Docker Hub link for automated builds : https://hub.docker.com/r/mishraakshay91/

Why we may not want to 'go cloud' with this app?

As users of this app, we may not deploy this app to cloud for few reasons :

1. This app is not going to generate any revenue for the company but only serves as a tool for your web developers. So one may not want to spend financial resources on deploying this to cloud environment. Also as this is not a client-facing website, you also do not need to think about latency, etc fro different geographical location perspective.

2. For reasons like compliance and/or security. you may not want to ship your logs to another location, as logs may contain confidential data like client details, locations, IPs etc.

3. This app will have a fixed and calculated load of constant amount, so you can assing a permanent VM to this app.

4. It is not a business critical app so even though downtime is not recommended, it won't be as critical as your prod app.

5. We are using docker-containers so already app is light-weight as it wil only use as much system resoueces as it needs, keeping rest of the system resources free for other applications, if any running on the same host.

Scatterplots using ELK stack on Docker as backend

We are making scatterplot possible in ELK now. Traditionally ELK stack makes us possible to create beautiful visualizations of aggregate data. But sometimes, requirement is to spot and analyse individual requests the system, their page loading performance, etc. Scatterplot makes this possible, and we are trying to integrate this idea with backend stack as ELK.

Getting Started

Install below on your linux machine :

• jdk 6+
• docker client (1.12+)
• docker server (1.12+)
• docker-compose v(1.9+)

Setting up this app on your linux machine

We have created a dummy log generator to test our code, as we need real-time logs as well as logs from the past dates. In prod system, filebeat will be monitoring actual log files instead of the ones created byour java utility here. so, change the mounted volume path to Logstash container and also the 'paths' in filebeat.yml which resides in docker-elk/filebeat/config directory.

Two simple steps to get the app running :

1. Run "random" & "increasing" java utilities to generated logs.
2. simply run docker-compose from "docker-elk" directiry.

For the first part,

You need sudo access because normally users don't have permission as a normal user to write in /var/log in Linux. (If you do not have sudo access, then change location of file generated to ~/iis.log in java code, mounted volume in docker-compose for logstash part and also in filebeat paths at the beginning.)

Create a folder called 'app' in your host inside /var/log/ directory as java utility that creates and dumps dummy logs in /var/log/app/iis.log file. Also, the iis.log file is created if not already existing, and logs are appended to it everytime we run the java utility.

As you can see, you have 2 versions of the utility in log-generator directory - random and increasing. Go to each one of them and run below commands in both of them.

javac ProjectLogs.java

This will create a .class file in the same folder, which then you can run using

sudo java ProjectLogs

For the second part,

just download docker-elk folder somewhere on your machine and run below command from that folder

docker-compose run --build 

One known issue that may come up while starting up elasticsearch containers is that, it may throw an error saying minimum acceptable value for vm.max_map_count should be 262144 for ES to start up. This is a highly recommended setting for prod system to avoid trouble later. So under using root access, run below command :

su

Now enter root password on prompt, and ten run

echo 262144 > /proc/sys/vm/max_map_count

Now run

docker-compose run --build

again and this should automatically fetch and build required images and also start running docker containers.

Your app will start listening on localhost:3000/elk_dashboard, as you can check in the logs. Check the browser. Username and pwd is "admin" and "admin".

Enjoy!!!