This error will also appear if you try to connect to an exposed port from within a Docker container, when nothing is actively serving the port.
On a host where nothing is listening/bound to that port you'd get a No connection could be made because the target machine actively refused it error instead when making a request to a local URL that is not served, eg: localhost:5000. However, if you start a container that binds to the port, but there is no server running inside of it actually serving the port, any requests to that port on localhost will result in:
[Errno 99] Cannot assign requested address (if called from within the container), or[Errno 0] Error (if called from outside of the container).You can reproduce this error and the behaviour described above as follows:
Start a dummy container (note: this will pull the python image if not found locally):
docker run --name serv1 -p 5000:5000 -dit python
Then for [Errno 0] Error enter a Python console on host, while for [Errno 99] Cannot assign requested address access a Python console on the container by calling:
docker exec -it -u 0 serv1 python
And then in either case call:
import urllib.request
urllib.request.urlopen('https://localhost:5000')
I concluded with treating either of these errors as equivalent to No connection could be made because the target machine actively refused it rather than trying to fix their cause - although please advise if that's a bad idea.
I've spent over a day figuring this one out, given that all resources and answers I could find on the [Errno 99] Cannot assign requested address point in the direction of binding to an occupied port, connecting to an invalid IP, sysctl conflicts, docker network issues, TIME_WAIT being incorrect, and many more things. Therefore I wanted to leave this answer here, despite not being a direct answer to the question at hand, given that it can be a common cause for the error described in this question.