Master JVM Usage when there is High Bulk-Indexing Requests Rate

Hi OpenSearch Experts,

We have seen our dedicated Master Node having high JVM memory usage when there are large request rate for bulk-indexing API calls. As far as I understand, OpenSearch bulk-indexing requests goes to Data Node only. Thus, I am puzzled why master node memory being a bottle neck for bulk-indexing.

Also, as a side question - do we have any good resources for learning OpenSearch internals?


Hi Sichen,

The following article [1] would hopefully clarify a lot of details for you regarding bulk indexing. Also, to understand better the context, could you please share:

  • size of the cluster (nodes / indices / shards)
  • topology (dedicated masters / coordinators / data nodes)
    Thank you.

[1] Why am I seeing bulk rejections in my Elasticsearch cluster? | Elastic Blog