[ci skip] Avoid waiting on a container that died

[michaelhixson]

This is happening for me locally with mofuw. Its CMD line fails:

/bin/sh: 1: ./start-servers.sh: not found

and then the toolset waits around for a long time for it to be running.
With the change, it detects that it has failed quickly.

[michaelhixson]

By the way, the build failure with mofuw was a local-only problem related to line endings. Nothing that needs to be fixed in the repo.

[NateBrady23]

There's a potential issue here that I found last week which was causing all of those start/stop errors on citrine. We are sometimes losing a race where we tell the docker daemon to start a container in detached mode and immediately check its availability.

[michaelhixson]

@msmith-techempower Did you see Nate's comment? Do you remember what that original fix for that race was?

[msmith-techempower]

@michaelhixson Yeah, it was #3574

Basically, we found rare instances where the call to start the container daemonized would return control to the toolset and we would query the docker sdk for the running container before it was actually running. This check is unnecessary now and it was removed.

[michaelhixson]

Ok, so just to be clear, I'm not reintroducing the race that Nate fixed. That's because our original check (which Nate removed) used client.containers.list(), which only returns running containers. My new check uses client.containers.get(id_or_name), which will return the container even if it isn't running yet.

[msmith-techempower]

Ah yes, this does appear to reintroduce a similar mechanism, but with your safeguard.