Empty snapshots in distributed deployment

I am faced with the problem of creating snapshots in a distributed qdrant deployment. The collection is very large, but its snapshot is created with a size of 11.8 kB.

Current Behavior

Qdrant is deployed in distributed mode on 5 nodes. The collection was created with 3 shards and a replication factor of 2. I use a load balancer in front of the qdrant cluster. Therefore, requests are randomly distributed across nodes. If the snapshot creation request is directed to a node that does not have any shards of this collection, an empty snapshot of 11 kilobytes is created as a result.
Request:
curl --location --request POST 'http://qdrant:6333/collections/mycollection/snapshots'
Response:
{
"result": {
"name": "mycollection-13542296025311659898-2023-11-10-08-34-23.snapshot",
"creation_time": "2023-11-10T08:34:23",
"size": 11776
},
"status": "ok",
"time": 0.181759348
}

Steps to Reproduce

1. Set up a distributed Qdrant deployment.
2. Create a collection with a replication factor less than the number of nodes. Or move the shards in such a way that there are no shards of this collection on some node.
3. Fill the collection with enough number of vectors.
4. Send a request to create a snapshot of the collection to a node that does not have shards of this collection.
5. Check created snapshot.

Expected Behavior

A normal snapshot is created regardless of which node the snapshot creation request came to.

Start by reproducing the snapshot request in a five-node distributed deployment with three shards and replication factor two, targeting a node without collection shards. Trace the collection snapshot endpoint and request routing to determine why it creates an 11.8 kB snapshot; done means the same request creates a normal snapshot regardless of the receiving node.