Sentry Github Actions App

NOTE: If this is a project you would like us to invest in, please let us know in this issue. You can try this app by following the steps in the "Try it out" section below.

This app allows your organization to instrument Github Actions with Sentry. You can use this to get insights of what parts of your CI are slow or failing often.

It works by listening to Github workflow events via a webhook configured in a Github App. These events are then stored in Sentry as performance transactions. The best value for this project is when you have Sentry's Discover and Dashboards in order to slice the data and answer the questions you care about. In a sense, this project turns Github's CI into an app which Sentry monitors, thus, you can use many of the features you're already familiar with.

NOTE: The Discover feature is only available on either a Business or Trial Plan.

NOTE: The Dashboards feature is only available on the Team Plan or higher.

There's likely products out there that can give you insights into Github's CI, however, this may be cheaper, it saves you having to add another vendor and your developers are already familiar with using Sentry.

Examples

This screenshot shows that you can group Github Actions through various tags in Sentry's Discover:

This screenshot shows the transaction view for an individual Github Action showing each step of the job:

Features

Here's a list of benefits you will get if you use this project:

• Visualize a Github job and the breakdown per step
  • For instance, it makes it easy to see what percentage of the jobs are dedicated to set up versus running tests
• Create CI alerts
  • For instance, if your main branch has a failure rate above a certain threshold
  • Notice when on: schedule workflows fail which currently Github does not make it easy to notice
• Create widgets and dashboards showing your CI data
  • For instance you can show: slowest jobs, most failing jobs, jobs requirying re-runs, repos consuming most CI
  • Some of the main tags by which you can slice the data: workflow file, repo, branch, commit, job_status, pull_request, run_attempt

Watch this video showing the features described above:

2022-08-29.09-15-51.mp4

Try it out

Steps to follow:

• Create a Sentry project to gather your Github CI data
  • You don't need to select a platform in the wizard or any alerts
  • Find the DSN key under the settings of the project you created
• Create a private repo named .sentry
• Create a file in that repo called sentry_config.ini with these contents (adjust your DSN value)

[sentry-github-actions-app]
; DSN value of the Sentry project you created in the
dsn = https://[email protected]/foobar

• Install this Github App for all your repos
  • We only request workflow jobs and a single file permissions

NOTE: In other words, we won't be able to access any of your code.

Give us feedback in this issue.

Local development

Prerequisites:

• ngrok (for local dev)

Set up:

python3 -m venv .venv
source .venv/bin/activate
pip install wheel
pip install -r requirements.txt -r requirements-dev.txt

You can ingest a single job without webhooks by using the cli. For example:

# This is a normal URL of a job on Github
python3 cli.py https://github.com/getsentry/sentry/runs/5759197422?check_suite_focus=true
# From test fixture
python3 cli.py tests/fixtures/jobA/job.json

Steps to ingest events from a repository:

• Install ngrok, authenticate and start it up (ngrok http 5001)
  • Take note of the URL
• Create a Github webhook for the repo you want to analyze
  • Choose Workflow jobs events & make sure to choose application/json
  • Put Ngrok's URL there

Table of commands:

Command	Description
flask run -p 5001	Start the Flask app on http://localhost:5001
pre-commit install	Install pre-commit hooks
tox	Run tests in isolated environment
pytest	Run Python tests
pytest --cov=src --cov-report=html	Generate code coverage.

Sentry staff info

Google Cloud Build will automatically build a Docker image when the code merges on main. Log-in to Google Cloud Run and deploy the latest image.