Scaling services

We should always run at least two instances of any given service. That way they can share the load and, if one of them fails, there will be no downtime. We'll explore Swarm's failover capability soon and leave load balancing for the next chapter.

We can, for example, tell Swarm that we want to run five replicas of then go-demo service:

docker service scale go-demo=5

With the service scale command, we scheduled five replicas. Swarm will make sure that five instances of go-demo are running somewhere inside the cluster.

We can confirm that, indeed, five replicas are running through the, already familiar, service ls command:

docker service ls

The output is as follows (IDs are removed for brevity):

NAME       MODE       REPLICAS IMAGE
go-demo    replicated 5/5      vfarcic/go-demo:1.0
go-demo-db replicated 1/1      mongo:3.2.10

As we can see, five out of five REPLICAS of the go-demo service are running.

The service ps command provides more detailed information about a single service:

docker...