We Keep Coding

Install local package through a Dockerfile

I have started learning Docker and I have developed a Python package (not published anywhere, it is just used internally) that installs and works fine locally (here I will call it mypackage). However, when trying to install it in a Docker container, Python in the container fails to recognise it even though during the build of the image no error was raised. The Dockerfile looks like this:
# install Ubuntu 20.04
FROM ubuntu:20.04
# update Ubuntu packages
ARG DEBIAN_FRONTEND=noninteractive
RUN apt update
RUN apt upgrade -y
RUN apt install -y apt-utils \
build-essential \
curl \
mysql-server \
libmysqlclient-dev \
libffi-dev \
libssl-dev \
libxml2-dev \
libxslt1-dev \
unzip \
zlib1g-dev
# install Python 3.9
RUN apt-get install -y software-properties-common gcc && \
add-apt-repository -y ppa:deadsnakes/ppa
RUN apt-get update && apt-get install -y python3.9 python3.9-dev python3.9-distutils python3-pip python3-apt python3.9-venv
# make symlink (overriding default Python3.8 installed with Ubuntu)
RUN rm /usr/bin/python3
RUN ln -s /usr/bin/python3.9 /usr/bin/python3
# copy package files and source code
RUN mkdir mypackage
COPY pyproject.toml setup.cfg setup.py requirements.txt ./mypackage/
COPY src mypackage/src/
# add path
ENV PACKAGE_PATH=/mypackage/
ENV PATH="$PACKAGE_PATH/:$PATH"
# install mypackage
RUN pip3 install -e ./mypackage
CMD ["python3.9", "main.py"]
So the above runs successfully, but if I run sudo docker run -it test_image bin/bash and run pip3 list, the package will not be there and a ModuleNotFoundError when running code depending on mypackage. Interestingly if I create a virtual environment by replacing this:
ENV PACKAGE_PATH=/mypackage/
ENV PATH="$PACKAGE_PATH/:$PATH"
by this:
ENV VIRTUAL_ENV=/opt/venv
RUN python3.9 -m venv $VIRTUAL_ENV
ENV PATH="$VIRTUAL_ENV/bin:$PATH"
it works. Ideally, I want to know why I need to create a virtual environment and how can I run local packages in a container without creating virtual environments.

changing python version in docker

I am trying to have this repo on docker: https://github.com/facebookresearch/detectron2/tree/main/docker
but when I want to docker compose it, I receive this error:
ERROR: Package 'detectron2' requires a different Python: 3.6.9 not in '>=3.7'
The default version of the python I am using is 3.10 but I don't know why through docker it's trying to run it on python 3.6.9.
Is there a way for me to change it to a higher version of python while running the following dockerfile?
FROM nvidia/cuda:11.1.1-cudnn8-devel-ubuntu18.04
# use an older system (18.04) to avoid opencv incompatibility (issue#3524)
ENV DEBIAN_FRONTEND noninteractive
RUN apt-get update && apt-get install -y \
python3-opencv ca-certificates python3-dev git wget sudo ninja-build
RUN ln -sv /usr/bin/python3 /usr/bin/python
# create a non-root user
ARG USER_ID=1000
RUN useradd -m --no-log-init --system --uid ${USER_ID} appuser -g sudo
RUN echo '%sudo ALL=(ALL) NOPASSWD:ALL' >> /etc/sudoers
USER appuser
WORKDIR /home/appuser
ENV PATH="/home/appuser/.local/bin:${PATH}"
RUN wget https://bootstrap.pypa.io/pip/3.6/get-pip.py && \
python3 get-pip.py --user && \
rm get-pip.py
# install dependencies
# See https://pytorch.org/ for other options if you use a different version of CUDA
RUN pip install --user tensorboard cmake # cmake from apt-get is too old
RUN pip install --user torch==1.10 torchvision==0.11.1 -f https://download.pytorch.org/whl/cu111/torch_stable.html
RUN pip install --user 'git+https://github.com/facebookresearch/fvcore'
# install detectron2
RUN git clone https://github.com/facebookresearch/detectron2 detectron2_repo
# set FORCE_CUDA because during `docker build` cuda is not accessible
ENV FORCE_CUDA="1"
# This will by default build detectron2 for all common cuda architectures and take a lot more time,
# because inside `docker build`, there is no way to tell which architecture will be used.
ARG TORCH_CUDA_ARCH_LIST="Kepler;Kepler+Tesla;Maxwell;Maxwell+Tegra;Pascal;Volta;Turing"
ENV TORCH_CUDA_ARCH_LIST="${TORCH_CUDA_ARCH_LIST}"
RUN pip install --user -e detectron2_repo
# Set a fixed model cache directory.
ENV FVCORE_CACHE="/tmp"
WORKDIR /home/appuser/detectron2_repo
# run detectron2 under user "appuser":
# wget http://images.cocodataset.org/val2017/000000439715.jpg -O input.jpg
# python3 demo/demo.py \
#--config-file configs/COCO-InstanceSegmentation/mask_rcnn_R_50_FPN_3x.yaml \
#--input input.jpg --output outputs/ \
#--opts MODEL.WEIGHTS detectron2://COCO-InstanceSegmentation/mask_rcnn_R_50_FPN_3x/137849600/model_final_f10217.pkl

