Volumes in a non default location breaks the setup

[Luxtech]

I cloned the git repo and edited the docker-compose file so that volumes are located in a known location. This is something that I always do for docker projects so that I can backup the volumes easily. When i changed the location of the volumes for this partkeepr container the http://localhost:8080/setup/ gives this error: Not Found The requested URL was not found on this server
If I keep the original docker-compose file the setup works.

Can someone help me to solve this problem? Or is it not possible to change the volume locations?

My docker-compose file:

version: "3"
services:

  database:
    image: mariadb:10.0
    restart: on-failure
    environment:
      - MYSQL_RANDOM_ROOT_PASSWORD=yes
      - MYSQL_DATABASE=partkeepr
      - MYSQL_USER=partkeepr
      - MYSQL_PASSWORD=partkeepr
    volumes:
      - ./volumes/mysql-data:/var/lib/mysql

  partkeepr:
    image: mhubig/partkeepr:latest
    restart: on-failure
    environment:
      - PARTKEEPR_DATABASE_HOST
      - PARTKEEPR_DATABASE_NAME
      - PARTKEEPR_DATABASE_PORT
      - PARTKEEPR_DATABASE_USER
      - PARTKEEPR_DATABASE_PASS
      - PARTKEEPR_FR3D_LDAP_DRIVER_ACCOUNTCANONICALFORM
      - PARTKEEPR_FR3D_LDAP_DRIVER_ACCOUNTDOMAINNAME
      - PARTKEEPR_FR3D_LDAP_DRIVER_ACCOUNTDOMAINNAMESHORT
      - PARTKEEPR_FR3D_LDAP_DRIVER_ACCOUNTFILTERFORMAT
      - PARTKEEPR_FR3D_LDAP_DRIVER_BASEDN
      - PARTKEEPR_FR3D_LDAP_DRIVER_BINDREQUIRESDN
      - PARTKEEPR_FR3D_LDAP_DRIVER_HOST
      - PARTKEEPR_FR3D_LDAP_DRIVER_OPTREFERRALS
      - PARTKEEPR_FR3D_LDAP_DRIVER_PASSWORD
      - PARTKEEPR_FR3D_LDAP_DRIVER_PORT
      - PARTKEEPR_FR3D_LDAP_DRIVER_USESSL
      - PARTKEEPR_FR3D_LDAP_DRIVER_USESTARTTLS
      - PARTKEEPR_FR3D_LDAP_DRIVER_USERNAME
      - PARTKEEPR_FR3D_LDAP_USER_ATTRIBUTE_EMAIL
      - PARTKEEPR_FR3D_LDAP_USER_ATTRIBUTE_USERNAME
      - PARTKEEPR_FR3D_LDAP_USER_BASEDN
      - PARTKEEPR_FR3D_LDAP_USER_ENABLED
      - PARTKEEPR_FR3D_LDAP_USER_FILTER
      - PARTKEEPR_LOCALE
      - PARTKEEPR_MAILER_AUTH_MODE
      - PARTKEEPR_MAILER_ENCRYPTION
      - PARTKEEPR_MAILER_HOST
      - PARTKEEPR_MAILER_PASSWORD
      - PARTKEEPR_MAILER_PORT
      - PARTKEEPR_MAILER_TRANSPORT
      - PARTKEEPR_MAILER_USER
      - PARTKEEPR_AUTH_MAX_USERS
      - PARTKEEPR_CATEGORY_PATH_SEPARATOR
      - PARTKEEPR_CRONJOB_CHECK
      - PARTKEEPR_FILESYSTEM_DATA_DIRECTORY
      - PARTKEEPR_FILESYSTEM_QUOTA
      - PARTKEEPR_MAINTENANCE
      - PARTKEEPR_MAINTENANCE_MESSAGE
      - PARTKEEPR_MAINTENANCE_TITLE
      - PARTKEEPR_OCTOPART_APIKEY
      - PARTKEEPR_PARTS_INTERNALPARTNUMBERUNIQUE
      - PARTKEEPR_PARTS_LIMIT
      - PARTKEEPR_USERS_LIMIT
      - PARTKEEPR_SECRET
    ports:
      - "8080:80"
    volumes:
      - ./volumes/partkeepr-conf:/var/www/html/app/config
      - ./volumes/partkeepr-data:/var/www/html/data
      - ./volumes/partkeepr-web:/var/www/html/web
    depends_on:
      - database

