Book Image

Monitoring Elasticsearch

By : Dan Noble, Pulkit Agrawal, Mahmoud Lababidi
Book Image

Monitoring Elasticsearch

By: Dan Noble, Pulkit Agrawal, Mahmoud Lababidi

Overview of this book

ElasticSearch is a distributed search server similar to Apache Solr with a focus on large datasets, a schema-less setup, and high availability. This schema-free architecture allows ElasticSearch to index and search unstructured content, making it perfectly suited for both small projects and large big data warehouses with petabytes of unstructured data. This book is your toolkit to teach you how to keep your cluster in good health, and show you how to diagnose and treat unexpected issues along the way. You will start by getting introduced to ElasticSearch, and look at some common performance issues that pop up when using the system. You will then see how to install and configure ElasticSearch and the ElasticSearch monitoring plugins. Then, you will proceed to install and use the Marvel dashboard to monitor ElasticSearch. You will find out how to troubleshoot some of the common performance and reliability issues that come up when using ElasticSearch. Finally, you will analyze your cluster’s historical performance, and get to know how to get to the bottom of and recover from system failures. This book will guide you through several monitoring tools, and utilizes real-world cases and dilemmas faced when using ElasticSearch, showing you how to solve them simply, quickly, and cleanly.
Table of Contents (15 chapters)
Monitoring Elasticsearch
About the Author
About the Reviewers


This chapter addressed some common performance and reliability issues that come up when using Elasticsearch. To reiterate some of the major points in this chapter:

  • Always double-check your Elasticsearch cluster's configuration for errors

  • Set the fielddata cache size, especially if you see OutOfMemoryError exceptions

  • Use the slow log to find what queries run slow on your cluster

  • Avoid aggregations on high-cardinality fields (such as millisecond timestamps)

  • Be cognizant of your data indexing strategy so that no one index grows too large

  • Use index warmers or enable eager_global_ordinals to ensure queries that use the fielddata cache are fast the first time we run them

  • If possible, use SSDs on nodes that index data, and avoid storing Elasticsearch indices on network storage

Most importantly, when diagnosing Elasticsearch issues, be meticulous about testing at each stage. For example, don't try to optimize a query by making changes to elasticsearch.yml, modifying the query criteria, and enabling...