Description
Reposting a similar issue I posted on the Rust client repository, which continues to affect our team and which we suspect is an issue with QDrant and not the Rust client: qdrant/rust-client#187
I am unable to upload a large collection of vectors without getting a timeout error.
Current Behavior
I am uploading a collection of 385,186,142 vectors of dimension 512. Several million vectors into the upload, the upload stops with an error: "Error: Error in the response: The operation was cancelled Timeout expired."
If I try to delete and recreate the collection after this timeout error, creating the collection times out after deleting.
If I try to create the collection again, I get an error message indicating that "Collection data already exists at ./storage/collections/..."
Steps to Reproduce
- Create collection
- Upload collection, encounter timeout error
- Delete collection
- Create collection, encounter timeout error (shown immediately)
- Create collection, encounter "Collection data already exists" error
Collection config
{
"params": {
"vectors": {
"size": 512,
"distance": "Cosine",
"on_disk": true,
"datatype": "float16"
},
"shard_number": 1,
"replication_factor": 1,
"write_consistency_factor": 1,
"on_disk_payload": true
},
"hnsw_config": {
"m": 16,
"ef_construct": 100,
"full_scan_threshold": 10000,
"max_indexing_threads": 0,
"on_disk": false
},
"optimizer_config": {
"deleted_threshold": 0.2,
"vacuum_min_vector_number": 1000,
"default_segment_number": 0,
"max_segment_size": null,
"memmap_threshold": null,
"indexing_threshold": 0,
"flush_interval_sec": 5,
"max_optimization_threads": null
},
"wal_config": {
"wal_capacity_mb": 32,
"wal_segments_ahead": 0
},
"quantization_config": null,
"strict_mode_config": {
"enabled": false
}
}
Expected Behavior
I should not get a timeout error uploading a collection (or should get a more informative error message).
I should also never be in a situation where I both cannot create a collection or delete it.