Building a docker container with multiple conda envs

Question

As a novice to the technology I have only a basic knowledge regarding how to build and deploy a docker so i was hoping someone could provide some help.

The situation is the following:

I need to create a docker container with two conda environments and then I want to call a command line from my native os (linux).

Firs i created two conda envs:

env-a.yml:

conda create --name env-a
conda activate env-a
conda install -c bioconda samtools
conda env export > env-a.yml

env-b.yml:

conda create --name env-b
conda activate env-b
conda install -c bioconda samtools
conda env export > env-b.yml

Next I build my docker container

Dockerfile:

    FROM continuumio/miniconda3
     COPY ./env-a.yml /opt/
     COPY ./env-b.yml /opt/
     RUN conda env create -f /opt/env-a.yml
     RUN conda env create -f /opt/env-b.yml
     RUN echo "source activate env-a" > ~/.bashrc
     ENV PATH /opt/conda/envs/env/bin:$PATH

my build :

docker build -t test ./

and now i run it

docker run -it test
(env-a) root@1c44fcc2e83d:/#

looks good… now i want to mount a location from which my files are read and execute two pipelines each calling tools from different environment:

docker run --init -v /media/data/home/booby/test/:/opt/ test /bin/bash -c "source activate env-a && env view -h /opt/my.bam > /opt/my.sam && source activate env-b && samtools view -b /opt/my.sam > /opt/my.bam"

but when i mount a volume (-v) , it seams I lose access to my conda envs.

/bin/bash: activate: No such file or directory

what would be the proper way to set this up?

Thank you

Answer 1

Better way to do this is to define both envs and env-a and env-b inside the single dockerfile, like so:

FROM continuumio/miniconda3

WORKDIR /usr/src/app

#Create scripts to run in env-a and env-b (This is for example; pipelines in your case) 

COPY run_env_a.sh /usr/src/app/run_env_a.sh
COPY run_env_b.sh /usr/src/app/run_env_b.sh

# Make the scripts executable
RUN chmod +x /usr/src/app/run_env_a.sh /usr/src/app/run_env_b.sh

# Install samtools in env-a
RUN conda create --name env-a -c bioconda samtools

# Install samtools in env-b
RUN conda create --name env-b -c bioconda samtools

# Run the script in env-a
RUN conda run -n env-a /usr/src/app/run_env_a.sh

# Run the script in env-b
RUN conda run -n env-b /usr/src/app/run_env_b.sh



# This is optional and can be removed to use base conda env generated from base image

# Set the default shell to use env-a
SHELL ["conda", "run", "-n", "env-a", "/bin/bash", "-c"]

# Set the default command to activate env-a
CMD ["conda", "run", "-n", "env-a", "/bin/bash"]


# Set the default command to activate env-a (After this command all commands will automatically run in env-a)
CMD ["conda", "run", "-n", "env-a", "/bin/bash”]

Answer 2

You're overwriting the contents of /opt/ in the container with the contents from /media/data/home/booby/test/. Mount it in some other directory - for example to a subdirectory of /opt as shown below as shown below

docker run --init -v /media/data/home/booby/test/:/opt/sam /bin/bash -c "source activate env-a && env view -h /opt/sam/my.bam > /opt/sam/my.sam && source activate env-b && samtools view -b /opt/sam/my.sam > /opt/sam/my.bam"