I think I stumbled across a bug. With this config:
and only net_a is available the pipeline fails without logging the actual error. It was only after careful study of the database that I discovered this entry in stages.stage_error:
Error response from daemon: network net_b not found
What I would expect is to see this error propagated to either the runner or the server logs (or both would be even better).
All I got was this on the server:
DEBU manager: updating step status step.id=557 step.name=clone step.status=running DEBU manager: updating step status step.id=557 step.name=clone step.status=error
and this on the runner:
[timestamp] level=debug msg="updated stage to running" build.id=158 build.number=88 repo.id=11 [repo details snipped] [timestamp] level=debug msg="destroying the pipeline environment" build.id=158 build.number=88 repo.id=11 [ repo details snipped]
The result of a missing network is catastrophic - the pipeline fails immediately without any indication of the actual error.