How to use Docker Compose to run complex multi container apps - Node on MongoDB

This docker files shows how to build your own base image for various cloudera components running on Centos6.6 This is a simple setup with 3 Nodes and one MongoDB as backend run along with Weave for service discovery. Later will show you how to use Apache Bench to test the nodes and the fault tolerance of failing one NODE intentionally

We'll be using the following tools, technologies, and services in this post:

• Docker
• Docker Compose
• NodeJS
• MongoDB from Mongo Labs
• Weave
• Docker Hub
• Git
• OMDB API

Composing our Services

We will be setting up three webNode[1-3] Nodes running node.js configured to run express. They will be frontended withe HA Proxy to load balance the queries. For the database, Initially design was to have mongo DB running locally as docker container, but then found that i could use Mongo Labs as that would reduce load on my machine and follows a more 'DBaaS' Model.

I usually run Weave before other nodes are started as that allows you to have name resolution working among my docker nodes without too much networking configuration. Scope allows you monitor the nodes in near real time and connect to them from the gui to do some quick fixes or validations.

cAdivsor is another monitoring tool from google gives you detailed statistics about your containers, but Scope's ability to connect to the containers and visually appealing gui for each container more appealing.

The node servers are configured to run on port 8081, The HA Proxy comes with an admin console running on port 70. Feel free to change them to different ports, It will have to be done in these files : Dockerfile, haproxy.cfg, docker-compose.yml

How to run

From your docker terminal, jump to the directory where you have the copied all the files from this repo and run,

docker-compose up -d

How does it all look at the end?

                       |WebNode1|
User <--> HAProxy <--> |WebNode2| <--> MongoDB-Node1
                       |WebNode3|

Performance testing using Apache Benchmark

If you are using linux (CentOS or Fedora or RHEL) you can install it by running

yum -y install httpd-tools

Create a post data file with data in the below format

<form_field1_name>=<form_field1_value>&<form_field2_name>=<form_field2_value>

mediaTitle=titanic&mediaYear=1997

Basically what we are doing here, The number-n of queries is being set 10000. The concurrency -c is being set to 30. The rest of the flags are mandatory if you are using ab to post data to the url and can be safely ignored if you are not using. Finally provide the url to which you are posting. If you want to know more, Apache Benchmark has very good documentation here.

ab -n 10000 -c 30 -p post_data_filepath -v 4 -T 'application/x-www-form-urlencoded' <NODE.JS_POST_URL>

Performance testing using cURL

Incase installing another tool like ab is too much work, you can do something like this, (Dont forget to fill in the fields <..>)

for i in {1..10000}
do
curl -X POST --data "mediaTitle=<movie_title_of_choice>" --data "mediaYear=<Year_of_release_of_movie>"  <NODE.JS_POST_URL> > /dev/null;
done

To Do

• Setup Apache Bench along with Docker Compose