Add task for swarm join in provision workflow about deploykit HOT 4 CLOSED

@wfarner @fermayo - Please comment.

from deploykit.

It will be useful here to clearly delineate between the responsibilities of libmachete and swarm. If the primary objective of libmachete is to bring a machine to a state where a Docker engine is running, we should consider this out of scope. If we consider that libmachete is a more general-purpose tool that supports execution of shell commands and the like, it is in scope.

One bit of context that might help guide: is docker swarm join executed via the engine remote API (i.e. is there an engine API call to join a swarm)?

from deploykit.

I have mixed feelings about this one. I understand the separation of concerns here and the willingness to have libmachete completely ignore swarm stuff, but the fact is that those nodes deployed through libmachete will always need to join or create a swarm. IMHO libmachete's objetive should be to create a new node with a docker engine on it that either is the first one in a swarm, or has to join a swarm. Either way I think having libmachete understand this would make the UX much better.

In any case, having it as a step in the template as proposed here is a good compromise for now, keeps it clean, should be simple to implement (as the user will be specifying the IP of the master on the template).

from deploykit.

Closing this as InfraKit has evolved quite a bit since the discussion began. Swarm-related behavior will now be delegated to a plugin, which currently does support join behavior.

from deploykit.