Change8

v1.13.5

📦 qdrantView on GitHub →
2 features🐛 16 fixes🔧 6 symbols

Summary

This release focuses heavily on stability and performance improvements, particularly around cluster operations like resharding and snapshotting, alongside optimizations for payload indexing and resource utilization.

✨ New Features

  • Split CPU budget into CPU and IO to better saturate resources during optimization.
  • Enhanced payload indices to handle IsEmpty and IsNull conditions much more efficiently.

🐛 Bug Fixes

  • Fixed user-defined sharding not being applied in consensus snapshots, potentially corrupting cluster.
  • Fixed malformed user-defined sharding data in consensus snapshots if using numeric shard keys, potentially corrupting cluster.
  • Fixed cluster metadata not being in consensus snapshots, potentially causing cluster state desync.
  • Fixed resharding state not being applied with consensus snapshots, potentially causing cluster state desync.
  • Fixed snapshot restore error when numeric user-defined shard keys are used.
  • Fixed potential panic while propagating point deletions during resharding.
  • Fixed not loading or restoring segments from hidden files, preventing breakage on hidden files in storage by other tools.
  • Fixed search panic after HNSW creation with GPU when on NVIDIA.
  • Fixed write rate limit not being properly set in strict mode.
  • Fixed reads not being rate limited for shard transfers in strict mode (internal operation).
  • Fixed shard cleanup operation not being rate limited in strict mode (internal operation).
  • Properly rate limited batch point updates.
  • Kept existing shard configuration if snapshot restore failed to prevent panic on startup.
  • Used configured CA certificate for internal snapshot transfers.
  • Fixed opt-in anonymization of various telemetry fields.
  • Fixed showing warning if bootstrap URI is not provided.

🔧 Affected Symbols

HNSW building (GPU)shard transferconsensus snapshotspoint delete propagationreshardingshard cleanup