Book Image

Apache Mesos Cookbook

By : David Blomquist, Tomasz Janiszewski
Book Image

Apache Mesos Cookbook

By: David Blomquist, Tomasz Janiszewski

Overview of this book

Apache Mesos is open source cluster sharing and management software. Deploying and managing scalable applications in large-scale clustered environments can be difficult, but Apache Mesos makes it easier with efficient resource isolation and sharing across application frameworks. The goal of this book is to guide you through the practical implementation of the Mesos core along with a number of Mesos supported frameworks. You will begin by installing Mesos and then learn how to configure clusters and maintain them. You will also see how to deploy a cluster in a production environment with high availability using Zookeeper. Next, you will get to grips with using Mesos, Marathon, and Docker to build and deploy a PaaS. You will see how to schedule jobs with Chronos. We’ll demonstrate how to integrate Mesos with big data frameworks such as Spark, Hadoop, and Storm. Practical solutions backed with clear examples will also show you how to deploy elastic big data jobs. You will find out how to deploy a scalable continuous integration and delivery system on Mesos with Jenkins. Finally, you will configure and deploy a highly scalable distributed search engine with ElasticSearch. Throughout the course of this book, you will get to know tips and tricks along with best practices to follow when working with Mesos.
Table of Contents (15 chapters)
Title Page
Credits
About the Authors
About the Reviewer
www.PacktPub.com
Customer Feedback
Preface

Tuning Marathon


In this recipe, you will learn how to configure Marathon to improve its performance.

Getting ready

Before you begin, you should capture Marathon metrics to see whether presented actions give the desired effect.

How to do it...

Marathon has many configuration options that could change its performance. It's written in Scala so it runs on JVM, which has even more options.

Most of the default settings are good enough even for big installations. However, there is a small set that should be changed.

You can gain better performance by changing Marathon's JVM options, especially to give it more memory and change the Garbage Collection algorithm. Tuning JVM is out of the scope of this book, so we will show only the basic methods. The following code will set Marathon to use 2 GB of memory and change the Garbage Collector to G1:

echo 'JAVA_OPTS=" -mx2g -ms2g -XX:+UseG1GC -XX:MetaspaceSize=100M"' >> /etc/default/marathon

The next thing that has an impact on performance is how many tasks...