You can use pyenv: https://github.com/pyenv/pyenv
Just google docker pyenv container, will give you some entries like: https://gist.github.com/jprjr/7667947
If you follow the gist you can see how it has been updated, very easy to update to latest python that pyenv support. anything since 2.2 to 3.11
Only drawback is that container becomes quite large because it holds all glibc development tools and libraries to compile cpython, but often it helps in case you need modules without wheels and need to compile because it is already there.
Below is a minimal Pyenv Dockerfile Just change the PYTHONVER or set a --build-arg to anything pythonversion pyenv support have (pyenv install -l):
FROM ubuntu:22.04
ARG MYHOME=/root
ENV MYHOME ${MYHOME}
ARG PYTHONVER=3.10.5
ENV PYTHONVER ${PYTHONVER}
ARG PYTHONNAME=base
ENV PYTHONNAME ${PYTHONNAME}
ARG DEBIAN_FRONTEND=noninteractive
RUN apt-get update && apt-get upgrade -y && \
apt-get install -y locales wget git curl zip vim apt-transport-https tzdata language-pack-nb language-pack-nb-base manpages \
build-essential libjpeg-dev libssl-dev xvfb zlib1g-dev libbz2-dev libreadline-dev libreadline6-dev libsqlite3-dev tk-dev libffi-dev libpng-dev libfreetype6-dev \
libx11-dev libxtst-dev libfontconfig1 lzma lzma-dev
RUN git clone https://github.com/pyenv/pyenv.git ${MYHOME}/.pyenv && \
git clone https://github.com/yyuu/pyenv-virtualenv.git ${MYHOME}/.pyenv/plugins/pyenv-virtualenv && \
git clone https://github.com/pyenv/pyenv-update.git ${MYHOME}/.pyenv/plugins/pyenv-update
SHELL ["/bin/bash", "-c", "-l"]
COPY ./.bash_profile /tmp/
RUN cat /tmp/.bash_profile >> ${MYHOME}/.bashrc && \
cat /tmp/.bash_profile >> ${MYHOME}/.bash_profile && \
rm -f /tmp/.bash_profile && \
source ${MYHOME}/.bash_profile && \
pyenv install ${PYTHONVER} && \
pyenv virtualenv ${PYTHONVER} ${PYTHONNAME} && \
pyenv global ${PYTHONNAME}
and the pyenv config to be saved as .bash_profile in Dockerfile directory:
# profile for pyenv
export PYENV_ROOT="$HOME/.pyenv"
export PATH="$PYENV_ROOT/bin:$PATH"
eval "$(pyenv init -)"
eval "$(pyenv init --path)"
eval "$(pyenv virtualenv-init -)"
build with:
docker build -t pyenv:3.10.5 .
Will build the image, but as said it is quite big:
docker images
REPOSITORY TAG IMAGE ID CREATED SIZE
pyenv 3.10.5 64a4b91364d4 2 minutes ago 1.04GB
very easy to test any python version only changing PYTHONVER
docker run -ti pyenv:3.10.5 /bin/bash
(base) root#968fd2178c8a:/# python --version
Python 3.10.5
(base) root#968fd2178c8a:/# which python
/root/.pyenv/shims/python
if I build with docker build -t pyenv:3.12-dev --build-arg PYTHONVER=3.12.dev . or change the PYTHONVER in the Dockerfile:
docker run -ti pyenv:3.12-dev /bin/bash
(base) root#c7245ea9f52e:/# python --version
Python 3.12.0a0