  cronjob:
    image: mhubig/partkeepr:latest
    restart: on-failure
    entrypoint: []
    command: bash -c "crontab /etc/cron.d/partkeepr && cron -f"
    volumes:
      - ./volumes/partkeepr-conf:/var/www/html/app/config:ro
      - ./volumes/partkeepr-data:/var/www/html/data
      - ./volumes/partkeepr-web:/var/www/html/web
    depends_on:
      - partkeepr

volumes:
  partkeepr-conf:
  partkeepr-data:
  partkeepr-web:
  mysql-data:

[m4v]

I believe that what is happening is that the image has files in those directories, thus directories like /var/www/html/data are overridden by empty ones when the volumes are mounted. Docker initializes named volumes with the contents of the image so this problem doesn't happen.

[ukoda]

I am trying to do the same thing for the same reason and seeing the same issue. I have tried running the original docker-compose file to create a good set of volume related files in /var/lib/docker/volumes, shutting it down, then copying them to where I really want them and updating the docker-compose file to point there and then bringing it back up. It throws a forbidden error so clearly I don't understand enough about how docker works to solve this problem at this time.

[ngtech]

You can use a third one time "init" container to copy the correct files to the non default location bind mounted volumes. I am attaching an excerpt of a working docker-compose.yaml file that bind mounts /data/docker/partkeepr/{conf,data,web} as volumes and attaches them to the partkeepr , partkeepr-cronjob and partkeepr-init containers.

1. Modify your compose file to bind mound the named volumes
2. Create the three host folders
3. Copy the partkeepr-init service definition below (Mounts volumes under /www-dest/)
4. Run docker-compose up partkeepr-init to initialize your bind mounted folders and wait until it rsyncs the data.
5. Comment out / delete partkeepr-init service.
6. Proceed with installation / usage.

  partkeepr:
    image: mhubig/partkeepr:latest
    container_name: partkeepr
    environment:
	...	     
    volumes:
      - partkeepr-conf:/var/www/html/app/config
      - partkeepr-data:/var/www/html/data
      - partkeepr-web:/var/www/html/web
	
  partkeepr-cronjob:
         ...
    volumes:
      - partkeepr-conf:/var/www/html/app/config
      - partkeepr-data:/var/www/html/data
      - partkeepr-web:/var/www/html/web    

  partkeepr-init:
    image: mhubig/partkeepr:latest
    restart: "no"
    entrypoint: []
    command: bash -c "apt update && apt install -y rsync && rsync -avHhSp /var/www/ /var/www-dest/"
    volumes:
      - partkeepr-conf:/var/www-dest/html/app/config
      - partkeepr-data:/var/www-dest/html/data
      - partkeepr-web:/var/www-dest/html/web

volumes:
  partkeepr-conf:
    driver: local
    driver_opts:
      o: bind
      type: none
      device: /data/docker/partkeepr/conf
  partkeepr-data:
    driver: local
    driver_opts:
      o: bind
      type: none
      device: /data/docker/partkeepr/data      
  partkeepr-web:
    driver: local
    driver_opts:
      o: bind
      type: none
      device: /data/docker/partkeepr/web

[Rascalov]

ngtech's solution worked for me. I also copied over the database seperately from the standard location to the custom location and it works as intended.

[crazyelectron-io]

I had a similar issue with a container running in a Kubernetes Pod and using an init-container like @ngtech mentioned worked like a charm. Thanks!