ticker

ticker is a simple web API that fetches upstream data regarding stocks using the stock symbol.

The referenced upstream API in this example is https://www.alphavantage.co/.

ticker
- Setup
  - Docker
  - Source
- Environment Variables
- Documentation
- Usage
- Deployment
- Resiliency
- Caveats

Setup

Docker

A Dockerfile is provided for building locally. The image is already built and available on Docker Hub via eb129/ticker:v0.1.1.

If you wish to build it yourself:

$ docker build . -t ticker:v0.1.0

You can adjust the image name and tag as you see fit, but be sure to update those when attempting to run via Docker.

Source

Alternatively, you can build the binary locally without the image and run:

$ make build
$ bin/ticker_darwin_amd64

The application runs on port 8080 - you can adjust this in main.go if you see fit.

Running this from the root of the repository will work since .env is available if you have populated it. You can optionally use the ticker_linux_amd64 build depending on your environment. If you require compatibility for arm arch, you may adjust the contents of the Makefile.

Environment Variables

The following environment variables are required via the .env file if running locally. Alternatively, the .env will be skipped if it isn't present and environment variables will be referenced directly. See the section on deployment below for additional details. All of these are required to be set.

APIKEY - The API key for the alphavantage API.
SYMBOL - The stock symbol to look up - i.e. MSFT.
NDAYS - When using the /avg endpoint, this is the number of days to look back for data.

Documentation

Swagger docs are available via http://localhost:8080/swagger/index.html when the API is running.

Usage

Once the API is available either locally or through deployment to Kubernetes, you can use the following calls to test. The endpoint will need to be adjusted depending on where you run it.

$ curl -X 'GET' \
  'http://localhost:8080/api/v1/stock' \
  -H 'accept: application/json'

This will return the full range of data available for the stock, bypassing NDAYS.

$ curl -X 'GET' \
  'http://localhost:8080/api/v1/stock/avg' \
  -H 'accept: application/json'

This will return any items greater than or equal to the NDAYS value - meaning, anything older will not be returned. This also returns the average close price over that same time period.

Deployment

Kustomize manifests are provided for deployment. You can switch to the deploy/kubernetes/ticker/overlays/development folder and run the following commands to work with the files.

You must have a Kubernetes cluster running and configured for the apply step. This was tested in development using minikube.

$ kubectl kustomize .
...outputs rendered manifests

$ kubectl apply -k .
...applies to cluster

Once the application is deployed, you can port-forward directly to the service if you wish to test locally:

$ kubectl port-forward -n ticker service/ticker 8080:8080
$ curl...

Note: The manifests use a secret referencing a file called .env.api in the base/ folder that will generate a Secret using secretGenerator. For the sake of this demonstration, SYMBOL and NDAYS are provided as literals for the configMapGenerator. You must provide your API key in .env.api and you may adjust the literals in the configMapGenerator as you wish.

Resiliency

A HorizontalPodAutoscaler and a PodDisruptionBudget are provided here as they are fundamental constructs for managing resiliency in a real-world application. A base health check (liveness/readiness) is also provided. These settings would need to be adjusted in production to reflect the following requirements:

Enough Pods at baseline to support realistic traffic on the HPA in terms of min/max settings.
Appropriate configuration of the PDB to guarantee the application remains available during events like rollouts, cluster upgrades, etc.
Appropriate configuration of RollingUpdate configuration on the Deployment to tolerate graceful rollouts.
Appropriate configuration of health checks.

The configuration provided here is oversimplified, but this would warrant additional consideration when deploying an application.

Caveats

In production, you'd want to use a real data store for rate limiting implementation. This uses an in-memory store so restarting the app negates the usefulness of rate limiting. The rate limiting implemenation is 5/minute - the alphavantage API also instates 500/day, but that wasn't implemented here.
Parsing of sensitive data, like the API key or anything passed in from .env, is done in the function calls to save time. In a real-world application, I'd recommend using a struct to instantiate the data so it can be reused properly.
This was tested on minikube using minikube v1.25.2 on Darwin 12.4.
I would recommend writing tests for this application in real-world usage.
Ingress configuration, in my opinion, is beyond the scope of this exercise. You may need to add annotations or adjust the manifest for the Ingress depending on your environment.
HorizontalPodAutoscaler functionality only works if metrics are available via the API. This is beyond the scope of this exercise.

echoboomer / ticker