This is an open issue with facebookresearch/detectron2. The developers updated the base Python requirement from 3.6+ to 3.7+ with commit 5934a14 last week but didn't modify the Dockerfile.
I've created a Dockerfile based on Nvidia CUDA's CentOS8 image (rather than Ubuntu) that should work.
FROM nvidia/cuda:11.1.1-cudnn8-devel-centos8
RUN cd /etc/yum.repos.d/ && \
sed -i 's/mirrorlist/#mirrorlist/g' /etc/yum.repos.d/CentOS-* && \
sed -i 's|#baseurl=http://mirror.centos.org|baseurl=http://vault.centos.org|g' /etc/yum.repos.d/CentOS-* && \
dnf check-update; dnf install -y ca-certificates python38 python38-devel git sudo which gcc-c++ mesa-libGL && \
dnf clean all
RUN alternatives --set python /usr/bin/python3 && alternatives --install /usr/bin/pip pip /usr/bin/pip3 1
# create a non-root user
ARG USER_ID=1000
RUN useradd -m --no-log-init --system --uid ${USER_ID} appuser -g wheel
RUN echo '%wheel ALL=(ALL) NOPASSWD:ALL' >> /etc/sudoers
USER appuser
WORKDIR /home/appuser
ENV PATH="/home/appuser/.local/bin:${PATH}"
# install dependencies
# See https://pytorch.org/ for other options if you use a different version of CUDA
ARG CXX="g++"
RUN pip install --user tensorboard ninja cmake opencv-python opencv-contrib-python # cmake from apt-get is too old
RUN pip install --user torch==1.10 torchvision==0.11.1 -f https://download.pytorch.org/whl/cu111/torch_stable.html
RUN pip install --user 'git+https://github.com/facebookresearch/fvcore'
# install detectron2
RUN git clone https://github.com/facebookresearch/detectron2 detectron2_repo
# set FORCE_CUDA because during `docker build` cuda is not accessible
ENV FORCE_CUDA="1"
# This will by default build detectron2 for all common cuda architectures and take a lot more time,
# because inside `docker build`, there is no way to tell which architecture will be used.
ARG TORCH_CUDA_ARCH_LIST="Kepler;Kepler+Tesla;Maxwell;Maxwell+Tegra;Pascal;Volta;Turing"
ENV TORCH_CUDA_ARCH_LIST="${TORCH_CUDA_ARCH_LIST}"
RUN pip install --user -e detectron2_repo
# Set a fixed model cache directory.
ENV FVCORE_CACHE="/tmp"
WORKDIR /home/appuser/detectron2_repo
# run detectron2 under user "appuser":
# curl -o input.jpg http://images.cocodataset.org/val2017/000000439715.jpg
# python3 demo/demo.py \
#--config-file configs/COCO-InstanceSegmentation/mask_rcnn_R_50_FPN_3x.yaml \
#--input input.jpg --output outputs/ \
#--opts MODEL.WEIGHTS detectron2://COCO-InstanceSegmentation/mask_rcnn_R_50_FPN_3x/137849600/model_final_f10217.pkl
Alternatively, this is untested as the following images don't work on my machine (because I run arm64) so I can't debug...
In the original Dockerfile, changing your FROM line to this might resolve it, but I haven't verified this (and the image mentioned in the issue (pytorch/pytorch:1.10.0-cuda11.3-cudnn8-devel) might work as well.
FROM nvidia/cuda:11.1.1-cudnn8-devel-ubuntu20.04

