Problem

Bookinfo application should be deployed to the Kubernetes cluster and tested. The solution for testing can be written on Python or Bash.

Requirements

Solution for testing should meet the listed below requirements:

Kubernetes cluster should be deployed using Minikube in the scope of test script/framework;
Kubernetes cluster should be provisioned and configured by the test script/framework;
Bookinfo application should be deployed by the test script/framework;
Bookinfo application should be tested by the test script/framework;
Test results should be collected by the test script/framework.

Solution

Task solved using Python framework placed in the current repo.

Environment requirements

As an environment, I have picked up a bare-metal server running Debian 10.

To start working please clone this project on the machine where the script will be executed. For the acceptable performance I recommend to run all this stuff on the bare-metal server with the minimum requirements:

4 Core CPU;
8 Gb RAM;
30 Gb free HDD space.

Environment preparation

Please, clone this repo into the home directory of the root user:

git clone git@github.com:SerhiyMakarenko/bookinfo-tester.git

To save your time I have prepared the bash script that will prepare the environment for running python tester script. Since the script should install additional packages and create a minikube user it requires superuser privileges. Please, run it under root:

cd tools
./environment_setup.sh

After that please login into the system using newly created minikube user and install required python dependencies:

pip3 install -r ./bookinfo_tester/requirements.txt

When dependencies would be installed, please run the tester script:

cd bookinfo_tester
./bookinfo_tester.py

Some explanations of the chosen architecture

For providing an ability to reach K8s services locally and assign an external IP address to ingress I chose MetaLB load balancer.

This requires a slight modification of the default minikube installation. And this is the reason why the folder tools appears in this repo.

Folder contains listed below files:

- tools
   |-- environment_setup.sh
   |-- metallb_configmap.yaml
    -- minikube-net.xml

About aim of the environment_setup.sh I told earler.

minikube-net.xml provides a modified network config for minikube-net KVM network that usually created by default during minikube run. I have made this to free IP addresses in range 192.168.39.200/29 from DHCP to give them to MetalLB.

metallb_configmap.yaml -- config map for K8s to assign IP addresses to Istio ingress load balancer.

What script do

bookinfo_tester.py script performs listed below set of activities in mentioned order:

K8s provisioning tasks

Start K8s cluster using minikube;
Install in K8s cluster MetalLB Load Balancer;
Initialize Helm;
Install Istio into K8s cluster;
Deploy Bookinfo application.

Bookinfo test tasks

Request Routing case: route all requests to v1 and check the results;
Request Routing case: route based on user identity and check the results;
Request Routing case: route all requests to review v3 and check the results.

What can be improved

Listed below things clear for me now:

Invoking kubectl command for applying manifests. This approach was chosen to cut corners and save development time. Unfortunately, K8s API does not provide an ability to deploy a set of resources from a single YAML manifest with a mix of resources. To implement those steps in the python-way require developing custom parser. This task beyond the scope of this work;
Clear and simple error handling during subprocess invocation.
Increasing test coverage. I cover in the script only Request Routing cases. As an improvement test range can be extended to cover route cases based on user identity, Injecting an HTTP delays fault, weight-based routing and more;
Make bookinfo_tester.py simplier and elegant.