Label your Docker Swarm Nodes

by Ruan Bekker

You can label your nodes to place specific containers on specific nodes.

This becomes handy when you have different type of nodes in your swarm. One example would be:

  • 1 Manager Node
  • 3 Worker Nodes (1GB Memory)
  • 1 Worker Node (4GB Memory)

Now we can place memory intensive tasks on our 4GB memory node.

That is one example, if you want to group specific tasks to specific data centers, you can also do that.

Use Case

In our use case today, I have 5 nodes in my raspberry pi docker swarm cluster:

$ docker node ls
ID                            HOSTNAME            STATUS              AVAILABILITY        MANAGER STATUS      ENGINE VERSION  
u261mbr3dkwiubga70stm4q80 *   rpi-01              Ready               Active              Leader              19.03.5  
6s48hk0kzrp2er06zltd7h4bg     rpi-02              Ready               Active                                  18.06.3-ce  
81br4uaayop19edv43es1hthb     rpi-03              Ready               Active                                  18.09.1  
7ixvmd6gr2vm1csnvwu6t44ei     rpi-04              Ready               Active                                  18.09.1  
igx5njlbw6zt1a3bew8ub7auh     rpi-05              Ready               Active                                  19.03.5

Nodes: rpi-02, rpi-03, rpi-04 are workers and have 1GB of Memory and rpi-05 is a worker and have 4GB of Memory.

We want to label rpi-05 with spec=memory as this way we would identify the node spec. Then when we create a service we will add a constraint that the task can only reside on a node with the label node.labels.spec==memory

Labeling the Node

From our manager, let's label our node rpi-05 with the key spec and the value memory:

$ docker node update --label-add spec=memory rpi-05

Let's inspect the node to verify that the node has been labeled:

$ docker node inspect rpi-05
[
    {
        "CreatedAt": "2020-02-11T14:05:41.930599473Z",
        "UpdatedAt": "2020-02-11T14:06:14.390067308Z",
        "Spec": {
            "Labels": {
                "spec": "memory"
            },
            "Role": "worker",
            "Availability": "active"
        },
        "Description": {
            "Hostname": "rpi-05",
            "Platform": {
                "Architecture": "armv7l",
                "OS": "linux"
            },
...
    }
]

Deploy some Workloads

Deploy workloads that will be hosted on our high memory node with a constraint flag:

$ docker service create \
  --name high-memory-spec-task \
  --constraint node.labels.spec==memory \ 
  pistacks/alpine ping localhost

When we verify:

$ docker service ps high-memory-spec-task
ID                  NAME                      IMAGE                    NODE                DESIRED STATE       CURRENT STATE            ERROR               PORTS  
05sjl7pf71h2        high-memory-spec-task.1   pistacks/alpine:latest   rpi-05              Running             Running 30 seconds ago

We can see that our task is running on our rpi-05 node.

Other types of constraints can look like this:

  • --constraint node.labels.spec!=memory
  • --constraint node.role==worker