Offheap Buffers for querying only?

are offheap buffers used for anything other than query processing?




If so, is there a reason that they are allocated for indexing jobs that can not accept real time queries (index_hadoop for instance)? This seems like a waste of resources if the buffers aren’t used for anything during the duration of the ingestion task.


The offheap buffers are only used for query processing.

Ideally indexing jobs that aren’t queryable shouldn’t get these resources, but presently all peon processes created by a given MiddleManager inherit the same configuration.

It’s undocumented but I believe you can override those configs on a per-task basis: