If the script plugin spawns a shellscript that spins off some background subprocesses and exits, and those subprocesses do not close stdout/stderr, then the script plugin will hang forever (or at least until all the subprocesses have exited) trying to consume stdout/stderr output.
There should probably be a timeout here, limiting the time spent waiting for stdout/stderr EOF after the immediate child process exits. Otherwise it leads to very non-obvious behaviour: on the host running the task, all you can see is the celery worker sitting there doing nothing with no subprocesses, and there's no simple way to trace which other process somewhere else in the system is making the task hang.
This situation will also prevent the host from being uninstalled/deleted (as the worker uninstall task will block forever waiting for celery to finish the stuck task)
@mutability Hi, sorry but could you also "move" this to one of the Cloudify google groups like you did with the issue from the Openstack plugin?
Thank you.
If the script plugin spawns a shellscript that spins off some background subprocesses and exits, and those subprocesses do not close stdout/stderr, then the script plugin will hang forever (or at least until all the subprocesses have exited) trying to consume stdout/stderr output.
There should probably be a timeout here, limiting the time spent waiting for stdout/stderr EOF after the immediate child process exits. Otherwise it leads to very non-obvious behaviour: on the host running the task, all you can see is the celery worker sitting there doing nothing with no subprocesses, and there's no simple way to trace which other process somewhere else in the system is making the task hang.
This situation will also prevent the host from being uninstalled/deleted (as the worker uninstall task will block forever waiting for celery to finish the stuck task)