Update a docker compose service properly

Question

I'm currently running an executable in Docker that has been mounted as a Docker volume in the foreground. The executable (or the service) continuously generates console output and can be interrupted by the Ctrl-C signal. From time to time, I have to update the service executable "manually". To do this, I wrote a sh script that temporarily stops the Docker container, downloads the new executable, unpacks it and restarts the Docker container. Is this a common way for my use case? Or can I update it in the container itself?

Below parts from:

docker-compose.yml

  myservice:
    image: debian:12
    init: true
    stop_grace_period: 1m
    user: 1000:1000
    volumes:
      - /home/tobi/myservice/:/myservice/
    command: >
      bash -c "cd /myservice/ && ./myserviceExec"
    labels:
      - "traefik.enable=true"
      - "traefik.http.routers.myservice.rule=Host(`sub.mydomain.com`)"
      - "traefik.http.services.myservice.loadbalancer.server.port=80"
      - "traefik.http.routers.myservice.entrypoints=websecure"
      - "traefik.http.routers.myservice.tls.certresolver=myresolver"

update_script.sh

#!/bin/sh
docker stop proxy-myservice-1
sleep 5
sh script_backup.sh
sleep 5
wget https://...afile.zip
sleep 5
unzip afile.zip
sleep 5
mv -v myserviceExec myservice/
rm -v afile.zip
sleep 5
docker start proxy-myservice-1
sleep 5
ls -la --color=auto
ls -la --color=auto myservice/
sleep 30
docker logs -t proxy-myservice-1

Answer 1

Your manual approach, using a single docker container, sounds like it is routinely accompanied by a brief outage as you switch to a new rev of code.

Is this a common way

No.

can I update it in the container

Yes, but I wouldn't recommend it. Let the container query an endpoint to learn the current rev of code, probably an executable name with a SemVer suffix. When it changes, download the new code, ask the current daemon to exit, and respawn on the same TCP port.

Here is a better and more common approach. A load balancer listens for internet client requests on some well known (address, port), and distributes requests to multiple backends which routine healthchecks have shown are currently up. Healthy backends are "in the rotation", getting a share of incoming requests.

When you rev the code, deploy new containers with that code so they will be added to the rotation. Once they are seen to be working correctly, tear down containers running downrev code.