Add descriptions for UCX config options

distributed/distributed-schema.yaml

@@ -803,44 +803,51 @@ properties:
   rmm:
     type: object
     description: |
-      Configuration options for the RAPIDS Memory Manager
+      Configuration options for the RAPIDS Memory Manager.
     properties:
       pool-size:
-        type:
-        - integer
-        - "null"
-        description:
-          The size of the memory pool in bytes
+        type: [integer, 'null']
+        description: |
+          The size of the memory pool in bytes.
   ucx:
     type: object
     description: |
-      UCX provides access to other network interconnects like Infiniband and NVLINK
+      UCX provides access to other transport methods including NVLink and InfiniBand.
     properties:
+      cuda_copy:
+        type: boolean
+        description: |
+          Set environment variables to enable CUDA support over UCX. This may be used even if
+          InfiniBand and NVLink are not supported or disabled, then transferring data over TCP.
       tcp:
-        type:
-        - boolean
-        - "null"
+        type: boolean
+        description: |
+          Set environment variables to enable TCP over UCX, even if InfiniBand and NVLink
+          are not supported or disabled.
       nvlink:
-        type:
-        - boolean
-        - "null"
+        type: boolean
+        description: |
+          Set environment variables to enable UCX over NVLink, implies ``ucx.tcp=True``.
       infiniband:
-        type:
-        - boolean
-        - "null"
+        type: boolean
+        description: |
+          Set environment variables to enable UCX over InfiniBand, implies ``ucx.tcp=True``.
       rdmacm:
-        type:
-        - boolean
-        - "null"
-      cuda_copy:
-        type:
-        - boolean
-        - "null"
+        type: boolean
+        description: |
+          Set environment variables to enable UCX RDMA connection manager support,
+          requires ``ucx.infiniband=True``.
       net-devices:
-        type:
-        - string
-        - "null"
-        description: Define which Infiniband device to use
+        type: [string, 'null']
+        description: |
+          Interface(s) used by workers for UCX communication. Can be a string (like
+          ``"eth0"`` for NVLink or ``"mlx5_0:1"``/``"ib0"`` for InfiniBand), ``"auto"``
+          (requires ``ucx.infiniband=True``) to pick the optimal interface per-worker based on
+          the system's topology, or ``None`` to stay with the default value of ``"all"`` (use
+          all available interfaces). Setting to ``"auto"`` requires UCX-Py to be installed
+          and compiled with hwloc support. Unexpected errors can occur when using
+          ``"auto"`` if any interfaces are disconnected or improperly configured.
       reuse-endpoints:
         type: boolean
-        description: Whether to reuse endpoints or not, default True
+        description: | 
+          Whether to reuse endpoints or not.

distributed/distributed.yaml

@@ -188,10 +188,10 @@ distributed:
 rmm:
   pool-size: null
 ucx:
-  tcp: null  # enable tcp
-  nvlink: null  # enable cuda_ipc
-  infiniband: null # enable Infiniband
-  rdmacm: null # enable RDMACM
-  cuda_copy: null  # enable cuda-copy
-  net-devices: null  # define which Infiniband device to use
+  cuda_copy: False  # enable cuda-copy
+  tcp: False  # enable tcp
+  nvlink: False  # enable cuda_ipc
+  infiniband: False # enable Infiniband
+  rdmacm: False # enable RDMACM
+  net-devices: null  # define what interface to use for UCX comm
   reuse-endpoints: True  # enable endpoint reuse