MacroPower / docker-diskover

A Docker container for the Diskover space mapping application

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

linuxserver.io

Blog Discord Discourse Fleet GitHub Open Collective

The LinuxServer.io team brings you another container release featuring:

  • regular and timely application updates
  • easy user mappings (PGID, PUID)
  • custom base image with s6 overlay
  • weekly base OS updates with common layers across the entire LinuxServer.io ecosystem to minimise space usage, down time and bandwidth
  • regular security updates

Find us at:

  • Blog - all the things you can do with our containers including How-To guides, opinions and much more!
  • Discord - realtime support / chat with the community and the team.
  • Discourse - post on our community forum.
  • Fleet - an online web interface which displays all of our maintained images.
  • GitHub - view the source for all of our repositories.
  • Open Collective - please consider helping us by either donating or contributing to our budget

linuxserver/diskover

GitHub Stars GitHub Release GitHub Package Repository GitLab Container Registry MicroBadger Layers Docker Pulls Docker Stars Jenkins Build LSIO CI

diskover is a file system crawler and disk space usage software that uses Elasticsearch to index and manage data across heterogeneous storage systems.

diskover

Supported Architectures

Our images support multiple architectures such as x86-64, arm64 and armhf. We utilise the docker manifest for multi-platform awareness. More information is available from docker here and our announcement here.

Simply pulling ghcr.io/linuxserver/diskover should retrieve the correct image for your arch, but you can also pull specific arch images via tags.

The architectures supported by this image are:

Architecture Tag
x86-64 amd64-latest
arm64 arm64v8-latest
armhf arm32v7-latest

Usage

Here are some example snippets to help you get started creating a container.

docker-compose (recommended)

Compatible with docker-compose v2 schemas.

version: '2'
services:
  diskover:
    image: linuxserver/diskover
    container_name: diskover
    environment:
      - PUID=1000
      - PGID=1000
      - TZ=Europe/London
      - REDIS_HOST=redis
      - REDIS_PORT=6379
      - ES_HOST=elasticsearch
      - ES_PORT=9200
      - ES_USER=elastic
      - ES_PASS=changeme
      - RUN_ON_START=true
      - USE_CRON=true
    volumes:
      - /path/to/diskover/config:/config
      - /path/to/diskover/data:/data
    ports:
      - 80:80
      - 9181:9181
      - 9999:9999
    mem_limit: 4096m
    restart: unless-stopped
    depends_on:
      - elasticsearch
      - redis
  elasticsearch:
    container_name: elasticsearch
    image: docker.elastic.co/elasticsearch/elasticsearch:5.6.9
    volumes:
      - ${DOCKER_HOME}/elasticsearch/data:/usr/share/elasticsearch/data
    environment:
      - bootstrap.memory_lock=true
      - "ES_JAVA_OPTS=-Xms2048m -Xmx2048m"
    ulimits:
      memlock:
        soft: -1
        hard: -1
  redis:
    container_name: redis
    image: redis:alpine
    volumes:
      - ${HOME}/docker/redis:/data

docker cli

docker run -d \
  --name=diskover \
  -e PUID=1000 \
  -e PGID=1000 \
  -e TZ=Europe/London \
  -e REDIS_HOST=redis \
  -e REDIS_PORT=6379 \
  -e ES_HOST=elasticsearch \
  -e ES_PORT=9200 \
  -e ES_USER=elastic \
  -e ES_PASS=changeme \
  -e INDEX_NAME=diskover- \
  -e DISKOVER_OPTS= \
  -e WORKER_OPTS= \
  -e RUN_ON_START=true \
  -e USE_CRON=true \
  -p 80:80 \
  -p 9181:9181 \
  -p 9999:9999 \
  -v /path/to/diskover/config:/config \
  -v /path/to/diskover/data:/data \
  --restart unless-stopped \
  ghcr.io/linuxserver/diskover

Parameters

Container images are configured using parameters passed at runtime (such as those above). These parameters are separated by a colon and indicate <external>:<internal> respectively. For example, -p 8080:80 would expose port 80 from inside the container to be accessible from the host's IP on port 8080 outside the container.

Parameter Function
-p 80 diskover Web UI
-p 9181 rq-dashboard web UI
-p 9999 diskover socket server
-e PUID=1000 for UserID - see below for explanation
-e PGID=1000 for GroupID - see below for explanation
-e TZ=Europe/London Specify a timezone to use EG Europe/London
-e REDIS_HOST=redis Redis host (optional)
-e REDIS_PORT=6379 Redis port (optional)
-e ES_HOST=elasticsearch ElasticSearch host (optional)
-e ES_PORT=9200 ElasticSearch port (optional)
-e ES_USER=elastic ElasticSearch username (optional)
-e ES_PASS=changeme ElasticSearch password (optional)
-e INDEX_NAME=diskover- Index name prefix (optional)
-e DISKOVER_OPTS= Optional arguments to pass to the diskover crawler (optional)
-e WORKER_OPTS= Optional argumens to pass to the diskover bots launcher (optional)
-e RUN_ON_START=true Initiate a crawl every time the container is started (optional)
-e USE_CRON=true Run a crawl on as a cron job (optional)
-v /config Persistent config files
-v /data Default mount point to crawl

Environment variables from files (Docker secrets)

You can set any environment variable from a file by using a special prepend FILE__.

