ray · Issues· 3586 open

[Data] Allow disabling Task Fusion / Documenting how to avoid it

[Clusters] rsync incompatibility error when using other docker images

[autoscaler][docs] Examples with `head_start_ray_commands` should listen on dashboard host 0.0.0.0

[Serve] [RFC] Computed Retry-After for backpressure rejections

[Core] Worker hangs at exit and leaks as ray::IDLE when log rotation is on and a child process holds the redirected stream

[Core][Train] Support Torch runtime with TPUs

[RLLib] Change connector configuration callbacks to transform ConnectorPipelineV2

[serve][llm] Governance middleware layer for Ray Serve LLM — PII detection, cost budgets, policy enforcement, and audit trails

[data] ObjectRefs passed to map UDF are not automatically deref'ed

[Core] After GCS failover, raylet grants lease requests queued by the previous GCS; worker stays ray::IDLE and leaks the node's GPU

[Data] Actor pools cannot scale down to zero in multi-step pipelines with downstream bottleneck

[Data] read_clickhouse materializes an entire ReadTask result in one block

[Tune] Add examples of using XGBoost with the sklearn API and add examples of regression tasks

[RFC][Data] Add read_hive datasource for Apache Hive

[Data] from_pandas with override_num_blocks argument is broken on pandas version 3