Install pip with specefic python version

I am building a ubuntu docker image that is going to run my python application, and I have some libraries that require python <= 3.6 to work otherwise it will throw errors.
My problem is that when I install pip, it will always automatically use python 3.8, and I'm not sure how to let pip use an older version of python, this is the installation in my Dockerfile
RUN apt-get update && \
apt-get upgrade -y && \
apt-get install -y software-properties-common && \
add-apt-repository ppa:deadsnakes/ppa && \
apt-add-repository universe && \
apt-get update && \
apt-get install -y \
libmysqlclient-dev \
netcat \
python3 \
python-dev \
build-essential \
python3-setuptools \
python3-pip \
supervisor && \
pip install -U pip setuptools && \
rm -rf /var/lib/apt/lists/*
I tried to change python3-pip by just python-pip but when I run it it gives me the following error
E: Unable to locate package python-pip
I've tried a lot of solutions but always the same problem

Outside of Docker, if python3.6 is the python you need, you can do:
python3.6 -m pip install
In Docker right now obviously python3 is pointing to Python 3.8 so you must first install python3.6 and find out how to call it (python3.6 or python3). You might need to compile it from source and probably create some symbolic link. This can get very ugly to do inside a Docker, but you can try to write a shell script with all commands and to run the shell script inside a Docker. Or if you are lucky you may find a ready Python3.6 Docker package that works for you and apt-get install it instead of python3 the same way as you do now.

Docker how to make python 3.8 as default

I'm trying to update an existing Dockerfile to switch from python3.5 to python3.8, previously it was creating a symlink for python3.5 and pip3 like this:
RUN ln -s /usr/bin/pip3 /usr/bin/pip
RUN ln -s /usr/bin/python3 /usr/bin/python
I've updated the Dockerfile to install python3.8 from deadsnakes:ppa
apt-get install python3-pip python3.8-dev python3.8-distutils python3.8-venv
if I remove python3-pip, it complains about gcc
C compiler or Python headers are not installed on this system. Try to run: sudo apt-get install gcc python3-dev
with these installations in place I'm trying to update existing symlink creation something like this:
RUN ln -s /usr/bin/pip3 /usr/local/lib/python3.8/dist-packages/pip
RUN ln -s /usr/bin/pip /usr/local/lib/python3.8/dist-packages/pip
RUN ln -s /usr/bin/python3.8 /usr/bin/python3
it fails, saying
ln: failed to create symbolic link '/usr/bin/python3': File exists
which I assume fails because python3 points to python3.6.
if I try: RUN ln -s /usr/bin/python3.8 /usr/bin/python it doesn't complain about symlink and image gets build successfully, but fails while installing requirements later (we use Makefile targets to install dependencies inside the container using pip and pip-sync):
ERROR: Cannot uninstall 'python-apt'. It is a distutils installed project and thus we cannot accurately determine which files belong to it which would lead to only a partial uninstall.
which I assume because python-apt gets installed as part of the default python3.6 installation and python3.8 pip can't uninstall it.
PS: my Dockerfile image is based on Ubunut 18.04 which comes with python3.6 as default.
How can I properly switch Dockerfile / image from python3.5 to python3.8? so I can later use pip directly and it points to python3.8's pip

Replacing the system python in this way is usually not a good idea (as it can break operating-system-level programs which depend on those executables) -- I go over that a little bit in this video I made "why not global pip / virtualenv?"
A better way is to create a prefix and put that on the PATH earlier (this allows system executables to continue to work, but bare python / python3 / etc. will use your other executable)
in the case of deadsnakes which it seems like you're using, something like this should work:
FROM ubuntu:bionic
RUN : \
&& apt-get update \
&& DEBIAN_FRONTEND=noninteractive apt-get install -y --no-install-recommends \
software-properties-common \
&& add-apt-repository -y ppa:deadsnakes \
&& DEBIAN_FRONTEND=noninteractive apt-get install -y --no-install-recommends \
python3.8-venv \
&& apt-get clean \
&& rm -rf /var/lib/apt/lists/* \
&& :
RUN python3.8 -m venv /venv
ENV PATH=/venv/bin:$PATH
the ENV line is the key here, that puts the virtualenv on the beginning of the path
$ docker build -t test .
...
$ docker run --rm -ti test bash -c 'which python && python --version && which pip && pip --version'
/venv/bin/python
Python 3.8.5
/venv/bin/pip
pip 20.1.1 from /venv/lib/python3.8/site-packages/pip (python 3.8)
disclaimer: I'm the maintainer of deadsnakes

