Struggling to write to a file using Docker volumes and Flask

Question

My first question is: What is a docker volume? Is it an instance of the code? So if I write to a file, the file is written to that volume's version of the file?

Here is my data structure:

static   
---scripts  
---styles  
---templates 

matrices  
--- sampleMatrix.txt    <- this is the file I'm trying desperately to write to     
app.py  
dockerfile  
docker-compose.yml

App.py code:

try:
    matrixFile= open(fileName, "w+")
    matrixFile.write("text: \n")               <- should print test:
    print("wrote to: ",fileName, file=sys.stderr)
    matrixFile.close()

    matrixFile= open(fileName, "w+")
    print("About to read: ",fileName, file=sys.stderr)
    print(matrixFile.read(), file=sys.stderr)   <- prints a newline and nothing else
    print("Read from: ",fileName, file=sys.stderr)
    matrixFile.close()   
except:
    print("cant open file", file=sys.stderr)  <- never prints

With this, I know that flask is FINDING the file but seems not to be writing anything to it. Also no changes to the file locally

dockerfile:

FROM python:3.6.1-alpine
ADD . .
RUN pip install -r requirements.txt
RUN pip install requests
VOLUME [ "/matrices/" ]
CMD ["python","app.py"]

docker-compose.yml

version: '2'

services:
  <containerName>:
    build: . 
    ports: 
      - 82:80
    container_name: <containerName>
    volumes:
      - ~<userName>/Desktop/<folderName>/matrices/:/matrices/

I thought possibly the path of the volume could be wrong but I can't tell how to check. If this was theoretically working correctly, would I see changes to the local file immediately?

Answer 1

A Docker volume is essentially a "directory" you can mount to a container that is only managed and accessed through Docker. You are using a bind mount, that is some directory on your local filesystem managed by the host OS and made accessible to your Docker container. If mounted properly, you will be able to read/write this volume both in your Docker container and on your local filesystem. This picture from the docs helps visualize this:

In your Dockerfile, you've copied your code into your image. When you start your Docker container with a volume, that is mounted to the container and is accessible by your python code in the image. This isn't the only way to run it, for example you could also bind mount your Python code and run the container that way.

If you gave the wrong path in services.<containerName>.volumes, then Docker will try to create matrices if it doesn't exist. It looks like your path isn't right if it starts with ~username.

I'd suggest you double check the volume's path in the docker-compose.

Unix

ls ~<userName>/Desktop/<folderName>/matrices/

# if this fails, it will say so with something like:
# ls: cannot access '<YOUR_PATH>': No such file or directory

Windows Powershell (just in case)

Test-Path ~<userName>/Desktop/<folderName>/matrices/
# this will output False if the path is bad

Once you've verified the path is correct, you should be see whatever the Python code writes to appear in that directory on the host OS.

For more about Docker volumes, see these docs. For more about bind mounts, see these docs.