immediately stopped cluster will still start instances

[johny-b]

Cluster object is available before instances are created.
Cluster.stop() does nothing except for stopping current instances.

So if we do this:

cluster = await golem.run_service(...)
cluster.stop()

instances will still be created and another cluster.stop() will be required later to stop them.

IMHO the smallest necessary change would be to rename stop() to stop_current_instances() or something like that (but I'm not saying this is the best solution).

[mateuszsrebrny]

Leaving in backlog, but we still need to define how this should work

[mateuszsrebrny]

not sure if this is for this patch.3 - just labelling not to forget

[johny-b]

@shadeofblue Could you explain why this is in the Icebox? This seems pretty important to me.