Different behaviour running command via bash vs command line

Docker Desktop for Linux
1

I have build a docker with the follow command

docker build    \
-t bethelsight/november-bix:latest    \
--build-arg HOST=192.168.0.14    \
--build-arg PORT=8000    \
--build-arg CONDA_PACKAGE=Anaconda3-2022.10-Linux-x86_64.sh    \
--file Dockerfile .

The content of the Dockerfile is

FROM bethelsight/november-ubuntu-22

# most bioinformatics tools are installed via conda
# See this for conda activation within a Dockerfile - https://pythonspeed.com/articles/activate-conda-dockerfile/#working
ARG HOST
ARG PORT
ARG CONDA_PACKAGE

ENV ENV_HOST=$HOST
ENV ENV_PORT=$PORT
ENV ENV_CONDA_PACKAGE=$CONDA_PACKAGE

# Conda environment configuration file
COPY environment.yml /tmp

LABEL maintainer="yue.nicholas@gmail.com"

RUN apt -y update && apt install -y build-essential cmake ninja-build

RUN cd /tmp && curl -L -O http://${ENV_HOST}:${ENV_PORT}/${ENV_CONDA_PACKAGE}

# Installation
RUN sh /tmp/${ENV_CONDA_PACKAGE} -b

# Setup/initialization
# NOTE : Keep everything in a SINGLE call
RUN . /root/anaconda3/bin/activate && \
    conda init bash && \
    conda config --set auto_activate_base false && \
    conda update -y -n base -c defaults conda && \
    conda env create -f /tmp/environment.yml

# Clean up
RUN cd /tmp && rm *.sh

# Make RUN commands use the new environment:
RUN echo "export PATH=/root/anaconda3/envs/bix/bin:\$PATH" >> ~/.bashrc
RUN echo "export PATH=/usr/examples/bioinformatics/bin:\$PATH" >> ~/.bashrc
RUN echo "conda activate bix" >> ~/.bashrc

# This ensures that a shell will source ~/.bashrc since
# we are making it a "login" shell
SHELL ["/bin/bash", "--login", "-c"]

The parent docker image has a custom built (c++) command line application call cnov

cnov when run without and options simply prints out the required arguments

Allowed options:
  --version             print version string
  --plugins             list all plugins
  --help                produce help message
  --info                produce summary of project without executing target 
                        [Optional]
  --project arg         name of a project file (GML) [Mandatory].
  --inputs arg          name of an inputs definition file (JSON) [Optional].
  --outputs arg         name of an outputs definition file (JSON) [Optional].
  --maps arg            name of an globals mapping file (JSON) [Required if 
                        either inputs/outputs JSON is specified].
  --target arg          target node Name [Optional if requesting info].

When I run it via the following way

docker run \
-v /data1/nyue/temp/bix:/tmp/data     \
-v .:/tmp/json     bethelsight/november-bix cnov

I get the expected print out. This tells me that cnov is available and in the path
Note: I did check that cnov is not in my current host machine

However, when I tried it out with actual parameters that I have tested as working, it runs part way and terminated

docker run \
-v /data1/nyue/temp/bix:/tmp/data     \
-v .:/tmp/json     \
bethelsight/november-bix cnov     \
--project /tmp/data/input/rnaseq.gml     \
--inputs /tmp/json/inputs.json     \
--outputs /tmp/json/outputs.json     \
--maps /tmp/json/globals-map.json     \
--target trimmomatic_1

Whereas if I launch a bash shell

docker run \
-v /data1/nyue/temp/bix:/tmp/data     \
-v .:/tmp/json     \
bethelsight/november-bix bash

and invoke cnov inside the shell, it runs through to completion and produces the expected output data files

root@066d3644efd3:/# cnov --project /tmp/data/input/rnaseq.gml --inputs /tmp/json/inputs.json --outputs /tmp/json/outputs.json --maps /tmp/json/globals-map.json --target trimmomatic_1
Project file is /tmp/data/input/rnaseq.gml
project_file variable /tmp/data/input/rnaseq.gml
Inputs definition file is /tmp/json/inputs.json
inputs_file variable /tmp/json/inputs.json
Outputs definition file is /tmp/json/outputs.json
outputs_file variable /tmp/json/outputs.json
Globals mapping file is /tmp/json/globals-map.json
maps_file variable /tmp/json/globals-map.json
Target node is trimmomatic_1
target variable trimmomatic_1
2023-09-19 17:50:24: <info> CommandLineNode::compute() command being executed is '/root/anaconda3/envs/bix/bin/trimmomatic SE -threads 4 /tmp/data/input/CPCT12345678R_HJJLGCCXX_S1_L001_R2_001.fastq.gz /tmp/data/output/trimmed.fastq TRAILING:10 -phred33'
TrimmomaticSE: Started with arguments:
 -threads 4 /tmp/data/input/CPCT12345678R_HJJLGCCXX_S1_L001_R2_001.fastq.gz /tmp/data/output/trimmed.fastq TRAILING:10 -phred33
Input Reads: 8952 Surviving: 8952 (100.00%) Dropped: 0 (0.00%)
TrimmomaticSE: Completed successfully
2023-09-19 17:50:25: <info> CommandLineNode::compute() _stdout = ''

I am trying to determine if the way I am building docker image via my Dockerfile, I am not setting it up correctly.

Thank you in advanced.

2

Answering my own question, my solution for my particular situation with conda involved was to use ENTRYPOINT

Here is the current Dockerfile which I am using successfully, hope it helps others in the future

FROM bethelsight/november-ubuntu-22

ARG HOST
ARG PORT
ARG CONDA_PACKAGE

ENV ENV_HOST=$HOST
ENV ENV_PORT=$PORT
ENV ENV_CONDA_PACKAGE=$CONDA_PACKAGE

COPY environment.yml /tmp

LABEL maintainer="yue.nicholas@gmail.com"

RUN apt -y update && apt install -y build-essential cmake ninja-build

RUN cd /tmp && curl -L -O http://${ENV_HOST}:${ENV_PORT}/${ENV_CONDA_PACKAGE}

RUN sh /tmp/${ENV_CONDA_PACKAGE} -b

RUN . /root/anaconda3/bin/activate && \
    conda init bash && \
    conda config --set auto_activate_base false && \
    conda update -y -n base -c defaults conda && \
    conda env create -f /tmp/environment.yml

RUN cd /tmp && rm *.sh

ENTRYPOINT ["/root/anaconda3/condabin/conda", "run", "--no-capture-output", "-n", "bix"]