deploy-rs gives up on life after a single deployment fails

[crabdancing]

I have like 8 machines to deploy to. One of them is offline. If deploy-rs tries to push to it first, it will fail, and then print 🚀 ❌ [deploy] [ERROR] Failed to push profile: Nix copy command resulted in a bad exit code: Some(1). No machines are updated.

Looking at the arg spec, I can't see any way to get it to continue pushing to the rest of the target machines.

[jakehamilton]

I ran into this the other day. It would be great if there was a way to allow for skipping offline hosts explicitly while still failing on other deployment errors. Perhaps some opt-in flag like --online-only which skips the host if it's unreachable?

[crabdancing]

My workflow currently involves a lot of experimental changes to very different machine profiles that can break on some sets of machines but not others. That being the case, I would prefer a mode that just keeps going until it's deployed everything it could, and reverted only the changes that would make machines inaccessible. :P

Ideally we'd have flags for --skip-offline, --ignore-errors, etc, which could be combined together into a "I don't care, just update everything that you can" mode. I believe there (probably) is a flag for the latter already, but it doesn't seem to actually work in my use case (e.g., deployment gives up when the nix profile tries to update vmware & a vmware service gets needlessly restarted, which reads as a deployment error and cancels the whole deployment).

Of course, that issue is outside the scope of this thread. I should probably check if an issue is already tracking it, and if not, make one.

[notgne2]

this is intentionally modeled after my own behavior during deployments

It wasn't much of an issue initially, but there definitely needs to be a way to do this, I've had to comment things out in my home infra flake so frequently and it's very tiring.

I also do agree that if we add this, we probably need a --idgaf flag which would probably be something along the lines of --auto-rollback false --magic-rollback false --skip-checks --skip-failed, to give you the fastest and most reckless blasting of your deployments onto your machines possible.