GitHub provides robust support for Deployments via its API, but when it comes to understanding the performance and lifecycle of those Deployments, things can get tricky. While the GitHub CLI is incredibly useful for many tasks, it currently lacks native support for querying Deployments.
I was searching for something similar to this extension for GitHub Workflows but couldn’t find anything for Deployments. So, I turned to GitHub’s REST API to gain insights into Deployment performance.
A naive approach would be to calculate the time between a Deployment’s created_at and updated_at timestamps. After all, Deployments update their updated_at value every time they receive a new status, and we could assume that the success status marks the end of a Deployment.
However, this assumption is flawed. When a deployment receives a success status, GitHub automatically adds an inactive status to all previous non-transient, non-productive deployments with the same environment name. This behavior can be disabled, but it means that the success status isn’t always a terminal status for a deployment. The deployment might receive an inactive status much later, artificially extending the updated_at value.
To address this, we can calculate the Deployment’s duration as the time between its created_at timestamp and the creation of the success status (if any). Unfortunately, the listDeploymentStatus endpoint doesn’t support filtering by status, so we must pull all statuses for each Deployment until we find the one we’re looking for.
Steps to Gather Deployment Stats
To gather statistics for GitHub Deployments in a given repo and environment:
- List a sample of Deployments.
- For each Deployment:
- List all statuses, and filter out the
successstatus. - Calculate the duration as the difference between the Deployment’s
created_attimestamp and thesuccessstatus’s creation time.
- List all statuses, and filter out the
- Compute the count, average, min, and max duration values.
This approach involves a significant number of queries, so I built a quick NodeJS script to automate the process using Octokit. You can view the script below:
Once I validated that the plan worked, I decided to explore GitHub CLI extensions. Here’s how easy it is to create one:
$ gh extension create chuchu
✓ Created directory gh-chuchu
✓ Initialized git repository
✓ Made initial commit
✓ Set up extension scaffolding
gh-chuchu is ready for development!
Next Steps
- run 'cd gh-chuchu; gh extension install .; gh chuchu' to see your new extension in action
- run 'gh repo create' to share your extension with others
For more information on writing extensions:
https://docs.github.com/github-cli/github-cli/creating-github-cli-extensionsHowever, since I had already built the script in NodeJS, I opted to use the --precompiled option, which allows you to use any language as long as you provide a script to compile the necessary binaries.
You can find the code for the NodeJS version here. After some trial and error, I decided to port the tool to Go, which was better supported by GitHub CLI’s tooling.
Final Result
The Go version of the extension works as expected. Here’s an example:
$ gh deployment-stats -cutoff 2024-08-01T12:00:00Z foo bar main
Fetched 500 deployments for main:
... fetching deployment statuses .................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................... Done
- 182 old successful deployments: avg 181 secs, min/max: 127/664 secs
- 301 new successful deployments: avg 206 secs, min/max: 55/684 secsCheck the code at ggalmazor/gh-deployment-stats. There are several improvements I plan to implement, but for now, it’s good enough. You can install it using the command:
$ gh extension install ggalmazor/gh-deployment-stats