df k8s.gcr.io/coredns "/coredns -conf /etc…" 30 minutes ago Up 30 minutes k8s_coredns_coredns-c4cffd6dc-xn62h_kube-system_961c845a-e424-11e8-bae2-080027501972_0
308bd132b35a gcr.io/k8s-minikube/storage-provisioner "/storage-provisioner" 31 minutes ago Up 30 minutes k8s_storage-provisioner_storage-provisioner_kube-system_963bf651-e424-11e8-bae2-080027501972_0
The list was in fact a little bit longer: it contained 22 containers.
Build, deploy and run an image on your local k8s setup
To setup a local registry, so Kubernetes can pull the image(s) from there:
docker run -d -p 5000:5000 --restart=always --name registry registry:2
To test whether there is a local registry, ask docker:
BA92-C02VP224HTDF:RAP stefjoosten$ docker ps | grep registry
97dc5ed7102d registry:2 "/entrypoint.sh /etc…" 4 days ago Up 4 days 0.0.0.0:5000->5000/tcp registry
BA92-C02VP224HTDF:RAP stefjoosten$
Build
First of, store all files (Dockerfile, my-app.yml, index.html) in this gist locally in some new (empty) directory.
You can build the Dockerfile below locally if you want to follow this guide to the letter. Store the Dockerfile locally, preferably in an empty directory and run:
docker build . --tag my-app
You should now have an image named 'my-app' locally, check by using docker images (or your own image of course). You can then publish it to your local docker registry:
docker tag my-app localhost:5000/my-app:0.1.0
Running docker images should now output the following:
REPOSITORY TAG IMAGE ID CREATED SIZE
my-app latest cc949ad8c8d3 44 seconds ago 89.3MB
localhost:5000/my-app 0.1.0 cc949ad8c8d3 44 seconds ago 89.3MB
httpd 2.4-alpine fe26194c0b94 7 days ago 89.3MB
Deploy and run from docker-compose.yml
I have a file called docker-compose.yml, which I have used to deploy RAP with docker-compose (rather than Kubernetes). I used kompose to transform that into the necessary Kubernetes .yaml files.
brew install kompose
I deployed straight from the docker-compose.yml file:
kompose up
However, I can also create the necessary Kubernetes files and deploy from kubectl:
kompose convert
This converts docker-compose.yml into the following files:
db-claim0-persistentvolumeclaim.yaml
db-deployment.yaml
phptools-deployment.yaml
phptools-service.yaml
rap3-claim0-persistentvolumeclaim.yaml
rap3-claim1-persistentvolumeclaim.yaml
rap3-deployment.yaml
rap3-service.yaml
Having run kompose up to deploy the three RAP services, I now have the following situation on my minikube:
BA92-C02VP224HTDF:RAP stefjoosten$ kubectl get all
NAME READY STATUS RESTARTS AGE
pod/db-56c6b7b6cc-442k9 1/1 Running 1 3d
pod/phptools-555cc5db49-w7z6k 1/1 Running 0 3d
pod/rap3-668d94bb95-cp8vs 1/1 Running 0 3d
NAME TYPE CLUSTER-IP EXTERNAL-IP PORT(S) AGE
service/kubernetes ClusterIP 10.96.0.1 <none> 443/TCP 4d
service/phptools ClusterIP 10.107.212.219 <none> 8080/TCP 3d
service/rap3 ClusterIP 10.101.81.211 <none> 80/TCP 3d
NAME DESIRED CURRENT UP-TO-DATE AVAILABLE AGE
deployment.apps/db 1 1 1 1 3d
deployment.apps/phptools 1 1 1 1 3d
deployment.apps/rap3 1 1 1 1 3d
NAME DESIRED CURRENT READY AGE
replicaset.apps/db-56c6b7b6cc 1 1 1 3d
replicaset.apps/phptools-555cc5db49 1 1 1 3d
replicaset.apps/rap3-668d94bb95 1 1 1 3d
(I got even more information with the command kubectl get all --output=wide.)
So we see three pods (containers) running. Each of the pods is wrapped in a service of its own and exposed internally on the local network of this cluster. (In Kubernetes, nodes, pods and services all have their own IPs. In many cases, the node IPs, pod IPs, and some service IPs on a cluster will not be routable, so they will not be reachable from a machine outside the cluster, such as your desktop machine.)
To access these services from a browser, we now need to expose these services to localhost. I did this by exposing both services from the minikube platform:
BA92-C02VP224HTDF:RAP stefjoosten$ minikube service rap3
Opening kubernetes service default/rap3 in default browser...
BA92-C02VP224HTDF:RAP stefjoosten$ minikube service phptools
Opening kubernetes service default/phptools in default browser...
Deploy and run from Kubernetes yml files
Store the file below my-app.yml on your system and run the following:
kubectl create -f my-app.yml
You should now see your pod and your service:
kubectl get all
The configuration exposes my-app outside of the cluster, you can get the address to access it by running:
minikube service my-app --url
This should give an output like http://192.168.99.100:30304 (the port will most likely differ). Go there with your favorite browser, you should see "Hello world!". You just accessed your application from outside of your local Kubernetes cluster!
Kubernetes GUI
minikube dashboard
Delete deployment of my-app
kubectl delete deploy my-app
kubectl delete service my-app
You're now good to go and deploy other images!
Reset everything
minikube stop;
minikube delete;
rm -rf ~/.minikube .kube;
brew uninstall kubectl;
brew cask uninstall docker virtualbox minikube;
Version
Last tested on 2017 October 20th macOS Sierra 10.12.6