Starting from version 7.11, Elasticsearch automatically sizes JVM heap based on a node’s roles and total memory. For more information, see Heap size settings. When running as a container within Kubernetes, the total available memory for Elasticsearch is determined by the resource limits defined for the container. By default, ECK applies a memory resource limit of 2 GiB to the container, which usually results in an adjusted heap size of 1 GiB. You can increase the total memory available to each Elasticsearch node by defining your own limits in the
podTemplate section of the manifest. An example can be found in Set compute resources for Elasticsearch.
If you define
resource.requests but not
resource.limits, it could result in the Elasticsearch container assuming that it has access to all the memory available in the Kubernetes node. This can result in bad performance or unexpected behaviour if other memory-consuming workloads gets scheduled to the same node.
The following instructions apply if you are using:
- Versions of Elasticsearch before 7.11
- Elasticsearch 7.11 or later, and you want to override the default heap size.
To manually change the heap size of Elasticsearch, set the
ES_JAVA_OPTS environment variable in the
podTemplate. Make sure you set the resource
limits at the same time so that the Pod gets enough resources allocated within the Kubernetes cluster. See Set compute resources for Elasticsearch for an example and more information.