Search

Performance Tuning Guide

Performance Tuning Guide

10.5

Back Next

Recommendations for Elasticsearch

Recommendations for Elasticsearch

The Elasticsearch parameters can be optimized for better performance.

The following table lists the recommendations for Elasticsearch:

Parameter	Recommended Setting	Description
Heap Size	-Xms8192m -Xmx8192m	Based on your requirement, contact the system administrator to increase the values as per the available memory on the system. However, do not exceed 31 GB of heap. Instead, you can add additional clustered nodes. Informatica recommends that you have the same value for Xms and Xmx.
Elasticsearch indexing commit interval during the initial index job.	Default is 30 seconds.	es.index.refresh.interval=30 This property is found in the cmxserver.properties file. Sets the interval in seconds for Elasticsearch to commit the changes to the data after an Initially Index Smart Search Data batch job is run. The data is available for search after this time interval. Modifying this value has a performance impact. Before you modify this property, you must contact Informatica Global Customer Support.
Elastic search server number of shards	Default is 2.	search.provisioning.numshards=2 Number of shards to create on your Elasticsearch environment. The value depends on the maximum number of shards and the total number of nodes. For example, if the maximum number of shards is 1and the number of nodes is 3, you can create 3 shards. Based on the requirement and available nodes, you can increase the value. If you want to modify this value, you require re-indexing.
Elasticsearch server number of replicas.	1 or more. Default is 0 The number of total shards, including the primary and replicas are evenly divided by the number of nodes in the elastic cluster.	search.provisioning.numreplicas=1 Number of copies of the Elasticsearch engine documents that you want to create on different nodes. Use the replication factor to create multiple copies of the documents in the shards of different nodes. You require multiple copies of the documents to achieve high availability if one or more nodes shut down unexpectedly. For example, if the replication factor is 2, you get two copies of the documents in two nodes.
Elastic search maximum connection per host.	Default is 100.	ex.max.conn.per.host=100 Sets the maximum number of Elasticsearch nodes that you want to connect to the host. The maximum number of total connection value for the Elasticsearch server is the number of nodes x maximum connection. For example, if you use 3 nodes for Elasticsearch server, then 3 x100 = 300 maximum number of total connection value is available for the Elasticsearch server.
Number of Elasticsearch nodes.	Odd number. For example: 1, 3, 5.	To avoid split-brain issue, you must set the number of nodes in odd numbers in the Elasticsearch server cluster.
Elasticsearch server maximum worker threads.	Default 1.	ex.max.threads=1 Sets the maximum number of threads that you want the Apache asynchronous non-blocking receiver to use for each node in the Elasticsearch cluster. You can set the maximum of number of processors available for the node. Contact Informatica Global Customer Support before modifying this property.
Elasticsearch server facets.	Limit the use of facets and choose this fields with less unique values.	Search performance is good with limited facets.
Elasticsearch server searchable fields.	Mark only the required fields as searchable.	If you have more searchable fields, it slows down the performance of jobs that needs to delete the indexes.
Elasticsearch server disk space monitoring.	You must have a monitoring tool to keep an eye on the disk space of the Elasticsearch server.	If you have less than 15% of free disk space, Elasticsearch server marks all indexes as read.
Elasticsearch server maximum virtual memory areas.	OS command : sysctl -w vm.max_map_count=262144	The MMap FS type stores the shared index on the file system by mapping a file into memory. Ensure that you set the values based on your requirement. See the Informatica Knowledge Base article number 589503 for further details.

Recommendations

Watch

Comments

0 COMMENTS

We’d like to hear from you! Log in to comment.