Basics

Docker Swarm is a container orchestration tool that has 1 manager and multiple worker nodes.

Create & Add Workers

docker swarm init --advertise-addr IP → Create a swarm with that machine being manager

docker swarm join --token TOKEN IP:PORT → Add workers to the swarm

Leave the Swarm

docker swarm leave [--force] (ON NODE)→ Leave the swarm, flag is not necessary

• Shows the node as “Down” in Manager docker node rm [-f] <node-id> (ON LEADER) → Remove the node from the swarm.
• Leave first, remote after. -f forces to remove before node leaving itself

Note: Forcefully removing a node creates problems, like node not knowing it is not a part of the swarm. Not Recommended

To completely destroy a swarm, the last remaining node, which is a manager, should leave —force

Information

• docker info shows info about swarm.
• docker node ls → Show node info

Services

Start & Stop

docker service create --name <name> --replicas <how many instance> -p 80:80 nginx

• This creates a service

docker service rm <name>

• This removes the service

docker service inspect [--pretty] <service-name>

• Inspects the service, shows info like parallelism, IP etc

docker service ps <service-name>

• Shows which nodes runs the service

Note

When you create a service with replicas, you can reach that service from every node even when that specific node doesn’t run that container. This is because of ingress routing mesh, which provide access even when a node is down. It creates high-availability and load-balancing

Services Info

docker service ls → Shows service info

When you stop a container in a working node, the manager automatically restarts

Scaling

docker service scale web-server=5 → Scale up to 5 replicas docker service ls → show the services docker service ps <service> → show the replicas

Updates

docker service create --replicas 3 --name redis --update-delay 10s redis:8.0 docker service update --image redis:8.2 redis → updates to 8.2

Warning: If the update name is wrong, that replica stops

Note

You can put --update-delay 10s when creating service to delay the update

Draining Node

Draining means stopping this node from getting any tasks for maintenance or planned shutdown. docker node update --availability drain <id>

If there are containers on that node, it will transferred to another node

docker node update --availability active <id>

Networking

Overlay network: It is used to connect different docker daemons together. It is mostly used in Swarm, but can be used to connect standalone containers. DockerHub

Create a Network

docker network create --driver overlay my-network docker service create --replicas 3 --network my-network --name test nginx:latest

docker service inspect test --pretty

Add to a Network

docker service update --network-add my-network test docker service update --network-rm my-network test

These commands are used to add/remove a network from a service

Storage

You can either add volume while creating or afterwards.

Update (Add Later)

docker service update --mount-add type=volume,source=new-vol,target=/app test

Create (While Creating)

docker service create --mount source=new-vol,target=/app -name myservice <image>

Note: You can also bind mount instead of volume mount, which is mounting a specific folder. Not suggested