When I first started working with open source most the tools just worked! Life was wonderful on my personal machine and various package managers were easy to use. Then the big bad corporate firewall came into play and I had no idea what I was doing anymore, tools that used to work we’re encountering SSL/TLS errors and connections weren’t happening.

If you already know what you’re doing… this will be obvious to you. If you’re struggling with setting up environments behind your firewall, hopefully you find something useful here.

I’ll go over a few modifications to base Docker images that allow connections to private (corporate) and public package repositories for Python, Node, .NET and Debian.

Link to repo files here

Python (pip)

The main package manager folks use here is pip (conda is similar). So we’ll need a pip.conf file on Linux, or a pip.ini file on Windows. For full reference click this link. For Debian, it’s nice to the proxy configs to apt as well with apt.conf. If you want to add a private repo for apt, I’d probably use add-apt-repository instead of adding the settings manually. Ok, so together we create these two files

# pip.conf
# the default index-url is https://pypi.python.org/simple
# fill in the blanks below to the location of your private repo (maybe you need to authenticate with it, [*] means optional)
index-url = https://[<username>:<password>@]<index_host>:<index_port>/<route_to_repo_index>

# if you need a proxy to access the index-url, you'll need to pass the proxy server details here
# this is equivalent to pip install --proxy http://[<username>:<password>@]<proxy_host>:<proxy_port> <package_name>
proxy = http://[<username>:<password>@]<proxy_host>:<proxy_port>

# also, if you're mucking around with some self-signed certs, don't forget to trust that host for good measure.
trusted-host = <host>
# apt.conf
Acquire::http::proxy "http://[<username>:<password>@]<proxy_host>:<proxy_port>"
Acquire::https::proxy "http://[<username>:<password>@]<proxy_host>:<proxy_port>"
# .curlrc
proxy = http://[<username>:<password>@]<proxy_host>:<proxy_port>

With this two files, we can create a python base Dockerfile so any internal info to server references can be left outside future Dockerfiles. i.e, if we host the following docker image internally



LABEL python_version=${PYTHON_VERSION}

COPY pip.conf /etc/pip.conf
COPY apt.conf /etc/apt/apt.conf
COPY .curlrc /root/.curlrc

CMD ["bash"]

I can build an app internally with docker, but also build it externally if something changes down the road and we open source the image or something.


FROM <python_internal>:${PYTHON_VERSION}

RUN apt-get update
RUN curl https://packages.microsoft.com/keys/microsoft.asc | apt-key add -
RUN pip install pandas


Don’t forget to be mindful of if you’re installing from internal apt repo’s or curling from an internal host. Then you’re not really doing a fully open source solution internally, it requires slightly more setup at the repo file level.

Node (npm)

Now-a-days most tools follow similar patterns, so all we’ll have to change for Node is the npm configs! Create a .npmrc file.

# .npmrc

# the default registry is https://registry.npmjs.org/
# fill in the blanks below to the location of your private repo (maybe you need to authenticate with it, [*] means optional)

# if you need a proxy to access the index-url, you'll need to pass the proxy server details here
# this is equivalent to npm install --proxy http://[<username>:<password>@]<proxy_host>:<proxy_port> <package_name>

# also, if you're mucking around with some self-signed certs, you'll have to ignore them or...

# you can add the certs of all the hosts you'd like to trust like follows
# ca[]="cert 1 base64 string"
# ca[]="cert 2 base64 string"

Equipped with our configs, we can create a node base image by following the same logic as before



LABEL node_version=${NODE_VERSION}

COPY .npmrc /root/.npmrc
COPY apt.conf /etc/apt/apt.conf
COPY .curlrc /root/.curlrc

CMD ["bash"]

Now you can use all the commands without worrying about the proxy!

.NET (NuGet)

Even with Microsoft, it’s the same story with .NET Core! Just create NuGet.Config with the following:

<?xml version="1.0" encoding="utf-8"?>
    <add key="dependencyVersion" value="Highest" />
    <add key="https_proxy" value="http://[<username>:<password>@]<proxy_host>:<proxy_port>" />
    <add key="http_proxy" value="http://[<username>:<password>@]<proxy_host>:<proxy_port>" />
    <add key="NuGet official package source" value="https://api.nuget.org/v3/index.json" />
    <author name="microsoft">
      <certificate fingerprint="3F9001EA83C560D712C24CF213C3D312CB3BFF51EE89435D3430BD06B5D0EECE" hashAlgorithm="SHA256" allowUntrustedRoot="false" />
    <repository name="nuget.org" serviceIndex="https://api.nuget.org/v3/index.json">
      <certificate fingerprint="0E5F38F57DC1BCC806D8494F4F90FBCEDD988B46760709CBEEC6F4219AA6157D" hashAlgorithm="SHA256" allowUntrustedRoot="false" />

and then our Dockerfile follows the same format as python and node


FROM mcr.microsoft.com/dotnet/core/sdk:${DOTNET_CORE_VERSION}

LABEL dotnet_core_version=${DOTNET_CORE_VERSION}

COPY NuGet.Config /root/.nuget/NuGet/
COPY apt.conf /etc/apt/apt.conf
COPY .curlrc /root/.curlrc

CMD ["bash"]

Now you can dotnet restore away


If you’ve made it this far, thanks for reading! Good luck with your hunt for knowledge, the corporate world is a political one! Stay Sane!