Fixed. (It isn't indeed necessary to copy args and kwargs)
[4]
As discussed on IRC, I initially went for a single-socket approach, but later realized that the monitor and the manager needs to respond to the watchdog pings and exit calls over AMP, and hence should listen to a socket as well. It looks safer to have the broker communicate with the manager and the monitor over the same socket the watchdog uses, so if case of problems the watchdog can restart the failing service.
[6]
I didn't know about this limitation, we'll have to split large messages in chunks, as you suggest.
Thanks Jamu!
[1], [2], [3], [5]
Fixed. (It isn't indeed necessary to copy args and kwargs)
[4]
As discussed on IRC, I initially went for a single-socket approach, but later realized that the monitor and the manager needs to respond to the watchdog pings and exit calls over AMP, and hence should listen to a socket as well. It looks safer to have the broker communicate with the manager and the monitor over the same socket the watchdog uses, so if case of problems the watchdog can restart the failing service.
[6]
I didn't know about this limitation, we'll have to split large messages in chunks, as you suggest.