ES6.1 Heap memory used by the index writer

Elastic Stack Elasticsearch
6

Might also be relevant to add that this problem index gets a lot of updates, and the updates happen via a Painless script:

{
  "script": {
    "source": "ctx._source.child_ids.add(params.child_id)",
    "params": {
      "child_id": 1234
    }
  }
}

It's maintaining an array that has an average length of 3 but a maximum length of 100,000.

Perhaps if all 100,000 of those Painless updates happened to a given document in series it would place an unusual stress on the shard?

Reading @nik9000's post about how Lucene handles _source has me wondering.