opensourcemirror
/
netdata
mirror of https://github.com/netdata/netdata.git
1
0
Fork 0
Code Issues Releases Wiki Activity
netdata/packaging/docker/README.md

13 KiB
Raw Blame History

Install Netdata with Docker

Running Netdata in a container works best for an internal network or to quickly analyze a host. Docker helps you get set up quickly, and doesn't install anything permanent on the system, which makes uninstalling Netdata easy.

See our full list of Docker images at Docker Hub.

Limitations running Netdata in Docker

For monitoring the whole host, running Netdata in a container can limit its capabilities. Some data, like the host OS performance or status, is not accessible or not as detailed in a container as when running Netdata directly on the host.

A way around this is to provide special mounts to the Docker container so that Netdata can get visibility on host OS information like /sys and /proc folders or even /etc/group and shadow files.

Also, we now ship Docker images using an ENTRYPOINT directive, not a COMMAND directive. Please adapt your execution scripts accordingly. You can find more information about ENTRYPOINT vs COMMAND in the Docker documentation.

Package scrambling in runtime (x86_64 only)

Our x86_64 Docker images use Polymorphic Polyverse Linux package scrambling. For increased security, you can enable rescrambling of Netdata packages during runtime by setting the environment variable RESCRAMBLE=true while starting Netdata with a Docker container.

Run Netdata with the docker command

Quickly start Netdata with the docker command. Netdata is then available at http://host:19999.

docker run -d --name=netdata \
  -p 19999:19999 \
  -v /etc/passwd:/host/etc/passwd:ro \
  -v /etc/group:/host/etc/group:ro \
  -v /proc:/host/proc:ro \
  -v /sys:/host/sys:ro \
  -v /etc/os-release:/host/etc/os-release:ro \
  --cap-add SYS_PTRACE \
  --security-opt apparmor=unconfined \
  netdata/netdata

The above can be converted to a docker-compose.yml file for ease of management:

version: '3'
services:
  netdata:
    image: netdata/netdata
    hostname: example.com # set to fqdn of host
    ports:
      - 19999:19999
    cap_add:
      - SYS_PTRACE
    security_opt:
      - apparmor:unconfined
    volumes:
      - /etc/passwd:/host/etc/passwd:ro
      - /etc/group:/host/etc/group:ro
      - /proc:/host/proc:ro
      - /sys:/host/sys:ro

Some of the bind-mounts are optional depending on how you use Netdata:

  • If you don't want to use the apps.plugin functionality, you can remove the mounts of /etc/passwd and /etc/group (they are used to get proper user and group names for the monitored host) to get slightly better security.

  • Most modern linux distos supply /etc/os-release although some older distros only supply /etc/lsb-release. If this is the case you can change the line above that mounts the file inside the container to -v /etc/lsb-release:/host/etc/lsb-release:ro.

  • If your host is virtualized then Netdata cannot detect it from inside the container and will output the wrong metadata (e.g. on /api/v1/info queries). You can fix this by setting a variable that overrides the detection using, e.g. --env VIRTUALIZATION=$(systemd-detect-virt -v). If you are using a docker-compose.yml then add:

    environment:
      - VIRTUALIZATION=${VIRTUALIZATION}

This allows the information to be passed into docker-compose using:

VIRTUALIZATION=$(systemd-detect-virt -v) docker-compose up

Starting with v1.12, Netdata collects anonymous usage information by default and sends it to Google Analytics. Read about the information collected, and learn how to-opt, on our anonymous statistics page.

The usage statistics are vital for us, as we use them to discover bugs and priortize new features. We thank you for actively contributing to Netdata's future.

Docker container names resolution

There are a few options for resolving container names within Netdata. Some methods of doing so will allow root access to your machine from within the container. Please read the following carefully.

Docker socket proxy (safest option)

Deploy a Docker socket proxy that accepts and filters out requests using something like HAProxy so that it restricts connections to read-only access to the CONTAINERS endpoint.

The reason it's safer to expose the socket to the proxy is because Netdata has a TCP port exposed outside the Docker network. Access to the proxy container is limited to only within the network.

Below is an example repository (and image) that provides a proxy to the socket.

You run the Docker Socket Proxy in its own Docker Compose file and leave it on a private network that you can add to other services that require access.

version: '3'
services:
  netdata:
    image: netdata/netdata
    # ... rest of your config ...
    ports:
      - 19999:19999
    environment:
      - DOCKER_HOST=proxy:2375
  proxy:
    image: tecnativa/docker-socket-proxy
    volumes:
     - /var/run/docker.sock:/var/run/docker.sock:ro
    environment:
      - CONTAINERS=1

Note: Replace 2375 with the port of your proxy.

Giving group access to the Docker socket (less safe)

Important Note: You should seriously consider the necessity of activating this option, as it grants to the netdata user access to the privileged socket connection of docker service and therefore your whole machine.

If you want to have your container names resolved by Netdata, make the netdata user be part of the group that owns the socket.

To achieve that just add environment variable PGID=[GROUP NUMBER] to the Netdata container, where [GROUP NUMBER] is practically the group id of the group assigned to the docker socket, on your host.

This group number can be found by running the following (if socket group ownership is docker):

grep docker /etc/group | cut -d ':' -f 3

Running as root (unsafe)

You should seriously consider the necessity of activating this option, as it grants to the netdata user access to the privileged socket connection of docker service, and therefore your whole machine.

version: '3'
services:
  netdata:
    image: netdata/netdata
    # ... rest of your config ...
    volumes:
      # ... other volumes ...
      - /var/run/docker.sock:/var/run/docker.sock:ro
    environment:
      - DOCKER_USR=root

Pass command line options to Netdata

Since we use an ENTRYPOINT directive, you can provide Netdata daemon command line options such as the IP address Netdata will be running on, using the command instruction.

