-
Notifications
You must be signed in to change notification settings - Fork 232
Description
This issue requires some changes in the file aiida.engine.processes.control.py:_resolve_futures.
The timeout has not been applied on the correct future object. In PR #6793 we tried to fix it but is now just differently buggy. Now the command timeouts correctly but the message is not correct. As it mentions that the action has been cancelled while the timeout is just cancelling the waiting for the response and not the action itself. Due to the message broker design there is no way without extending the current message builder interface in plumpy to cancel the action, so we can only cancel the waiting on the response from the worker performing the action (so unblock the CLI of the user). However I am still confused why in the code, even before PR #6793, the timeout was used at two places. I dont fully understand this yet but it seems to me that sending the message and the unwrapping of the result both of a timeout. But the sending of the message has itself a timeout defined in the config.
Discussed this with @khsrali a bit
Metadata
Metadata
Assignees
Labels
Type
Projects
Status