Skip to content

v1.19.1

Choose a tag to compare

@timvisee timvisee released this 04 Sep 07:59
· 526 commits to master since this release
v1.19.1
6ab21ca

Change log

Improvements

  • #10342, #10420 - Speed up and saturate memory bandwidth in quantized scoring with prefetching
  • #10391, #10392, #10362, #10437, #10439, #10381 - Rework and batch 4-bit TurboQuant SIMD implementation for better performance
  • #10441 - Limit query plane allocations for better search performance
  • #8650 - Improve L2 cosine preprocessing performance on AVX
  • #10042 - Batch deletion checks for faster full scans
  • #10054, #10010, #10194 - Batch searches in HNSW for faster search
  • #10099 - Skip bad items before pushing into HNSW priority queue
  • #10312, #10333 - Skip point ID resolution in search post processing to improve search performance
  • #10040, #10066 - Use raw payloads in shard transfer to make payload heavy shard transfers 1.5x faster
  • #9760 - Make set replica state and abort transfer/resharding crash safe to prevent consensus desync
  • #9786 - Make restart transfer operation crash safe to prevent consensus desync
  • #10345 - Query shard status more efficiently, don't build full segment info
  • #10455 - Faster memory reporting on Linux with cachestat
  • #10457 - Probe segment memory utilization outside of read locks
  • #10027 - Honor max_segment_size on filter based updates
  • #10426 - Support fast HNSW healing cancellation to prevent blocking consensus on shard drop
  • #10306 - Raise HTTP2 max pending resets on gRPC API to allow higher connection concurrency
  • #10243 - Expose used CPU cores in /metrics (cpu_cores_used)
  • #10132 - Extend container runtime detection, include Kubernetes and other runtimes

Bug Fixes

  • #10201 - Fix data consistency, flush CoW segments before building payload index
  • #10399 - Fix Gridstore panic on new page after unsafe shutdown, repairs preexisting failures
  • #10205 - On aborting resharding down, don't block consensus with deferred points for along time
  • #10324 - Reject vector size larger than 65536
  • #10143 - Reject empty dense vectors
  • #10374 - Reject mismatched dimension count in positive/negative examples of recommend average API
  • #10382 - Not specifying scroll limit would bypass max_query_limit
  • #10128 - Prevent underestimating is_empty and is_null cardinality
  • #10101 - Fix is_null filter for unindexed payload fields
  • #10341 - Fix phrase matching on unindexed payload field not respecting token positioning
  • #10181 - Propagate mmap read errors back to user rather than panicking
  • #10338 - Fix concurrent consensus operations awaiters dropping each other
  • #9957 - Limit tracing ID by 256 bytes, not characters
  • #9959 - Clamp picking random HNSW layer to lower bound of zero

Security

  • #10242 - Reject . and .. in collection names