Docker Cookbook

7.3 Starting Containers on a Cluster with Docker Swarm | 199

---

docker
pull
swarm
SCRIPT
.
.
.
Once the nodes are up and Vagrant has returned, 
ssh
to the head node and start a
Swarm container by using the 
swarm
image that was pulled during the bootstrap
process:
$ vagrant ssh swarm-head
$ docker run -v /vagrant:/tmp/vagrant -p 1234:1234 -d swarm manage \
file://tmp/vagrant/swarm-cluster.cfg -H=0.0.0.0:1234
72acd5bc00de0b411f025ef6f297353a1869a3cc8c36d687e1f28a2d8f422a06
The Swarm server setup uses a file-based discovery mechanism.
The 
swarm-cluster.cfg
file contains the hardcoded lists of the Swarm
nodes started by Vagrant. Additional discovery 
services
are avail‐
able for Swarm. You can use a service hosted by Docker Inc., 
Con‐
sul
, 
Etcd
, or 
ZooKeeper
. You can also write your own discovery
interface.
With the Swarm server running and the worker nodes discovered, you will be able to
use the local Docker client to get information about the cluster and start containers.
You will need to use the 
-H
option of the Docker CLI to target the Swarm server run‐
ning in a container instead of the local Docker daemon:
$ docker -H 0.0.0.0:1234 info
Containers: 0
Nodes: 3
swarm-2: 192.168.33.12:2375
└ Containers: 0
└ Reserved CPUs: 0 / 1
└ Reserved Memory: 0 B / 490 MiB
swarm-3: 192.168.33.13:2375
└ Containers: 0
└ Reserved CPUs: 0 / 1
└ Reserved Memory: 0 B / 490 MiB
swarm-1: 192.168.33.11:2375
└ Containers: 0
└ Reserved CPUs: 0 / 1
└ Reserved Memory: 0 B / 490 MiB
Using the local Docker client and specifying the Swarm server as a Docker daemon
endpoint, you can start containers on the entire cluster. For example, let’s start Nginx:
$ docker -H 0.0.0.0:1234 run -d -p 80:80 nginx
8399c544b61953fd5610b01be787cb3802e2e54f220673b94d78160d0ee35b33
$ docker -H 0.0.0.0:1234 run -d -p 80:80 nginx
1b2c4634fc6d9f2c3fd63dd48a2580f466590ddff7405f889ada885746db3cbd
$ docker -H 0.0.0.0:1234 ps
CONTAINER ID IMAGE ... PORTS NAMES