Setup a Docker Swarm cluster Part III - Cluster Initialization

Build your own cheap but powerful self-hosted cluster and be free from any SaaS solutions by following this opinionated guide 🎉

This is the Part III of more global topic tutorial. Back to first part for intro.

Docker 🐳

Now we must do the classic Docker installation on each stateless servers. Repeat following commands on manager-01, worker-01 and runner-01.

curl -fsSL https://download.docker.com/linux/ubuntu/gpg | sudo gpg --dearmor -o /usr/share/keyrings/docker-archive-keyring.gpg
echo \
  "deb [arch=$(dpkg --print-architecture) signed-by=/usr/share/keyrings/docker-archive-keyring.gpg] https://download.docker.com/linux/ubuntu \
  $(lsb_release -cs) stable" | sudo tee /etc/apt/sources.list.d/docker.list > /dev/null

sudo apt-get update
sudo apt-get install -y docker-ce docker-ce-cli containerd.io docker-compose-plugin

sudo usermod -aG docker $USER

Then logout and use docker run hello-world and be sure all is OK. Follow official installation if not.

Hetzner & MTU

Thankfully since version 24 of Moby, we can set a default MTU for all future networks created through next docker stacks. For this create following JSON file on manager-01 :

1
{
2
  "default-network-opts": {
3
    "overlay": {
4
      "com.docker.network.driver.mtu": "1450"
5
    }
6
  }
7
}

Then restart docker daemon with sudo service docker restart. And that’s it, we don’t need to carry about anymore.

Enable Docker Swarm

Finally, enable Swarm mode on manager-01 :

docker swarm init --advertise-addr 10.0.0.2

The above command will show the command to launch to other worker nodes. Apply it on worker-01 and runner-01.

When done use docker node ls on manager node in order to confirm the presence of the 2 workers with Ready status and active.

Yeah, cluster is already properly configured. Far less overwhelming than Kubernetes, I should say.

CLI tools & environment labels

ctop is a very useful CLI tools that works like htop but dedicated for docker containers. Install it on every docker hosts :

echo "deb http://packages.azlux.fr/debian/ buster main" | sudo tee /etc/apt/sources.list.d/azlux.list
wget -qO - https://azlux.fr/repo.gpg.key | sudo apt-key add -
sudo apt update
sudo apt install -y docker-ctop

Before continue, let’s add some labels on nodes in order to differentiate properly production from build nodes :

# worker-01 is intended for running production app container
docker node update --label-add environment=production worker-01

# runner-01 is intended to build docker image through CI/CD pipeline
docker node update --label-add environment=build runner-01

Installing the Traefik - Portainer combo 💞

It’s finally time to start our first container services. The minimal setup will be :

• Traefik as main proxy with dynamic services discovery through that Docker API
• Portainer as main GUI for docker containers management and deployement

This 2 services will be deployed as docker services on manager-01.

Traefik 🛣️

The main task of traefik will be to redirect correct URL path to corresponding app service, according to regex rules (which domain or subdomain, which prefix URL path, etc.).

Thankfully, Traefik can be configured to take cares of all SSL certificates generation automatically without any intervention. We will use simple Let’s encrypt for this.

The static Traefik configuration

I should say that Traefik is not really easy to setup for new comers. The essential part to keep in mind is that this reverse proxy has 2 types of configuration, static and dynamic. Go here for detail explication of difference between these types of configuration.

Here we’ll talk about static configuration. Create next YAML file (TOML is also supported) :

1
entryPoints:
2
  https:
3
    address: :443
4
    http:
5
      middlewares:
6
        - gzip
7
      tls:
8
        certResolver: le
9
  http:
10
    address: :80
11
    http:
12
      redirections:
13
        entryPoint:
14
          to: https
15
          scheme: https
16
          permanent: true
17
  ssh:
18
    address: :22
19
certificatesResolvers:
20
  le:
21
    acme:
22
      email: admin@sw.dockerswarm.rocks
23
      storage: /certificates/acme.json
24
      tlsChallenge: {}
25
providers:
26
  docker:
27
    defaultRule: Host(`{{ index .Labels "com.docker.stack.namespace" }}.sw.dockerswarm.rocks`)
28
    exposedByDefault: false
29
    swarmMode: true
30
    network: traefik_public
31
api: {}
32
accessLog: {}
33
metrics:
34
  prometheus: {}

Traefik deployment

In order to deploy Traefik on our shiny new Docker Swarm, we must write a Docker Swarm deployment file that looks like to a classic Docker compose file. Create next file :