Install Netdata using docker-compose with SSL/TLS enabled HTTP Proxy

For a permanent installation on a public server, you should secure the Netdata instance. This section contains an example of how to install Netdata with an SSL reverse proxy and basic authentication.

You can use the following docker-compose.yml and Caddyfile files to run Netdata with Docker. Replace the domains and email address for Let's Encrypt before starting.

Caddyfile

This file needs to be placed in /opt with name Caddyfile. Here you customize your domain and you need to provide your email address to obtain a Let's Encrypt certificate. Certificate renewal will happen automatically and will be executed internally by the caddy server.

netdata.example.org {
  proxy / netdata:19999
  tls admin@example.org
}

docker-compose.yml

After setting Caddyfile run this with docker-compose up -d to have fully functioning Netdata setup behind HTTP reverse proxy.

version: '3'
volumes:
  caddy:

services:
  caddy:
    image: abiosoft/caddy
    ports:
      - 80:80
      - 443:443
    volumes:
      - /opt/Caddyfile:/etc/Caddyfile
      - $HOME/.caddy:/root/.caddy
    environment:
      ACME_AGREE: 'true'
  netdata:
    restart: always
    hostname: netdata.example.org
    image: netdata/netdata
    cap_add:
      - SYS_PTRACE
    security_opt:
      - apparmor:unconfined
    volumes:
      - /etc/passwd:/host/etc/passwd:ro
      - /etc/group:/host/etc/group:ro
      - /proc:/host/proc:ro
      - /sys:/host/sys:ro
      - /var/run/docker.sock:/var/run/docker.sock:ro

Restrict access with basic auth

You can restrict access by following official caddy guide and adding lines to Caddyfile.

Publish a test image to your own repository

At Netdata, we provide multiple ways of testing your Docker images using your own repositories. You may either use the command line tools available or take advantage of our Travis CI infrastructure.

Using tools manually from the command line

The script packaging/docker/build-test.sh can be used to create an image and upload it to a repository of your choosing.

Usage: packaging/docker/build-test.sh -r <REPOSITORY> -v <VERSION> -u <DOCKER_USERNAME> -p <DOCKER_PWD> [-s]
	-s skip build, just push the image
Builds an amd64 image and pushes it to the docker hub repository REPOSITORY

This is especially useful when testing a Pull Request for Kubernetes, since you can set image to an immutable repository and tag, set the imagePullPolicy to Always and just keep uploading new images.

Example:

We get a local copy of the Helm chart at https://github.com/netdata/helmchart. We modify values.yaml to have the following:

image:
  repository: cakrit/netdata-prs
  tag: PR5576
  pullPolicy: Always

We check out PR5576 and run the following:

./packaging/docker/build-test.sh -r cakrit/netdata-prs -v PR5576 -u cakrit -p 'XXX'

Then we can run helm install [path to our helmchart clone].

If we make changes to the code, we execute the same build-test.sh command, followed by helm upgrade [name] [path to our helmchart clone]

Inside Netdata organization, using Travis CI

To enable Travis CI integration on your own repositories (Docker and Github), you need to be part of the Netdata organization.

Once you have contacted the Netdata owners to setup you up on Github and Travis, execute the following steps

  • Preparation

    • Have Netdata forked on your personal GitHub account
    • Get a GitHub token: Go to GitHub settings -> Developer Settings -> Personal access tokens, and generate a new token with full access to repo_hook, read-only access to admin:org, public_repo, repo_deployment, repo:status, and user:email settings enabled. This will be your GITHUB_TOKEN that is described later in the instructions, so keep it somewhere safe.
    • Contact the Netdata team and seek for permissions on https://scan.coverity.com should you require Travis to be able to push your forked code to coverity for analysis and report. Once you are setup, you should have your email you used in coverity and a token from them. These will be your COVERITY_SCAN_SUBMIT_EMAIL and COVERITY_SCAN_TOKEN that we will refer to later.
    • Have a valid Docker hub account, the credentials from this account will be your DOCKER_USERNAME and DOCKER_PWD mentioned later.

  • Setting up Travis CI for your own fork (Detailed instructions provided by Travis team here)

    • Login to travis with your own GITHUB credentials (There is Open Auth access)
    • Go to your profile settings, under repositories section and setup your Netdata fork to be built by Travis CI.
    • Once the repository has been setup, go to repository settings within Travis CI (usually under https://travis-ci.com/NETDATA_DEVELOPER/netdata/settings, where NETDATA_DEVELOPER is your GitHub handle), and select your desired settings.

  • While in Travis settings, under Netdata repository settings in the Environment Variables section, you need to add the following:

    • DOCKER_USERNAME and DOCKER_PWD variables so that Travis can login to your Docker Hub account and publish Docker images there.
    • REPOSITORY variable to NETDATA_DEVELOPER/netdata, where NETDATA_DEVELOPER is your GitHub handle again.
    • GITHUB_TOKEN variable with the token generated on the preparation step, for Travis workflows to function properly.
    • COVERITY_SCAN_SUBMIT_EMAIL and COVERITY_SCAN_TOKEN variables to enable Travis to submit your code for analysis to Coverity.

Having followed these instructions, your forked repository should be all set up for integration with Travis CI. Happy testing!

analytics