Why not just build a new image from ubuntu:18.04 with the desired config you need?
Like this:
FROM ubuntu:18.04
RUN apt update && apt install software-properties-common -y
RUN add-apt-repository ppa:deadsnakes/ppa && install python3.8 -y
RUN ln -s /usr/bin/pip3 /usr/bin/pip && \
ln -s /usr/bin/python3.8 /usr/bin/python

You can install and enable your python version.
# Python 3.8 and pip3
RUN apt-get update
RUN apt-get install -y software-properties-common
RUN add-apt-repository ppa:deadsnakes/ppa -y
RUN apt-get install -y python3.8
RUN ln -s /usr/bin/python3.8 /usr/bin/python
RUN apt-get install -y python3-pip

Sometimes, modifying the OS (like getting new Ubuntu clean os) is not favorable, because the current OS is too complicated. For example, my base OS is FROM ufoym/deepo:all-cu101.
So, to modify the existing python (3.6) to python 3.8, I added these 2 lines:
RUN apt-get update -qq && apt-get install -y -qq python3.8
RUN rm /usr/bin/python && rm /usr/bin/python3 && ln -s /usr/bin/python3.8 /usr/bin/python && ln -s /usr/bin/python3.8 /usr/bin/python3 \
&& rm /usr/local/bin/python && rm /usr/local/bin/python3 && ln -s /usr/bin/python3.8 /usr/local/bin/python && ln -s /usr/bin/python3.8 /usr/local/bin/python3 \
&& apt-get install -y python3-pip python-dev python3.8-dev && python3 -m pip install pip --upgrade
The first step is to install the python3.8;
The second step is to modify the softlink of python and python3 to point to python3.8
After that, install python3-pip, and update it to make sure the pip is using the current python 3.8 environment.

How to install xapian with Python 3.6 on Ubuntu 16.04?

I installed Python 3.6 on Ubuntu 16.04 on Docker using the ppa:jonathonf/python-3.6 repository. Now I'd like to install xapian so I can use it with Python. I have not found any ready-made packages, so I am trying to build it from sources. I set PYTHON3 and PYTHON3_LIB parameters to point to Python 3.6. During the build process I get the following error:
ImportError: libxapian.so.30: cannot open shared object file: No such file or directory
I tried xapian versions 1.3.7 and 1.4.5 without luck.
How can I install xapian?
Here's a Dockerfile to reproduce my error:
FROM ubuntu:16.04
RUN apt-get update \
&& apt-get install -y software-properties-common python-software-properties
RUN add-apt-repository ppa:jonathonf/python-3.6
RUN apt-get update \
&& apt-get install -y python3-pip docker.io python3.6 python3.6-dev software-properties-common \
python-software-properties build-essential wget unzip cmake python3-sphinx \
&& cd /usr/local/bin \
&& ln -s /usr/bin/python3.6 python
RUN python -m pip install --upgrade pip
# install xapian 1.4.5
RUN apt-get update && apt-get install -y curl uuid-dev zlib1g-dev
WORKDIR /root
RUN curl --silent --show-error --fail --next -O https://oligarchy.co.uk/xapian/1.4.5/xapian-core-1.4.5.tar.xz
RUN curl --silent --show-error --fail --next -O https://oligarchy.co.uk/xapian/1.4.5/xapian-bindings-1.4.5.tar.xz
RUN tar xvf xapian-core-1.4.5.tar.xz
RUN tar xvf xapian-bindings-1.4.5.tar.xz
WORKDIR /root/xapian-core-1.4.5
RUN ./configure && make && make install
WORKDIR /root/xapian-bindings-1.4.5
RUN ./configure PYTHON3=/usr/bin/python3.6 PYTHON3_LIB=/usr/lib/python3.6 --with-python3 && make && make install
RUN python -c "import xapian"

The problem is that the Xapian library (libxapian.so.30) is being installed into /usr/local/lib by default, but Ubuntu doesn't know that it's been put there yet. You can tell it by adding:
RUN ldconfig
after installing the core (so before you change WORKDIR to build the bindings).
There's some helpful information about ldconfig and library search paths on Ubuntu in the answers to this Unix Stackexchange question.