vibecheck

a stateless web app that leverages sentiment analysis models to plot the vibe of tweeters over time.

what?

a full-stack project using the following technologies:

role	technologies
deployment	kubernetes, cri-o, nginx ingress
backend logic	golang
web routing	gin
frontend	<to do>
external apis	vader-go

note: we intentially picked tools that we haven't used before to make this project as much of a learning experience as possible.

available routes

/api/sentiment

[get|post]: for testing the sentiment analysis model. the get request returns a html form to send a post reques, which returns the results of the sentiment analysis.

{
    "input": "i am happy"
}

{
    "input": "i am happy",
    "output": {
        "positive": 0.6491228070175439,
        "neural": 0.350877192824561,
        "negative": 0
    }
}

how to get started locally with docker

although this project was developed to run on a cluster of nodes using tools like kubernetes and the cri-o container runtime, you can get started by running a single container locally with docker. follow the instructions here to install docker for your machine.

to spin up the api server locally, use the following commands to build and run the container:

docker build -t go-gin-api .
docker run --name go-gin-api-container -p 80:5000 go-gin-api

the api server can be reached on localhost throught the standard http port 80 (see the available routes section for more details). the terminal used to spin up the container will show you live request logs. after the container is stopped, if there are no changes to the api server's source code, you can start the container again with the latter of the two above commands.

to stop the server and remove the container, use the following commands:

docker stop go-gin-api-container
docker rm go-gin-api-container

how to get started locally with minikube

note: the following was tested on Linux 5.10.0-1008-oem #9-Ubuntu SMP Tue Dec 15 14:22:38 UTC 2020 x86_64 x86_64 x86_64 GNU/Linux.

this repo was designed to have the containers deployed across a cluster of nodes. the container orchestration tool of choice is kubernetes, with the cri-o container runtime. so, to test how this applciation scales locally, we used minikube, which leverages docker, to create an environment that simulates a cluster on a single machine.

after following the instructions to install kubernetes, cri-o, docker, and minikube, start a small 2 node minikube cluster with the following command:

$ minikube start --container-runtime crio-o --cri-socket /var/run/crio/crio.sock -n 2

to access the web server inside minikube, we used the nginx ingress controller, which can be enabled in minikube with the following command (this process will vary based on the provider):

$ minikube addons enable ingress

to create the components of the web app, the images need to be built and pushed to dockerhub. only then can the kubernetes specification be used to deploy the image across the cluster. do so using the following commands:

$ docker build -t nariddh/go-gin-api:0.1 .
$ docker push nariddh/go-gin-api:0.1
$ kubectl apply -f api-deployment.yml

at this stage, the web server should be running inside minikube behind the nginx ingress. with the following command, interogate the nginx ingress to find the IP where the web server can be reached locally.

$ kubectl get ingress
NAME            CLASS    HOSTS   ADDRESS        PORTS   AGE
nginx-ingress   <none>   *       192.168.49.3   80

with the above address, you can access the web server at 192.168.49.3