1
version: '3.8'
2

3
services:
4
  traefik:
5
    image: traefik:v2.6
6
    ports:
7
      - target: 22
8
        published: 22
9
        mode: host
10
      - target: 80
11
        published: 80
12
        mode: host
13
      - target: 443
14
        published: 443
15
        mode: host
16
    networks:
17
      - public
18
    volumes:
19
      - /etc/traefik:/etc/traefik
20
      - /var/run/docker.sock:/var/run/docker.sock:ro
21
      - certificates:/certificates
22
    deploy:
23
      placement:
24
        constraints:
25
          - node.labels.traefik-public.certificates == true
26
      labels:
27
        - traefik.enable=true
28
        - traefik.http.middlewares.gzip.compress=true
29
        - traefik.http.middlewares.admin-auth.basicauth.users=admin:${HASHED_PASSWORD?Variable not set}
30
        - traefik.http.middlewares.admin-ip.ipwhitelist.sourcerange=82.82.82.82
31
        - traefik.http.routers.traefik-public-api.entrypoints=https
32
        - traefik.http.routers.traefik-public-api.service=api@internal
33
        - traefik.http.routers.traefik-public-api.middlewares=admin-ip,admin-auth
34
        - traefik.http.services.traefik-public.loadbalancer.server.port=8080
35

36
networks:
37
  public:
38

39
volumes:
40
  certificates:

It’s finally time to test all this configuration !

Go to the manager-01, be sure to have above /etc/traefik/traefik.yml file, and do following commands :

# declare the current node manager as main certificates host, required in order to respect above deploy constraint
docker node update --label-add traefik-public.certificates=true manager-01

# generate your main admin password hash for any admin HTTP basic auth access into specific environment variable
export HASHED_PASSWORD=$(openssl passwd -apr1 aNyR4nd0mP@ssw0rd)

# deploy our 1st stack and cross the fingers...
docker stack deploy -c traefik-stack.yml traefik

# check status of the service, it should have 1 replica
docker service ls

# check logs for detail or any errors
docker service logs traefik_traefik

After few seconds, Traefik should launch and generate proper SSL certificate for its own domain. You can finally go to https://traefik.sw.dockerswarm.rocks. http:// should work as well thanks to permanent redirection.

If properly configured, you will be prompted for access. After entering admin as user and your own chosen password, you should finally access to the traefik dashboard !

Portainer ⛵

The hard part is done, we’ll finish this 2nd part by installing Portainer. Portainer is constituted of :

• A main GUI that can be exposed through Traefik
• An active agent for each docker node, done by the global deployment mode of Docker Swarm. This agent will be responsible for getting all running dockers through API and send them to Portainer manager.

Create portainer-agent-stack.yml swarm stack file with follogin content :

1
version: '3.8'
2

3
services:
4
  agent:
5
    image: portainer/agent:latest
6
    volumes:
7
      - /var/run/docker.sock:/var/run/docker.sock
8
      - /var/lib/docker/volumes:/var/lib/docker/volumes
9
    networks:
10
      - agent_network
11
    deploy:
12
      mode: global
13

14
  portainer:
15
    image: portainer/portainer-ce:latest
16
    command: -H tcp://tasks.agent:9001 --tlsskipverify
17
    volumes:
18
      - /mnt/storage-pool/portainer:/data
19
    networks:
20
      - agent_network
21
      - traefik_public
22
    deploy:
23
      placement:
24
        constraints: [node.role == manager]
25
      labels:
26
        - traefik.enable=true
27
        - traefik.http.routers.portainer.entrypoints=https
28
        - traefik.http.routers.portainer.middlewares=admin-ip
29
        - traefik.http.services.portainer.loadbalancer.server.port=9000
30

31
networks:
32
  agent_network:
33
  traefik_public:
34
    external: true

This is an adapted file from the official Portainer Agent Stack.

We use agent_network as overlay network for communication between agents and portainer. No need of admin-auth middleware here as Portainer has its own authentication.

For Traefik dynamic configuration, as you’ll see for the most of next stacks to be exposed, it’s required to :

• Set traefik.enable=true explicitly, when above exposedByDefault is set to false.
• Specify router entrypoints parameter to https, in order to force service discovery only to the above https Traefik entry point configured in its static configuration.
• Specify the web port of internal container where the HTTP request should be redirected with loadbalancer.server.port, this is required in Swarm mode because the services doesn’t use the EXPOSE instruction from containers.

Deploy the portainer stack :

# create the local storage for portainer in Gluster storage
sudo mkdir /mnt/storage-pool/portainer

# deploy portainer stack
docker stack deploy -c portainer-agent-stack.yml portainer

# check status
docker service ls

As soon as the main portainer service has successfully started, Traefik will detect it and configure it with SSL. The specific router for Portainer should appear in Traefik dashboard on HTTP section as below.

Go to the router detail for checking currently applied middlewares :

It’s time to create your admin account through https://portainer.sw.dockerswarm.rocks. If all goes well, a primary environment should be appearing, and you should have access to your cluster home environment with 2 stacks active.

Note as this primary endpoint was automatically created by the above tcp://tasks.agent:9001 command, which is the address of all portainer agents globally deployed. But you can add any other endpoints, aka clusters via the environments’ menu.

Some maintenance cluster tools 🐕‍🦺

It’s finally time to test our new cluster environment by testing some stacks through the Portainer GUI !

Keep the containers image up-to-date ⬆️

We’ll start by installing Diun, a very useful tool which notify us when used docker images has available update in its Docker registry.

