Details
-
Bug
-
Resolution: Fixed
-
Critical
-
7.6.2
-
7.6.2-3584
-
Untriaged
-
0
-
Unknown
Description
- Create a cluster and couchstore bucket and load 10M items with base64 encoded 1536 dimensional vector embedding in the document body.
- Create a vector index on this field.
- FTS is crashing immediately after.
Service 'fts' exited with status 1. Restarting. Messages:
2024-05-01T15:06:10.440-07:00 [INFO] feed_dcp_gocbcore: Start, name: GleamBookUsers0VolumeCollection0_fts_idx_0_55f5a768ea975261_62ea9b3d, num streams: 28, manifestUID: 2, streamOptions: {FilterOptions: &{ScopeID:0 CollectionIDs:[10]}, StreamOptions: &{StreamID:18}}, vbuckets: [996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023]
2024-05-01T15:06:12.210-07:00 [FATA] scorch AsyncError, path: /data/@fts/GleamBookUsers0VolumeCollection0_fts_idx_0_55f5a768ea975261_a0aa4aa5.pindex/store, treating this as fatal, err: got err persisting snapshot: Error in void faiss::Clustering::train_encoded(faiss::idx_t, const uint8_t*, const faiss::Index*, faiss::Index&, const float*) at /home/couchbase/jenkins/workspace/cbdeps-platform-build/faiss/faiss/Clustering.cpp:303: Error: 'std::isfinite(x[i])' failed: input contains NaN's or Inf's, stack dump: /data/@fts/dumps/1714601172.fts.stack.dump.txt -- main.initBleveOptions.func2() at init_bleve.go:113
Attachments
Issue Links
- relates to
-
MB-61757 Different query results with same query vector but different index type (base64 vs regular)
- Open
For Gerrit Dashboard: MB-61742 | ||||||
---|---|---|---|---|---|---|
# | Subject | Branch | Project | Status | CR | V |
209498,2 | MB-61134, MB-61742: Bump up bleve/v2 | trinity | cbft | Status: MERGED | +2 | +1 |
209499,2 | MB-61134, MB-61742: go mod tidy | trinity | cbftx | Status: MERGED | +2 | +1 |
209500,2 | MB-61134, MB-61742: go mod tidy | trinity | n1fty | Status: MERGED | +2 | +1 |
209501,2 | MB-61134, MB-61742: go mod tidy | trinity | query | Status: MERGED | +2 | +1 |