9.8 KiB
Docker Swarm (mode) cluster
This section explains how to create a multi-host docker cluster with swarm mode using docker-machine and how to deploy Træfik on it.
The cluster consists of:
- 3 servers
- 1 manager
- 2 workers
- 1 overlay network (multi-host networking)
Prerequisites
- You will need to install docker-machine
- You will need the latest VirtualBox
Cluster provisioning
First, let's create all the required nodes. It's a shorter version of the swarm tutorial.
docker-machine create -d virtualbox manager
docker-machine create -d virtualbox worker1
docker-machine create -d virtualbox worker2
Then, let's setup the cluster, in order:
- initialize the cluster
- get the token for other host to join
- on both workers, join the cluster with the token
docker-machine ssh manager "docker swarm init \
--listen-addr $(docker-machine ip manager) \
--advertise-addr $(docker-machine ip manager)"
export worker_token=$(docker-machine ssh manager "docker swarm \
join-token worker -q")
docker-machine ssh worker1 "docker swarm join \
--token=${worker_token} \
--listen-addr $(docker-machine ip worker1) \
--advertise-addr $(docker-machine ip worker1) \
$(docker-machine ip manager)"
docker-machine ssh worker2 "docker swarm join \
--token=${worker_token} \
--listen-addr $(docker-machine ip worker2) \
--advertise-addr $(docker-machine ip worker2) \
$(docker-machine ip manager)"
Let's validate the cluster is up and running.
docker-machine ssh manager docker node ls
ID HOSTNAME STATUS AVAILABILITY MANAGER STATUS
013v16l1sbuwjqcn7ucbu4jwt worker1 Ready Active
8buzkquycd17jqjber0mo2gn8 worker2 Ready Active
fnpj8ozfc85zvahx2r540xfcf * manager Ready Active Leader
Finally, let's create a network for Træfik to use.
docker-machine ssh manager "docker network create --driver=overlay traefik-net"
Deploy Træfik
Let's deploy Træfik as a docker service in our cluster. The only requirement for Træfik to work with swarm mode is that it needs to run on a manager node - we are going to use a constraint for that.
docker-machine ssh manager "docker service create \
--name traefik \
--constraint=node.role==manager \
--publish 80:80 --publish 8080:8080 \
--mount type=bind,source=/var/run/docker.sock,target=/var/run/docker.sock \
--network traefik-net \
traefik \
--docker \
--docker.swarmMode \
--docker.domain=traefik \
--docker.watch \
--api"
Let's explain this command:
| Option | Description |
|---|---|
--publish 80:80 --publish 8080:8080 |
we publish port 80 and 8080 on the cluster. |
--constraint=node.role==manager |
we ask docker to schedule Træfik on a manager node. |
--mount type=bind,source=/var/run/docker.sock,target=/var/run/docker.sock |
we bind mount the docker socket where Træfik is scheduled to be able to speak to the daemon. |
--network traefik-net |
we attach the Træfik service (and thus the underlying container) to the traefik-net network. |
--docker |
enable docker provider, and --docker.swarmMode to enable the swarm mode on Træfik. |
--api |
activate the webUI on port 8080 |
Deploy your apps
We can now deploy our app on the cluster, here whoami, a simple web server in Go.
We start 2 services, on the traefik-net network.
docker-machine ssh manager "docker service create \
--name whoami0 \
--label traefik.port=80 \
--network traefik-net \
emilevauge/whoami"
docker-machine ssh manager "docker service create \
--name whoami1 \
--label traefik.port=80 \
--network traefik-net \
emilevauge/whoami"
!!! note
We set whoami1 to use sticky sessions (--label traefik.backend.loadbalancer.stickiness=true).
We'll demonstrate that later.
!!! note
If using docker stack deploy, there is a specific way that the labels must be defined in the docker-compose file.
Check that everything is scheduled and started:
docker-machine ssh manager "docker service ls"
ID NAME MODE REPLICAS IMAGE PORTS
moq3dq4xqv6t traefik replicated 1/1 traefik:latest *:80->80/tcp,*:8080->8080/tcp
ysil6oto1wim whoami0 replicated 1/1 emilevauge/whoami:latest
z9re2mnl34k4 whoami1 replicated 1/1 emilevauge/whoami:latest
Access to your apps through Træfik
curl -H Host:whoami0.traefik http://$(docker-machine ip manager)
Hostname: 5b0b3d148359
IP: 127.0.0.1
IP: 10.0.0.8
IP: 10.0.0.4
IP: 172.18.0.5
GET / HTTP/1.1
Host: whoami0.traefik
User-Agent: curl/7.55.1
Accept: */*
Accept-Encoding: gzip
X-Forwarded-For: 10.255.0.2
X-Forwarded-Host: whoami0.traefik
X-Forwarded-Proto: http
X-Forwarded-Server: 77fc29c69fe4
curl -H Host:whoami1.traefik http://$(docker-machine ip manager)
Hostname: 3633163970f6
IP: 127.0.0.1
IP: 10.0.0.14
IP: 10.0.0.6
IP: 172.18.0.5
GET / HTTP/1.1
Host: whoami1.traefik
User-Agent: curl/7.55.1
Accept: */*
Accept-Encoding: gzip
X-Forwarded-For: 10.255.0.2
X-Forwarded-Host: whoami1.traefik
X-Forwarded-Proto: http
X-Forwarded-Server: 77fc29c69fe4
!!! note As Træfik is published, you can access it from any machine and not only the manager.
