You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Mite has the ability to propagate changes in the ConfigManager to runners after the initial setup. In each call to Controller.request_work, a dictionary diff is computed (via ConfigManager.get_changes_for_runner) and sent over the wire; this is unpacked in the runner.
But as far as I can tell, this feature is never used. The only call to ConfigManager.set in the mite code is in __main__._create_config_manager (which is called at startup). I could not find any calls either in the id-mite-nft data repo.
The text was updated successfully, but these errors were encountered:
aecay
changed the title
Do we need the ability to synchronize config across runners beyond initial startup?
Updating config while tests run
Jul 9, 2019
It would obviously need some thought about how we wanted a live config to operate. Either providing a zmq admin channel, a http endpoint or have it refresh from env vars at regular intervals.
We might want to implement a new message type for this, rather than passing a (almost always empty) dict of config changes on every communication from the controller to the runner.
Mite has the ability to propagate changes in the
ConfigManager
to runners after the initial setup. In each call toController.request_work
, a dictionary diff is computed (viaConfigManager.get_changes_for_runner
) and sent over the wire; this is unpacked in the runner.But as far as I can tell, this feature is never used. The only call to
ConfigManager.set
in the mite code is in__main__._create_config_manager
(which is called at startup). I could not find any calls either in the id-mite-nft data repo.The text was updated successfully, but these errors were encountered: