OSS-Datastore

The goal of this package is to gather data from GitHub via API calls, from both v3 and v4 APIs, and then store said data for further analysis in a data warehouse.

Goals

GitHub doesn't keep information around indefinitely but for a short amount of time, usually around 14 days worth of data. Because of this, it is difficult to track information around your GitHub organizations (orgs) and repositories (repos) for planning the best place to focus your time/resources. This is exacerbated by the fact that requests to the GitHub APIs are finicky when asking for multiple large sets of data and can get your IP address blocked from making API requests for a period of time.

To get around this, the OSS-Datastore will establish a pipeline for ingesting data from the GitHub APIs, run ETL (extract/transform/load) operations, store data into Redshift using the data vault modeling technique, and establish a base set of information/data marts for end user analysis and consumption. This modeling technique moves data from multiple sources to a staging area, ETL actions store that data into the data vault, and the data is then spun out into infomarts which are where the end users actually interacts with the data. Below is a really bad diagram to demonstrate the flow of data through the system and will, from a high-level, look like the following.

Items that will be tracked include:

Stars/forks/watchers
Repo traffic/stats
Issue metrics
Pull request activity
- New pull requests
- replies or updates required
Tracking user (member and external colalborator) access levels
Security issues in a repo and its dependencies

This isn't the definitive list of information that will be tracked and will expand as the package grows. The tentative data model can be found in

Setup for CLI

You need to have pipenv installed locally

pip install --user pipenv

To activate shell:

pipenv shell

To install new runtime dependencies

pipenv install <package name>

To install new dev dependencies

pipenv install <package name> --dev

To run, ensure you have copied sample.env to .env and source it or you export the environment variables when running from the command line, fill in the config info, and run

pipenv run python datastore.py

Additional setup for AWS

In order to get things setup for running thing in AWS you will need to export your completed .env file and run the aws-cdk bootstrap.

source .env

cdk boostrap

For more information see the AWS CDK docs.

Deploying to AWS

Due to limitations in the AWS-CDK you will need to follow AWS Sectretsmanager Create a Basic Secret and set the name and secret key to 'OSS-Datastore-GitHub-Token'. All you have left to do is run the following from the root folder:

./deploy_lambda.sh

This will create and launch a Cloudformation template to build the resources we need in AWS. **** THIS WILL COST YOU MONEY ****

Notes on running in AWS

This will charge YOUR account so you should keep that in mind when running this package.

The cron scheduler is setup to run once every 24 hours and is kicked off at 00:00 GMT, configurable in oss_datastore_lambda_stack.py. This was selected to ensure the pulls don't interfere with PST working hours. The schedule can be altered in oss_datastore_lambda_stack.py

Note on the Dead Letter Queue

In order to ensure that our data is being pulled and stored from the GitHub API I track and store failed requests in a Dead Letter Queue (DLQ) for manual action. My intention is to attach monitoring to the DLQ and alarm/alert when a new item is added so an admin can investigate why the request failed. Since the reason for the failure can occur for many different reasons I track the infomration about which API version failed, the query/request that failed, and the response body/headers/status code for analysis. From there it is on the admin to take action to ensure the request is retried and the data is successfully stored.

Development tracker

License

This library is licensed under the Apache 2.0 License.

JoeMoCode/oss-datastore