programming-basics-2022/10_docker/lesson04/README.md

Lesson 4: Persisting data

1. Change directory into lesson04.

2. We saw in Lesson 3 that data written to the final container layer does not persist. How do we deal with this? One option is to have the container write to an external system like a database system or cloud storage. There is, however, another method: mounting a local file system or directory.

3. Let's create a Docker container that saves the current date to a file. Here is our first attempt:

    # Dockerfile
    FROM debian:buster-slim
    LABEL maintainer="adamhl@stanford.edu"
   
    ADD date.sh /root/date.sh
    RUN chmod a+x /root/date.sh
    CMD /root/date.sh
   
    # date.sh
    #!/bin/sh
    echo "echoing date to /tmp/date.output"
    date > /tmp/date.output
   
    $ docker build . -t date
    $ docker run --rm --name=fuzzle date
    echoing date to /tmp/date.output
   

4. The above will create the file /tmp/date.output containing the date but we know that the file will not persist after the container stops running. To get around this we "mount" a local directory into the container's /tmp directory. By "local directory" we mean a directory on the computer where we run the our docker commands. We mount the colume when we run the container.

    $ mkdir -p /tmp/docker
    $ docker run --rm --name=fuzzle --volume=/tmp/docker:/tmp date
    echoing date to /tmp/date.output
    $ cat /tmp/docker/date.output
   

5. You can run a container and override the CMD command.

    $ docker run --rm --name=fuzzle date ls -ld /etc
    drwxr-xr-x 1 root root 4096 Jan 17 17:23 /etc/
    (The date.sh script did NOT run)
   

6. This is especially useful when you want to login to the container and debug the filesystem or CMD command.

7. You can overwrite a file in the image using the same --volume option. For example, let's overwrite /etc/debian_version with a different file.

    $ docker run --rm --name=fuzzle date cat /etc/debian_version
    10.7
    $ echo "fake-version" > /tmp/deb_ver
    $ docker run --rm --name=fuzzle --volume=/tmp/deb_ver:/etc/debian_version date cat /etc/debian_version
    fake-version
   

8. If you mount an external directory onto a container directory everything in the directory in the container is replaced with the external directory.

    $ docker run --rm --name=fuzzle date ls -l /usr
    total 32
    drwxr-xr-x  2 root root 4096 Jan 11 00:00 bin
    drwxr-xr-x  2 root root 4096 Nov 22 12:37 games
    ... more ...
   
    # Create a directory in /tmp with a single file.
    $ mkdir -p /tmp/usr; cp test.txt /tmp/usr
   
    # Mount /tmp/usr over /usr in the container
    $ docker run --rm --name=fuzzle --volume=/tmp/usr:/usr date ls -l /usr
    total 4
    -rw-r--r-- 1 52777 root 7 Jan 17 18:04 test.txt
   

9. You can mount an external directory in read-only mode. This is particularly useful when injecting secrets or configuration information into a container. Look for the ro in the --volume option below.

    $ mkdir -p /tmp/secrets
    $ echo "my-password" > /tmp/secrets/password
   
    $ docker run --rm --name=fuzzle --volume=/tmp/secrets:/secrets:ro date sleep 100000 &
   
    $ docker exec -ti fuzzle /bin/sh
    # cat /secrets/password
    my-password
    # echo "another secret" >> /secrets/password
    /bin/sh: 5: cannot create /secrets/password: Read-only file system
   
    $ docker kill fuzzle
   

10. You can run the entire container in read-only mode. The docker run option --read-only mounts the root filesystem (i.e., everything) in read-only excepting any externally mounted volumes. This lets you lock-down the filesystem except for those parts of the filesystem you know need to be written to (e.g., /tmp, /var/log, etc.).