`melos bootstrap` command sometimes takes very long

[nilsreichardt]

I'm not sure if this is a Melos issue or a Flutter issue, but the FlutterFire repository has sometimes the issue that the melos boostrap takes very. This makes the FlutterFire CI flaky. The chance for this is 1 - 5%. You find the logs in the following PR: firebase/flutterfire#8916 (keep in mind that the logs will be deleted by GitHub in 90 days).

• https://github.com/nilsreichardt/flutterfire/runs/6887635553?check_suite_focus=true
• https://github.com/nilsreichardt/flutterfire/runs/6887680143?check_suite_focus=true
• https://github.com/nilsreichardt/flutterfire/runs/6887693001?check_suite_focus=true
• https://github.com/nilsreichardt/flutterfire/runs/6887701442?check_suite_focus=true
• https://github.com/nilsreichardt/flutterfire/runs/6887696867?check_suite_focus=true

[blaugold]

The version of melos in FlutterFire is a bit behind. Since 1.1.0 we added support for bootstrapping packages in parallel as well as logging the output pub get if it does not finish after 10 seconds. Could you try running with the latest version of melos?

Probably best to set command/bootstrap/runPubGetInParallel to false, though, since pub has some fixes for running it concurrently which are not yet in stable.

[nilsreichardt]

Thanks, @blaugold. I will try out in the next days 👍

[nilsreichardt]

This seems to fix it 🚀 I ran the jobs in total 1000x (each job 200x) and only the analyze job failed 2x (the e2e test does not count). Unfortunately, I can't access the logs anymore, because GitHub can't load the page (the page crashes). But I think these failed runs are different issues.

[akku1906]

We prepared a working patch that stabilizes melos bootstrap in CI
(timeout + retry logic + early fail on inconsistent deps).

In our tests it reduced run time by ~70% and eliminated random flakiness.

Happy to share README + proof if there’s interest.