httpservice-python

HTTP Service written in python3 and flask. Ready to be hosted on k8s with ingress-nginx.

Pre-requisites:

This project is tested on below configuration, you might need to do some adjustments if your configuration differs.

OS: Debian GNU/Linux 11
Docker Version: 20.10.5+ (non-root user should be able to use docker commands without sudo)
kubectl v1.23+
All commands should be run with a non-root user having sudo root access to install minikube

Project Structure

resources directory contains the application code that will be used by the Dockerfile.
metrics-server directory contains k8s manifest to install everything related to metrics-server to collect metrics data from our application.
k8s directory contains the manifests for deployment, service, ingress and hpa (Horizontal Pod Autoscaler) to be used by Kubernetes.
bootstrap.sh script to bootstrap the minikube testing environment with ingress addon enabled.
deploy.sh script to deploy the k8s manifests in one shot on minikube kubernetes cluster.

Common Instructions

To use this project, clone this repo and cd into it using below command:

git clone https://github.com/ankur512512/httpservice-python.git
cd httpservice-python

Execute all the further commands given in this README form within this directory only.

Instructions to build the application

To build the application in a container image, please issue below command:

docker build -t httpservice:latest .

You will see output like this:

Sending build context to Docker daemon  90.11kB
Step 1/6 : FROM python:3.9.13-alpine
 ---> b908778bd1b0
Step 2/6 : WORKDIR /python-docker
 ---> Using cache
 ---> aeaf472495ff
Step 3/6 : COPY resources/* ./
 ---> Using cache
 ---> c149bbd4ee05
Step 4/6 : RUN pip3 install -r requirements.txt
 ---> Using cache
 ---> 1efc3454f913
Step 5/6 : ENV FLASK_APP=httpService.py
 ---> Using cache
 ---> 6cce8dbffca7
Step 6/6 : CMD [ "python3", "-m" , "flask", "run", "--host=0.0.0.0"]
 ---> Using cache
 ---> ec53334da30f
Successfully built ec53334da30f
Successfully tagged httpservice:latest

I have further tagged and pushed this image to my personal repository () which is then being used in the Kubernetes cluster deployed in further steps.

Instructions to bootstrap the testing environment

Provided your system already satisfies the Pre-requisites requirements stated at the beginning, please issue below command to automatically download, install and start a minikube cluster and enable the ingress addon on it.

./bootstrap.sh

Please note that it will automatically wait for the ingress-nginx-controller pod to get ready, so please wait for the script to complete.

It should return something like this:

  % Total    % Received % Xferd  Average Speed   Time    Time     Time  Current
                                 Dload  Upload   Total   Spent    Left  Speed
100 69.2M  100 69.2M    0     0   218M      0 --:--:-- --:--:-- --:--:--  218M
😄  minikube v1.25.2 on Debian 11.3 (amd64)
✨  Using the docker driver based on existing profile
👍  Starting control plane node minikube in cluster minikube
🚜  Pulling base image ...
🏃  Updating the running docker "minikube" container ...
🐳  Preparing Kubernetes v1.23.3 on Docker 20.10.12 ...
    ▪ kubelet.housekeeping-interval=5m
🔎  Verifying Kubernetes components...
    ▪ Using image gcr.io/k8s-minikube/storage-provisioner:v5
🌟  Enabled addons: default-storageclass, storage-provisioner
🏄  Done! kubectl is now configured to use "minikube" cluster and "default" namespace by default
    ▪ Using image k8s.gcr.io/ingress-nginx/controller:v1.1.1
    ▪ Using image k8s.gcr.io/ingress-nginx/kube-webhook-certgen:v1.1.1
    ▪ Using image k8s.gcr.io/ingress-nginx/kube-webhook-certgen:v1.1.1
🔎  Verifying ingress addon...
🌟  The 'ingress' addon is enabled
pod/ingress-nginx-controller-cc8496874-cgxv5 condition met

Instructions to deploy the application on minikube

Please issue below command to automatically deploy the Kubernetes manifests on your minikube cluster:

./deploy.sh

You will see something similar to this:

deployment.apps/python-deployment created
horizontalpodautoscaler.autoscaling/python-hpa created
ingress.networking.k8s.io/python-ingress created
service/python-service created

 Deployment completed. Sleeping for 10s until IP is assigned to ingress resource...

 Deployment completed. Sleeping for 10s until IP is assigned to ingress resource...

 Deployment completed. Sleeping for 10s until IP is assigned to ingress resource...

 Deployment completed. Sleeping for 10s until IP is assigned to ingress resource...

 Deployment ready now. Please update your hosts file with below entry:
 
 192.168.49.2 time-hostname.info

As you can see, in the end it asked you to do a host file entry as we are not using LoadBalancer services because we are using a minikube cluster. You can either do that by yourself manually or issue below command to take care of that:

echo "`minikube ip` time-hostname.info" | sudo tee -a /etc/hosts

Test the application

To test your application from command line itself using curl, use below commands:

For timestamp:

curl time-hostname.info/timestamp

It will return the timestamp like this:

1654327983.8266492

Then, for hostname:

curl time-hostname.info/hostname

It will return hostname (in our case, pod's name) like this:

"python-deployment-7b7954c4fc-dghlf"

As of now, you will see only two replicas. But if you keep on hitting the api using some script to increase the cpu load. It will automatically scale up the replica set to a maximum of 3 using the hpa.

Load generation

If you want to generate some load to test the autoscaling feature, please open a separate terminal window and issue below command:

kubectl run load-generator  -it --rm --image=busybox:1.28 --restart=Never -- /bin/sh -c "while sleep 0.01; do wget -q -O- python-service:5000/hostname; done"

Wait for around 5-15 minutes and you will see that replicas will grow from 2 to 3 as the load increases and crosses the 50% threshold. It is normal for the hpa to show <unknown>/50% initially in the TARGETS field as it is yet to fetch the metrics from the metrics-server pod.

You can verify using below command:

kubectl get hpa

You will see results like this:

NAME         REFERENCE                      TARGETS    MINPODS   MAXPODS   REPLICAS   AGE
python-hpa   Deployment/python-deployment   133%/50%   2         3         3          74m

Stretch Goals

Metrics: CPU usage is used as a metric to scale up/down the application pods using HPA.
Image size: Image size is kept small using the alpine version of python3.

ankur512512 / httpservice-python