As an example:

-e FILE__PASSWORD=/run/secrets/mysecretpassword

Will set the environment variable PASSWORD based on the contents of the /run/secrets/mysecretpassword file.

Umask for running applications

For all of our images we provide the ability to override the default umask settings for services started within the containers using the optional -e UMASK=022 setting. Keep in mind umask is not chmod it subtracts from permissions based on it's value it does not add. Please read up here before asking for support.

User / Group Identifiers

When using volumes (-v flags) permissions issues can arise between the host OS and the container, we avoid this issue by allowing you to specify the user PUID and group PGID.

Ensure any volume directories on the host are owned by the same user you specify and any permissions issues will vanish like magic.

In this instance PUID=1000 and PGID=1000, to find yours use id user as below:

  $ id username
    uid=1000(dockeruser) gid=1000(dockergroup) groups=1000(dockergroup)

 

Application Setup

Once running the URL will be http://<host-ip>/ initial application spinup will take some time so please reload if you get an empty response. We highly reccomend using Docker compose for this image as it includes multiple database backends to link into. If you are looking to mount the elasticsearch and redis data to your host machine for access neither of them currently support setting a custom UID or GID they will run by default as:

  • Redis - UID=999 GID=999
  • Elasticsearch - UID=1000 GID=1000

ElasticSearch also requires a sysctl setting on the host machine to run properly. Running sysctl -w vm.max_map_count=262144 will solve this issue. To make this setting persistent through reboots, set this value in /etc/sysctl.conf.

If you simply want the application to work you can mount these to folders with 0777 permissions, otherwise you will need to create these users host level and set the folder ownership properly.

By default this compose example is pointed to a single directory and the UID and GID you pass to the diskover container needs to match that folders ownership. If these are shared folders with many owners the indexing will likely fail.

For specific questions or help setting up diskover in your environment please refer to the project's Github page Diskover.

Docker Mods

Docker Mods Docker Universal Mods

We publish various Docker Mods to enable additional functionality within the containers. The list of Mods available for this image (if any) as well as universal mods that can be applied to any one of our images can be accessed via the dynamic badges above.

Support Info

  • Shell access whilst the container is running: docker exec -it diskover /bin/bash
  • To monitor the logs of the container in realtime: docker logs -f diskover
  • container version number
    • docker inspect -f '{{ index .Config.Labels "build_version" }}' diskover
  • image version number
    • docker inspect -f '{{ index .Config.Labels "build_version" }}' ghcr.io/linuxserver/diskover

Updating Info

Most of our images are static, versioned, and require an image update and container recreation to update the app inside. With some exceptions (ie. nextcloud, plex), we do not recommend or support updating apps inside the container. Please consult the Application Setup section above to see if it is recommended for the image.

Below are the instructions for updating containers:

Via Docker Compose

  • Update all images: docker-compose pull
    • or update a single image: docker-compose pull diskover
  • Let compose update all containers as necessary: docker-compose up -d
    • or update a single container: docker-compose up -d diskover
  • You can also remove the old dangling images: docker image prune

Via Docker Run

  • Update the image: docker pull ghcr.io/linuxserver/diskover
  • Stop the running container: docker stop diskover
  • Delete the container: docker rm diskover
  • Recreate a new container with the same docker run parameters as instructed above (if mapped correctly to a host folder, your /config folder and settings will be preserved)
  • You can also remove the old dangling images: docker image prune

Via Watchtower auto-updater (only use if you don't remember the original parameters)

  • Pull the latest image at its tag and replace it with the same env variables in one run:
    docker run --rm \
    -v /var/run/docker.sock:/var/run/docker.sock \
    containrrr/watchtower \
    --run-once diskover
    
  • You can also remove the old dangling images: docker image prune

Note: We do not endorse the use of Watchtower as a solution to automated updates of existing Docker containers. In fact we generally discourage automated updates. However, this is a useful tool for one-time manual updates of containers where you have forgotten the original parameters. In the long term, we highly recommend using Docker Compose.

Image Update Notifications - Diun (Docker Image Update Notifier)

  • We recommend Diun for update notifications. Other tools that automatically update containers unattended are not recommended or supported.

Building locally

If you want to make local modifications to these images for development purposes or just to customize the logic:

git clone https://github.com/linuxserver/docker-diskover.git
cd docker-diskover
docker build \
  --no-cache \
  --pull \
  -t ghcr.io/linuxserver/diskover:latest .

The ARM variants can be built on x86_64 hardware using multiarch/qemu-user-static

docker run --rm --privileged multiarch/qemu-user-static:register --reset

Once registered you can define the dockerfile to use with -f Dockerfile.aarch64.

Versions

  • 19.12.19: - Rebasing to alpine 3.11.
  • 28.06.19: - Rebasing to alpine 3.10.
  • 12.04.19: - Rebase to Alpine 3.9.
  • 23.03.19: - Switching to new Base images, shift to arm32v7 tag.
  • 01.11.18: - Initial Release.

About

A Docker container for the Diskover space mapping application

License:GNU General Public License v3.0


Languages

Language:Dockerfile 56.7%Language:Shell 43.3%