smrowley / koncepts

A Python Flask app for demonstrating Kubernetes concepts

Koncepts

Image is hosted on quay.io

Koncepts is an application for exercising Kubernetes concepts to newer Kubernetes users, as well as testing various configurations.

A Python Flask application that displays /etc/hostname and other info about the host as an html page. When the application is deployed in k8s, this will be the pod name and other pod info.

The app may be helpful for demos of pod deployments, scaling, deletions, etc.

The service is a Flask Python app running on gunicorn. It is using a Jinja template for the index.html. Please forgive the messy html with inline styles and such.

Deploying the app

You can quickly deploy the app in k8s with the following after cloning the repo:

kubectl apply -k kube/base

Example: Deploy with config files mounted as volume

kubectl apply -k kube/overlays/with-config-files

Example: Deploy with LoadBalancer service ingress

kubectl apply -k kube/overlays/expose-lb

Example: Deploy with Ingress configured for Traefik

kubectl apply -k kube/overlays/expose-traefik

Index page

The index URL returns an html page that provides information about the container.

Command line inspection

The app will also handle requests to any path, log the request, and provide a response.

Query Parameters

There are also a few query parameters that you can specify to will cause various status codes and messages to be given in the reponse:

• startup_delay parameter looks for a positive integer value, and represents the number of seconds after the python app has started before it will return 200 responses. Until that time, it will return 503 response codes.
  • eg. $HOSTNAME/test?startup_delay=10 will return 503 responses for 10 seconds.
  • Default value is -1, which will immediately return a 200 on startup.
• failure_delay parameter works similar to startup_delay, but will return 500 status codes after the integer value in seconds beyond the python app starting.
  • eg. $HOSTNAME/test?failure_delay=10 will return 500 responses after the first 10 seconds of the app being started.
  • Default value is -1, which will never return a 500 failure.
  • Note: this parameter is absolute, and has no effect from the value set by startup_delay.
    • Example: $HOSTNAME/test?startup_delay=10&failure_delay=10 will never return a 200 response. Only 503 prior to 10 seconds, and 500 after 10 seconds has lapsed.
• readiness_period and readiness_duration parameters can be used to set a consistent interval of readiness/unreadiness.
  • readiness_period expects a positive integer value, which sets the interval period.
  • readiness_duration also expects a positive integer, and is the amount of time at the beginning of each period the response will return a 200.
  • eg. $HOSTNAME/test?readiness_period=10&readiness_duration=5 will return a 200 for the first five seconds of every ten second interval, and a 503 for the other five seconds.
  • Note: the readiness parameters only come into effect after startup_delay and failure_delay success conditions are met. However, the start time of the intervals is absolute to the startup time of the app.

Curl loop for testing

The following is an example script that will curl the app at one second intervals and return the pod name. This can be used to demonstrate the effects of pod scaling with ReplicaSets or Deployment rollouts.

while [ true ]
do
  curl '$HOSTNAME/test?startup_delay=8&failure_delay=45&readiness_period=10&readiness_duration=5'
  sleep 1
done

koncepts's Issues

Add prometheus metrics endpoint

Add a Prometheus /metrics endpoint that exposes info on the app.

Useful info may include:

• Request counts on unique paths
• Response counts with label for http status code
• Number of thread spawns (future)
• Number of items in cache (future)

Helpful links:

• https://github.com/prometheus/client_python

Add container architecture info to the web page

It would be helpful to display architecture info on the html page for those that are running on multi-architecture clusters.
eg:

architecture: linux/amd64

Add feature for increasing number of threads

Make a request that will increase a pool of long running threads to increase cpu utilization. This will be helpful for testing cpu limits

This thread may be useful:
https://stackoverflow.com/questions/2905965/creating-threads-in-python#2906014