We are encountering an issue, wherein multiple nodes in our elasticsearch cluster goes down due to Elasticsearch:java.lang.OutOfMemoryError: Java heap space.
As a result of this all the shards get unassigned and takes a long time to recover.
Urgent help will be highly appreciated. We have encountered this issue thrice in 5 days.
If Elasticsearch hits an OutOfMemoryError then, by default, it will write a heap dump before exiting. You will need to load this heap dump (e.g. using MAT) to work out why it ran out of memory.
Also note that Elasticsearch 7 has much better protection against OutOfMemoryErrors than earlier versions. It would be wise to upgrade.
We are looking for some tuning parameters for our cluster. Please suggest. Below are cluster specs.
26 nodes
1,488 indices
11,170 shards
38,558,392,038 docs
18.28TB
Xmx: 31GB
Xms: 31 GB
Node configuration: 16c64G.
Apache, Apache Lucene, Apache Hadoop, Hadoop, HDFS and the yellow elephant
logo are trademarks of the
Apache Software Foundation
in the United States and/or other countries.