Create the next stack through Portainer :

1
version: '3.8'
2

3
services:
4
  diun:
5
    image: crazymax/diun:latest
6
    command: serve
7
    volumes:
8
      - /mnt/storage-pool/diun:/data
9
      - /var/run/docker.sock:/var/run/docker.sock
10
    environment:
11
      TZ: Europe/Paris
12
      DIUN_WATCH_SCHEDULE: 0 */6 * * *
13
      DIUN_PROVIDERS_SWARM: 'true'
14
      DIUN_PROVIDERS_SWARM_WATCHBYDEFAULT: 'true'
15
      DIUN_NOTIF_MAIL_HOST:
16
      DIUN_NOTIF_MAIL_PORT:
17
      DIUN_NOTIF_MAIL_USERNAME:
18
      DIUN_NOTIF_MAIL_PASSWORD:
19
      DIUN_NOTIF_MAIL_FROM:
20
      DIUN_NOTIF_MAIL_TO:
21
    deploy:
22
      placement:
23
        constraints:
24
          - node.role == manager

Finally click on Deploy the stack, it’s equivalent of precedent docker stack deploy, nothing magic here. At the difference that Portainer will store the YML inside its volume, allowing full control, contrary to limited Traefik and Portainer cases.

Diun should now be deployed and manager host and ready to scan images for any updates !

You can check the full service page which will allows manual scaling, on-fly volumes mounting, environment variable modification, and show current running tasks (aka containers).

You can check the service logs which consist of all tasks logs aggregate.

Distributed cron jobs 🕰️

It’s frequent to have some crontab jobs for long-running maintenance tasks as database dumping, backups, exports and so on. But can we achieve that on this cluster environment ? Actually crazy-max has developed Cron precisely for this purpose ! Let’s add this service into same above maintenance stack.

1
#...
2
  cronjob:
3
    image: crazymax/swarm-cronjob
4
    volumes:
5
      - /var/run/docker.sock:/var/run/docker.sock
6
    environment:
7
      TZ: Europe/Paris
8
      LOG_LEVEL: info
9
      LOG_JSON: 'false'
10
    deploy:
11
      placement:
12
        constraints:
13
          - node.role == manager

This service will now search for any crontab related docker service labels across all Swarm cluster, and launch it accordingly schedule settings which uses standard crontab format. All you have to do is to use following structure for any service you want to run periodically :

1
#...
2
    deploy:
3
      labels:
4
        - swarm.cronjob.enable=true
5
        - swarm.cronjob.schedule=5 * * * *
6
        - swarm.cronjob.skip-running=true
7
      replicas: 0
8
      restart_policy:
9
        condition: none
10
#...

Get your own S3 💽

Let’s try with a tool with a web UI. Here is how get your own S3 bucket and be free from any external S3 provider. We’ll use our GlusterFS volume as a real storage.

Do sudo /mnt/storage-pool/minio on manager-01 and create following stack :

1
version: '3.8'
2

3
services:
4
  app:
5
    image: minio/minio
6
    volumes:
7
      - /mnt/storage-pool/minio:/data
8
    command: server /data --console-address ":9001"
9
    environment:
10
      MINIO_ROOT_USER: swarm
11
      MINIO_ROOT_PASSWORD:
12
      MINIO_BROWSER_REDIRECT_URL: https://minio.sw.dockerswarm.rocks
13
    networks:
14
      - traefik_public
15
    deploy:
16
      labels:
17
        - traefik.enable=true
18
        - traefik.http.routers.minio.entrypoints=https
19
        - traefik.http.routers.minio.rule=Host(`s3.sw.dockerswarm.rocks`)
20
        - traefik.http.routers.minio.service=minio
21
        - traefik.http.routers.minio-console.entrypoints=https
22
        - traefik.http.services.minio.loadbalancer.server.port=9000
23
        - traefik.http.routers.minio-console.service=minio-console
24
        - traefik.http.services.minio-console.loadbalancer.server.port=9001
25
      placement:
26
        constraints:
27
          - node.labels.environment == production
28

29
networks:
30
  traefik_public:
31
    external: true

The particularity of Minio is to have 2 web endpoints, one for web UI admin manager, and other as S3 API endpoint. So we need 2 Traefik routes in this case. Create an environment variable for MINIO_ROOT_PASSWORD and set your own admin password.

When deployed, wait few seconds for SSL auto generation (you can check it in the Traefik Dashboard) and go to https://minio.sw.dockerswarm.rocks in order to access the web administration by entering above credentials.

And yup, it’s done, create your 1st bucket through admin UI and you are ready to test the S3 API locally with https://s3.sw.dockerswarm.rocks/mybucket.

2nd check ✅

We’ve done the minimal viable Swarm setup with a nice cloud native reverse proxy, a containers GUI manager, and some other container sample tools. You can get a quick view of the current status of your cluster via the portainer visualizer !

It’s time to go further with self-hosted managed databases in next part.