Container.wait with timeout raises Connection Error

[sheridp]

    container = docker_client.containers.run(image = image, 
                                             command = cmd,
                                             detach = True, 
                                             auto_remove = False,
                                             )

container.wait(timeout = 10)

Without a timeout, everything works fine. With a timeout I get the following Exception:
ConnectionError(ReadTimeoutError('None: None',),)

The API says it should be raising a requests.exceptions.ReadTimeout

Name: docker
Version: 2.7.0
Summary: A Python library for the Docker Engine API.
Home-page: https://github.com/docker/docker-py
Author: Joffrey F
Author-email: joffrey@docker.com
License: Apache License 2.0
Location: /usr/local/lib/python3.6/site-packages
Requires: six, websocket-client, requests, docker-pycreds

[mpasternacki]

I am experiencing this as well. This started occuring after upgrading Docker from docker-engine 17.05 to docker-ce 18.03 on Ubuntu 16.04 LTS.

[mpasternacki]

I can see some entries in the backtrace that seem to be related to reading chunked response; might this be related to psf/requests#4402 ?

[orodbhen]

I already made this comment on #1374, but this seems to be a better place for it.

I'm still seeing this in 3.4.1:

OS: CentOS 7.5
Docker: 18.03

As before, it does not occur with the docker CLI, only the Python API.

In my case it occurs when running container.wait(). It originates with a socket.timeout exception in the urllib3 package, which sets off a cascade of other exceptions:

• socket.timeout
• urllib3.exceptions.ReadTimeoutError
• requests.exceptions.ConnectionError: UnixHTTPConnectionPool

It seems that requests is raising the wrong exception: requests.exceptions.ConnectionError instead of requests.exceptions.ReadTimeout. Shouldn't these exceptions be handled in the Docker API and converted to a public exception, rather than relying on the exception type from an underlying package?

[adnilsson]

I'm also experiencing this on 4.4.1.

I'm attempting to catch requests.exceptions.ReadTimeout as documented. However, no such error is catched. Instead, I'm presented with a trace of the same exceptions as @orodbhen

OS: Ubuntu 20.04.1 LTS
Docker: 20.10.0

[gsalzer]

As a workaround, until the low level exceptions are converted to a docker-timeout-exception, I use

import requests
try:
    result = container.wait(timeout=30)
except (requests.exceptions.ReadTimeout, requests.exceptions.ConnectionError):
    # timeout occurred
    pass

[zwimer]

I have this issue on Fedora 36 as well.

[jrab89]

I ran into this problem using version 7.0.0 of docker-py and version 3.11 of python. The docs say Container.wait raises requests.exceptions.ReadTimeout on a timeout, but it's raising requests.exceptions.ConnectionError.

[gsalzer]

@jrab89 I observed this in 2021, see my comment above, and wasn't the first one at that. It seems that the maintainers of docker-py cannot, or don't want to, change it. There must be some obstacle that prevents them from catching low-level exceptions at the right places and raising a "docker-timeout-occurred" exception instead.