manager: backend bug fixes
- improvement of backported
asyncio.run
in some of the previous MR's lead to a discovery of an hidden watchdog exception - manager's internal watchdog was not reporting errors properly, therefore it was completely useless (now it reports everything properly)
- subprocess monitoring was always failing because of a bug in worker ID allocation system
- ID allocation component was a mess, so I made a slight improvement
- while fixing linter errors, discovered and fixed issues with supervisord startup
@amrazek Please mainly check, that:
- if you manually stop a worker, manager will stop as well (roughly within 5 seconds)
- supervisord backend starts (at least without weird exceptions being thrown)
- you don't see any weird exceptions anywhere
Edited by Vaclav Sraier