Skip to content
This repository has been archived by the owner on Dec 10, 2023. It is now read-only.

Latest commit

 

History

History
128 lines (83 loc) · 4.75 KB

docs.md

File metadata and controls

128 lines (83 loc) · 4.75 KB
Raw

Prerequisites

You'll need to have the following installed:

Here's a handy tutorial for Ubuntu 18:04 users. The instructions written for this project are meant for minikube. However, they can be modified to adapt to a different Kubernetes implementation.

ENVIRONMENT SETUP

1. Configure Minikube

Once your minikube is up and running, enable the following addons:

$ minikube addons enable dashboard
$ minikube addons enable ingress # optional
$ minikube addons enable ingress-dns # optional

To access dashboard, type the following command from a terminal: minikube dashboard. You'll need to open a separate terminal and execute the following:

minikube tunnel

This will allow you to easily access ClusterPort services via your web browser.

2. Deploy Prometheus, and Kube State Metrics Server

We need to be able to collect and track metrics about our application. We'll use the following:

  • Kube State metrics: This is a service that extracts various metrics about all Kubernetes API objects such as deployments, pods e.t.c. Follow these instructions to deploy Kube State Metrics server to minikube.

  • Prometheus is an open source monitoring framework that scrapes(collects) metrics data from various sources such as the Kube State Metrics server. We can use it's query language to extract the information we need to determine if our application needs to be extracted. Use these instructions to deploy Prometheus to minikube.

3. Install InfluxDb and Grafana Locally

We can deploy InfluxDb and Grafana to Kubernetes as well. However, it's faster installing them locally:

  • InfluxDB
  • Grafana

Ensure the above services are running. If you are on Linux, you can start with the command:

sudo systemctl start influxdb grafana-server

Once Grafana is installed, you'll need to add Prometheus and InfluxDB as a data source.

4. Deploy Keda

  • Keda will scale up or down our application based on the threshold we provide it. Follow these instructions to deploy Keda to minikube.

At this point, Keda isn't configured to do anything.

5. Deploy Application

Execute the following commands to download and run the project in your workspace:

git clone git@github.com:k6io/example-kubernetes-autoscaling-nodejs-api.git
cd test-node-api-k6
npm i
npm run dev

Point your browser URL to localhost:4000. Any changes you make to your source code will restart the server. You can also test the API from your command line like this:

curl http://localhost:4000/crocodile

Next, deploy the application to Minikube, your Kubernetes node. Lucky for you, I've simplified the process. Just execute this one command:

make image # build and upload docker image to minikube environment
make apply # deploy crocodile-api and expose service

If you don't have GNU make on you platform, simply execute the following commands in a new terminal:

# build and upload docker image to minikube environment
eval $(minikube docker-env)
docker image prune -f
docker build -t brandiqa/crocodile-api .

# deploy crocodile-api and expose service
kubectl apply -f deploy/crocodile-deployment.yml

Execute the command to command kubectl get pods to confirm the pod is running.

[todo screenshot]

Execute the command to command kubectl get services to confirm the service is running. To ensure the service is exposed, perform a curl operation:

curl 10.98.55.109:4000/crocodiles # replace ip address

If you get a timeout, ensure minikube tunnel is running on a separate terminal.

6. Run Load Testing without AutoScaling

In this step, will run load testing test and send metrics to influxdb. Navigate to the root of the project and execute:

# replace ip address
ENDPOINT=http://10.98.55.109:4000/crocodiles k6 run -o influxdb=http://localhost:8086/k6 performance-test.js

7. Run Load Testing WITH AutoScaling

First deploy Keda's scaledobject configuration that scales up the project if the Prometheus expression, sum(rate(node_http_requests_total[2m])) exceeds 50 HTTP requests per second.

kubectl deploy -f keda/keda-prometheus-scaledobject.yml

Next, run the performance-test.js again. You should see the number of pods gradually increase as traffic increases.