Searchable snapshots

Versions (relevant - OpenSearch/Dashboard/Server OS/Browser):
2.16

Describe the issue:
We have repository-s3 setup in out environment. the S3 storage is on-prem located. Snapshots can be done and can be restored as searchable snapshots. When the snapshot has been restored as searchable snapshot (“storage_type”: “remote_snapshot”), only metadata should be imported and data should remain on the S3 filestorage. In our case the snapshot uses the full amount of the size of the index, not only the metadatas.

Does anone have the same issue?
Any Ideas are welcome…

Configuration:

Relevant Logs or Screenshots:

attributes of my_restored_snapshot
{
“index.codec”: “best_compression”,
“index.creation_date”: “1726444801009”,
“index.mapping.coerce”: “true”,
“index.mapping.ignore_malformed”: “true”,
“index.mapping.total_fields.limit”: “10000”,
“index.max_docvalue_fields_search”: “150”,
“index.number_of_replicas”: “0”,
“index.number_of_shards”: “2”,
“index.opendistro.index_state_management.policy_id”: “daily-30”,
“index.provided_name”: “my-index”,
“index.query.default_field”: [
“event.original”
],
“index.replication.type”: “DOCUMENT”,
“index.searchable_snapshot.index.id”: “GNgMCFkpTXWNsQ230avsKg”,
“index.searchable_snapshot.repository”: “s3-repo”,
“index.searchable_snapshot.snapshot_id.name”: “my-snapshot”,
“index.searchable_snapshot.snapshot_id.uuid”: “rS3ztxNqT1eXQRaKkKOJfQ”,
“index.store.type”: “remote_snapshot”,
“index.uuid”: “BB-POe_ZSaCdQWJpwDcZng”,
“index.version.created”: “136367827”
}

@andreborn, have you explored (tested) alternative index compression methods such as zstd?

more here: Index codecs - OpenSearch Documentation

Best,
mj

Thx, for the hint! I tried several methods, but after restoring the index fron the snapshot as a searchable snapshot is never smaller than the origin index. changing the index.codec does not eliminate the problem. do you have any other hints?