What is _recovery_source field?

[ruslaniv]

I have initiated indexing operation via python client that was abnormally terminated. When I ran

curl --location --request POST 'http://127.0.0.1:9200/index_name.vec/_disk_usage?run_expensive_tasks=true

to analyze the current index, I noticed that there is a _recovery_source field that based on its size contains raw dense_vector field data in float representation taking a lot of space.

"index_vector": {
                "store_size": "29.3mb",
                "all_fields": {
                    "total": "29.3mb",
                    "inverted_index": {
                        "total": "50.4kb",
                    },
                "stored_fields": "23.2mb",
                "doc_values": "286.6kb",
                "points": "14.2kb",
                "norms": "1.3kb",
                "term_vectors": "0b",
                "knn_vectors": "5.7mb",
        }

"_recovery_source": {
                "total": "21.8mb",
                "inverted_index": {
                    "total": "0b",
                },
                "stored_fields": "21.8mb",
                "doc_values": "0b",
                "points": "0b",
                "norms": "0b",
                "term_vectors": "0b",
                "knn_vectors": "0b",

 "_source": {
                "total": "1.3mb",
                "inverted_index": {
                    "total": "0b",
                },
                "stored_fields": "1.3mb",
                "doc_values": "0b",
                "points": "0b",
                "norms": "0b",
                "term_vectors": "0b",
                "knn_vectors": "0b",

Based on its name I can guess that this field is used for some recovery operations on the index but if it is storing raw dense vectors without any compression, then it's going to take a lot of space.
See:

Dense vector field space requirements

Hi! We are experimenting with dense vector field type for the purpose of similarity search. So we have a test index with approx 5_000_000 documents, each document has about 30 fields and 25 of these fields are mapped to both keyword AND text. So I created a new index, where 24 of these 25 fields were set only to keyword AND index set to False. And only one field mapped to dense_vector with 768 dimensions and index set to True. After indexing 500_000 documents we have noticed that the index s…

How to exclude dense_vector field from being stored

How can I exclude dense_vector field from being stored in the _source? I ran an experiment indexing approximately 6_000_000 documents and here is what I found out after running curl --location --request POST 'http://127.0.0.1:9200/index_name.vec/_disk_usage?run_expensive_tasks=true The dense_vector field indeed takes approximately 4 Kb in the index: (4 × 1024 + 4) / 1024 / 1024 / 1024 × (5872381 + 777907) = 25,3 Gb which is confirmed by: "knn_vectors": "26.2gb" But it is also storing dens…

So I was wondering what this field is used for and how can I disable it or at least disable dense vectors from being included in this field?

[BenTrent]

Recovery source is SUPPOSED to go away after segments are merged.

See: Indices with "_source.enabled: false" same size as indices with "_source.enabled: true" · Issue #41628 · elastic/elasticsearch · GitHub

It is effectively use by cross-cluster replication and shard disaster recoveries.

[ruslaniv]

Unfortunately it stays:

How to exclude dense_vector field from being stored

Yes I did run POST no_vector_in_source/_forcemerge?max_num_segments=1 but it did not help