Using pip in Docker container leads to ReadTimeoutError - docker

Question

I try to install a package via pip. However, every usage of pip, which needs an Internet connection (even the upgrade below) leads to a ReadTimeoutError. My basic Dockerfile which is working on another system is as follows:
FROM python:3-alpine
RUN wget google.com
RUN pip3 -V
RUN pip3 install --upgrade pip
Line two shows shows that I have an Internet connection. Output:
Connecting to google.com (216.58.206.14:80)
Connecting to www.google.com (108.177.126.103:80)
index.html 100% |*******************************| 10582 0:00:00 ETA
Line three shows that pip is installed. Output:
pip 10.0.1 from /usr/local/lib/python3.6/site-packages/pip (python 3.6)).
However, line four leads to:
Retrying (Retry(total=4, connect=None, read=None, redirect=None, status=None)) after connection broken by 'ReadTimeoutError("HTTPSConnectionPool(host='pypi.org', port=443): Read timed out. (read timeout=100.0)",)': /simple/pip/
I already tried to:
reinstall Docker
increase the default timeout with "--default-timeout=100" (which is why the read timeout is 100 in above's error message.)
I read that there are problems with pip when you are behind a Proxy, which is not the case here. Do you have any other ideas what is wrong here?
Thanks in advance!

Answer 1

There are two possible solutions:
This may not be a problem with network:
I just found a solution in another place:
Please add two lines after From layer
ENV http_proxy http://proxy-chain.xxx.com:911/
ENV https_proxy http://proxy-chain.xxx.com:912/
Or change to another mirror source:
Add the following commands before pip install
RUN mkdir ~/.pip && \
cd ~/.pip/ && \
echo "[global] \ntrusted-host = pypi.douban.com \nindex-url = http://pypi.douban.com/simple" > pip.conf
If it is because of the network, cuz it takes time to connect the source maybe:
Try to add the following limitation after pip install:
--default-timeout=1000 --no-cache-dir

Answer 2

This may also be caused when your host's network interface uses a smaller than default MTU (which is usually 1500), and the docker container does not know about this and uses a larger MTU.
We faced this on a Gitlab runner that was connected to a VLAN (MTU of 1400). Connections to some hosts worked just fine but some gave persistent issues (amongst which: PyPI, resulting in the error mentioned here).
The solution was to change the MTU for docker and all problems went away.
/etc/docker/daemon.json
{
"mtu": 1400
}