chore: import upstream snapshot with attribution

python/ray/autoscaler/spark/defaults.yaml

@@ -0,0 +1,49 @@
+# Example command to start a cluster with this config:
+#
+# ray start --autoscaling-config=default.yaml --head --block
+#
+cluster_name: spark
+max_workers: 8
+provider:
+    type: spark
+    # This must be true since the nodes share the same ip!
+    use_node_id_as_ip: True
+    disable_node_updaters: True
+    disable_launch_config_check: True
+available_node_types:
+    ray.head.default:
+        # You must set this manually to your "head" node resources!! The head
+        # node is launched via `ray start` and hence the autoscaler cannot
+        # configure its resources. The resources specified for its node type
+        # must line up with what Ray detects/is configured with on start.
+        resources:
+            CPU: 8   # <-- set this to num CPUs used/detected in `ray start`
+            GPU: 0   # <-- set this to num GPUs used/detected in `ray start`
+        node_config: {}
+        max_workers: 0
+    ray.worker:
+        resources:
+            CPU: 1
+            object_store_memory: 1000000000
+        node_config: {}
+        min_workers: 0
+        max_workers: 4
+head_node_type: ray.head.default
+upscaling_speed: 1.0
+idle_timeout_minutes: 1.0
+#
+# !!! Configurations below are not supported in spark cluster mode
+#
+auth: {}
+docker: {}
+initialization_commands: []
+setup_commands: []
+head_setup_commands: []
+worker_setup_commands: []
+head_start_ray_commands: []
+worker_start_ray_commands: []
+file_mounts: {}
+cluster_synced_files: []
+file_mounts_sync_continuously: false
+rsync_exclude: []
+rsync_filter: []