Keeping master always green & Emergency deployments

[eldad87]

Description

I'm evaluating shipit for my organization and there are several points that we would like to better understand and probably customize for everyone to use.

Master always green

Problem: Once a PR is enqueued & passed its checks, breaking changes can be added to the stack's branch.
Suggestion:

• Merge all pending PRs into a release branch.
• Trigger a CI process on the newly created release branch.
• If CI fails, reject all PRs.
• Otherwise, Merge the release branch into the stack`s branch.

Such approach will guarantee that our stack branch is always green and the process will minimize bottlenecks caused by long CI processes (Unit test, E2E test, Static code analysis/Linter, etc) by running it on all PRs at once.

Single/Emergency deployments

In extreme cases, developers required to deploy an individual PR due to its complexity or in an attempt to fix a critical bug on production.

Suggestion:
Developers can specify such meta-data on their PR as a comment, for example

deploy:single // Merge & deploy the given PR individually
deploy:emergency // Similar to deploy:single + Cancel all ongoing Stack processes & start immediately

I would like to get some pointers on how to get the job done or alternatives to achieve our goals
Please advise,
Thanks!

[JackTLi]

https://shopify.engineering/successfully-merging-work-1000-developers is the way this problem was solved at shopify, but requires a lot of development effort (this isn't available open-source afaik).

It seems like you're after something smaller scale. What you described should be buildable on top of shipit-engine, or https://bors.tech/ can maybe be an option that works well enough for you for less dev effort.