Repository Issue Activity (beta)

#65529 [Data] Add `max_concurrent_calls_per_actor` to ActorPoolStrategy

#65683 [common] BundleLocationIndex::Erase(node_id) leaves stale empty placement-group entries after all bundles on a node are removed

#61651 Supporting Host Name routing rather than primarily relying on IP address/port

#64696 [Core] Placement-group resource cleanup is abandoned after five transient RPC failures

#65677 [Core] Equal LabelSelectors can hash differently, leaking scheduling class ids

#65670 Path Traversal via unfiltered tar.extractall in file_transfer.py

#65672 [Core] Handle simultaneous poll and close completion in GCS AIO subscribers

#65673 [Serve] Cancel metrics pusher tasks that exceed graceful shutdown timeout

#65669 SSRF: databricks_uc_datasource.py unconditionally fetches external_url from server response

#65477 [serve][llm] s3:// model_source is dropped on Ray 2.57.0: engine config gets model_id instead of the staged local path

#32638 [Core][Feature request] Promote `actor_handle._actor_id` to public API

#65416 [Autoscaler][V2][KubeRay] Duplicate cloud_instance_ids reach terminate(), causing AssertionError and a permanent termination deadlock

#65454 [Core][KubeRay] Autoscaler sends all log records, including INFO, to stderr

#63924 [Data] AutoscalingCoordinator._reallocate_resources is O(R·N²) and holds its lock, starving concurrent datasets at scale (op_budget→0)

#65352 [Umbrella] Ray Sandboxing with gVisor

#65651 [Data] Migrate CSV Datasource to Datasource v2

#53793 [RLlib] Checkpointing fails with CUDA GPU learner using the new API stack

#64714 RDT: Enable ray.put from a non ray actor

#58679 [Rllib] new API stack uses >2X VRAM of old stack

#65648 [Core] ObjectRefGenerator async timeout polls leak waiter tasks

#65479 [Data] Fused map functions are serialized with every shuffle-map task

#65610 [RLlib] `Checkpointable.from_checkpoint` should have return type `typing.Self` instead of `Checkpointable`

#65640 [Data] MCAPDatasource: support file-level packed output to enable shuffle-free episode aggregation

#65641 [Data] MCAPDatasource: batched (streaming) message reads to reduce peak memory and enable intra-file pipelining

#64834 [Core][Data][CoreWorker]Discussion: remaining gRPC thread amplification with many task workers per node