Details
-
Bug
-
Resolution: Duplicate
-
Critical
-
7.6.2
-
Couchbase server Enterprise Edition 7.6.2 build 3573
-
Untriaged
-
Linux x86_64
-
-
0
-
Unknown
Description
- Create 6 node cluster. (cluster_services ::['kv', 'kv', 'kv', 'fts', 'fts', 'fts']).
- Create a bucket 'standard_bkt'. Load 1000000 (1M) docs into the bucket.
- The schema of the document is as follows:
Data : { id : documentID, sno : unique_no, sname : random_name }
Metadata : [ xattrs : { "vector_data" : <vector_embedding_1536_dimensions> } ]
- Create a search index "xattrs_search_index", indexing the field 'vector_data' in xattrs section in metadata.
- After indexing is completed and no more mutations occurring in the cluster, the memory usage of the nodes ( the 3/6 nodes in the cluster which run search service ) is still above 70% .
- Due to this, most of the time the search query is getting rejected.
Query
curl -XPOST -H "Content-Type: application/json" -u <username>:<password> \ http://172.23.106.155:8094/api/bucket/nishanthb1/scope/goloader/index/nishanth-1-index/query \ -d '{ "query": { "match_none": {} }, "explain": true, "knn": [ { "field": "_$xattrs.vector_data", "k": 6, "vector": <vector_embedding> } ], "size": 10, "from": 0 }' |
Response
{"status":{"total":1,"failed":1,"successful":0,"errors":{"ashokb1.s1.ashok-1-index_3227245fe315ab4c_13aa53f3":"query request rejected"}},"hits":null,"total_hits":0,"cost":0,"max_score":0,"took":0,"facets":null} |
|
Detailed view on the memory usage, statistics and logs are attached in the attachment section.
Attachments
Issue Links
- duplicates
-
-
- Closed
-