Visualizes how Docker nodes work together in a distributed environment.
This is a demo project for the trivago code workshop.
This will create the following setup:
+--------------+
| |
+-----> Server 1 |
| | |
| +--------------+
|
+------------+ +------------+ | +--------------+
| | | | | | |
| Client +----> Swarm +---------> Server 2 |
| | | | | | |
+------------+ +------------+ | +--------------+
|
| +--------------+
| | |
+-----> Server 3 |
| |
+--------------+
The servers are golang webservers that take an int and return true
if the int is a prime number and false otherwise.
The traffic to the containers is load-balanced with HAproxy.
Swarm is used for scaling the number of webservers.
First start haproxy and consul with ./start_haproxy.sh.
You can add new servers with ./start_primeserver.sh.
This will create a new instance of the primeserver container.
After that you can run the clients either locally or in another container.
E.g. to check each number from 1 to 1000 if it is prime in batches of 100, run:
go run primeclient.go --host=hostname --start=1 --stop=1000 --step=100
Of course you can start many clients at the same time to create more load.
To get an idea of the current cluster status, you can collect some metrics.
For that you can start the metrics backend, which is completely independent from the application.
You can enable and disable it at any time.
Start it with docker-compose up from the respective folder.
To make it work across hosts, also add the following entry to the /etc/hosts on cluster every machine:
127.0.0.1 influxdb
Please adjust the IP to the host where InfluxDB is running.
@jsacha @mendler