Hello! Anyone having experience with the Pulumi GitHub action? I have it up and running, but encountered multiple times, that if anything goes wrong or a pipeline is aborted while Pulumi is working, all subsequent pipelines will also fail, because the Pulumis state will be locked. I had the idea to add a step that runs the "cancel" command, when a step on the pipeline fails, but I thought that maybe someone has something smarter for this problem. Thank you!

What backend are you using? This hasn't happened to me, but I've only used the Pulumi backend.

This happens to both my Minio (S3) backend, and my current setup via file://, where I mount my state dir into the Runner pod

file:// in a GitHub action? Can you guarantee that the file system is even the same? Isn't GitHub allowed use different containers for every step?

Yes, I use my own selfhosted Forgejo instance and runner, to which I can mount what I want from my host system. And I mean Pulumi does work, if ran successfully, but if I would cancel the pipeline mid pulumi up, then the generated lock file would block every subsequent run, until I run pulumi cancel.

I'd guess that this is unrelated to GitHub. The file system issues are more likely to be causing the problem. But if you just want to solve it, then maybe remove the lock file after each Pulumi step?

Well the same does happen when using S3 as my backend. But yes, I also thought of removing the lock file after a failed step. Because usually the lock file is just removed after a successful run, right?

I believe so. Honestly it's been many years since I used any backend other than Pulumi's. But yes, that's my understanding.

What is the reason not to use pulumi cloud as backend?

Money would be the usual one 🙂 Also data residency concerns.