curl -H Host:whoami0.traefik http://$(docker-machine ip worker1)
Hostname: 5b0b3d148359
IP: 127.0.0.1
IP: 10.0.0.8
IP: 10.0.0.4
IP: 172.18.0.5
GET / HTTP/1.1
Host: whoami0.traefik
User-Agent: curl/7.55.1
Accept: */*
Accept-Encoding: gzip
X-Forwarded-For: 10.255.0.3
X-Forwarded-Host: whoami0.traefik
X-Forwarded-Proto: http
X-Forwarded-Server: 77fc29c69fe4
curl -H Host:whoami1.traefik http://$(docker-machine ip worker2)
Hostname: 3633163970f6
IP: 127.0.0.1
IP: 10.0.0.14
IP: 10.0.0.6
IP: 172.18.0.5
GET / HTTP/1.1
Host: whoami1.traefik
User-Agent: curl/7.55.1
Accept: */*
Accept-Encoding: gzip
X-Forwarded-For: 10.255.0.4
X-Forwarded-Host: whoami1.traefik
X-Forwarded-Proto: http
X-Forwarded-Server: 77fc29c69fe4
Scale both services
docker-machine ssh manager "docker service scale whoami0=5"
docker-machine ssh manager "docker service scale whoami1=5"
Check that we now have 5 replicas of each whoami service:
docker-machine ssh manager "docker service ls"
ID NAME MODE REPLICAS IMAGE PORTS
moq3dq4xqv6t traefik replicated 1/1 traefik:latest *:80->80/tcp,*:8080->8080/tcp
ysil6oto1wim whoami0 replicated 5/5 emilevauge/whoami:latest
z9re2mnl34k4 whoami1 replicated 5/5 emilevauge/whoami:latest
Access to your whoami0 through Træfik multiple times.
Repeat the following command multiple times and note that the Hostname changes each time as Traefik load balances each request against the 5 tasks:
curl -H Host:whoami0.traefik http://$(docker-machine ip manager)
Hostname: f3138d15b567
IP: 127.0.0.1
IP: 10.0.0.5
IP: 10.0.0.4
IP: 172.18.0.3
GET / HTTP/1.1
Host: whoami0.traefik
User-Agent: curl/7.55.1
Accept: */*
Accept-Encoding: gzip
X-Forwarded-For: 10.255.0.2
X-Forwarded-Host: whoami0.traefik
X-Forwarded-Proto: http
X-Forwarded-Server: 77fc29c69fe4
Do the same against whoami1:
curl -c cookies.txt -H Host:whoami1.traefik http://$(docker-machine ip manager)
Hostname: 348e2f7bf432
IP: 127.0.0.1
IP: 10.0.0.15
IP: 10.0.0.6
IP: 172.18.0.6
GET / HTTP/1.1
Host: whoami1.traefik
User-Agent: curl/7.55.1
Accept: */*
Accept-Encoding: gzip
X-Forwarded-For: 10.255.0.2
X-Forwarded-Host: whoami1.traefik
X-Forwarded-Proto: http
X-Forwarded-Server: 77fc29c69fe4
Because the sticky sessions require cookies to work, we used the -c cookies.txt option to store the cookie into a file.
The cookie contains the IP of the container to which the session sticks:
cat ./cookies.txt
# Netscape HTTP Cookie File
# https://curl.haxx.se/docs/http-cookies.html
# This file was generated by libcurl! Edit at your own risk.
whoami1.traefik FALSE / FALSE 0 _TRAEFIK_BACKEND http://10.0.0.15:80
If you load the cookies file (-b cookies.txt) for the next request, you will see that stickiness is maintained:
curl -b cookies.txt -H Host:whoami1.traefik http://$(docker-machine ip manager)
Hostname: 348e2f7bf432
IP: 127.0.0.1
IP: 10.0.0.15
IP: 10.0.0.6
IP: 172.18.0.6
GET / HTTP/1.1
Host: whoami1.traefik
User-Agent: curl/7.55.1
Accept: */*
Accept-Encoding: gzip
Cookie: _TRAEFIK_BACKEND=http://10.0.0.15:80
X-Forwarded-For: 10.255.0.2
X-Forwarded-Host: whoami1.traefik
X-Forwarded-Proto: http
X-Forwarded-Server: 77fc29c69fe4
