Stopping multiple processes should not fail fast

[zmitchell]

Feature Request

When stopping more than one process, all processes that can be stopped should be stopped, regardless of whether there is an error stopping another process.

Use Case:

For the sake of discussion, assume "bad" means a process either never existed or is no longer running.
Right now if you attempt to stop multiple processes and one of them is bad, the processes that actually get stopped depend on the order in which the processes are passed to process-compose. This is a surprising behavior, and we should aim for the least surprising behavior.

Proposed Change:

When stopping multiple processes, attempt to stop all of them before handling any errors. Some of the errors a user will encounter can be caught before stopping any processes e.g. you can check whether a process exists or whether it's running before trying to stop it, but there may be some errors that can only be caught while trying to stop the process. In that case you can't "unstop" the process and undo the the user's requested operation. In that case it's best to attempt to complete as much of it as possible and inform them of the error.

Who Benefits From The Change(s)?

Everyone

Alternative Approaches

[F1bonacc1]

Hi @zmitchell,

Let's say there are 2 processes good and bad. What do you think should be the behavior in terms of error reporting:

1. What should be the exit code?
2. What should be printed as an error (if at all)? Will it scale to 20 bad processes?
3. Today process compose returns the list of the stopped processes. Should it return 2 lists (stopped, failed + reason)? Obviously, that will be an API breaking change.

[zmitchell]

I think you should not run the command at all if you can tell that bad is not a process that was in the user's config e.g. you should check the process names before shutting anything down.

[zmitchell]

That eliminates a lot of the ambiguity that prompted this issue.

Regardless, to answer your questions:

1. I would expect this to be nonzero because we couldn't complete the entirety of the user's request.

2. It depends on whether you want full or summarized output. My personal preference is to default to full output, which would look like

:green-check: Successfully stopped proc1
:green-check: Successfully stopped proc2
:red-x: Failed to stop proc3
:green-check: Successfully stopped proc4
...

If you wanted more concise output you could do:

Successfully stopped some processes but encountered failures for: proc1, proc2, ...

or

Successfully stopped some processes but encountered failures for:
- proc1
- proc2
...

I think if I have 20 bad processes, that's something I want to know about, and furthermore if 19 of them are bad for one reason and 1 of them is bad for another reason, that's also something I want to know about.

3. I don't actually see where that list of stopped processes is returned. If I have two processes sleeping1 and sleeping2, then try to stop those and a bad process, this is what I see:

24-08-05 12:00:38.292 FTL failed to stop processes [sleeping1 sleeping2 sleeping3] error="process sleeping3 is not running"

[F1bonacc1]

Added in v1.18.0

Cheers ✌🏻