The easiest way to experiment with production-like Docker’s environment is to use Docker Swarm. And there are two ways to manipulate with a Swarm cluster - managing services or stacks. The below step by step guide provides you ability to configure a Swarm cluster, deploy it and do initial experiments.
Prerequisites
Installations
The required installations are
“Counting hits” application
The application counts how many time a page was open. It consists of a Flask application which stores data to a Redis.
The application’s images have 3 different tags:
extsoft/counting-hits:latestextsoft/counting-hits:v1extsoft/counting-hits:v2
All of them have equivalent functionality but represent different images.
“Counting hits” source
The sources below demonstrate how the images were created. You can just skip this section if you aren’t interested in.
app.py
import socket
from flask import Flask
from redis import Redis
app = Flask(__name__)
redis = Redis(host='redis', port=6379)
@app.route('/')
def hello():
count = redis.incr('hits')
return 'I have been seen {t} times. My hostname is: {h} \n'.format(
t=count, h=socket.gethostname()
)
if __name__ == "__main__":
app.run(host="0.0.0.0", debug=True)
requirements.txt
flask
redis
Dockerfile
FROM python:3.4-alpine
LABEL maintainer="Dmytro Serdiuk <dmytro.serdiuk@gmail.com>" site="extsoft.pro"
WORKDIR /code
COPY app.py requirements.txt ./
RUN pip install -r requirements.txt && mkdir -p logs
EXPOSE 5000
CMD ["python", "app.py"]
Prepare Swarm cluster
A couple of hosts are required to demonstrate a Swarm cluster. You can use any hosts available for yourself, but I’ll create several ones using docker-machine and VirtualBox. Then, the Swarm cluster will be configured for created hosts.
# create VMs
docker-machine create --driver virtualbox master
docker-machine create --driver virtualbox worker1
docker-machine create --driver virtualbox worker2
docker-machine ls
# configure SWARM cluster
docker-machine ssh master "docker swarm init --advertise-addr <master IP>"
docker-machine ssh worker1 "docker swarm join --token <token> <master IP>:2377"
docker-machine ssh worker2 "docker swarm join --token <token> <master IP>:2377"
# activate Docker CLI for "master" MV
eval $(docker-machine env master)
Deployment with docker service
Using docker service “Redis” service will be working along with “Counting hits” service. Communication between the services will be established using a custom network.
Run services
# create a network for services communication
docker network create --driver overlay --subnet 10.0.9.0/24 ch-network
# create a "redis" service
docker service create --detach --replicas 1 --network ch-network --name redis redis:alpine
# create 5 replicas of "ch" service
docker service create --detach --replicas 5 --network ch-network --publish 5000:5000 --name ch extsoft/counting-hits:v1
# wait around 10 seconds and check if all replicas are up
docker service ls
# check service status
docker service ps ch
# verify that all 5 replicas of "ch" service are working
for i in {1..10}; do curl http://192.168.99.100:5000; done
Scale services
# scale up to 7 replicas of "ch" service
docker service scale --detach ch=7
# see "ch" service stats
docker service inspect --pretty ch
# scale up to 2 replicas of "ch" service
docker service scale --detach ch=3
Update service image
docker service ps ch
# update "ch" service with new image version
docker service update --image extsoft/counting-hits:v2 ch
# wait some time and check a state of the "ch" service
docker service ps ch
# one more image update
docker service update --image extsoft/counting-hits:latest ch
Maintain a worker
Due to some reasons, you might want to shut down a worker (a host or a VM) for a maintenance purpose. If you want move working replicas from this worker to others, you need to drain availability of a worker and Swarm automatically will deploy replicas to other available workers. After this, the maintenance procedure can be performed.
# drain worker1 VM before maintenance
docker node update --availability drain worker1
# check Availability
docker node inspect --pretty worker1
# make sure nothing run on worker1
docker service ps ch
# activate worker1 VM after maintenance
docker node update --availability active worker1
Deployment with docker stack
Now let’s deploy the application using docker stack. First of all, need to prepare a compose file which describing the services structure. Then, just run the services using prepared configuration.
Prepare a compose file
stack-compose.yml
version: '3.0'
services:
ch:
image: extsoft/counting-hits:v1
depends_on:
- redis
ports:
- '5001:5000'
deploy:
replicas: 4
redis:
image: redis:alpine
volumes:
- redis-data:/data:rw
volumes:
redis-data:
driver: local
Run stack
docker stack deploy --compose-file stack-compose.yml ch
# look for available stacks
docker stack ls
# look for services in ch stack
docker stack services ch
# look for containers in ch stack
docker stack ps ch
# test app
for i in {1..10}; do curl http://192.168.99.100:5001; done
Also, you are able to manipulate stack services with docker service commands.
Clean env
eval $(docker-machine env -u)
docker-machine rm -y $(docker-machine ls -q)
Conclusion
Hope you enjoyed this guide! Any questions?