[Core] Add ray-start option 'session-name'

dashboard/head.py

@@ -102,7 +102,7 @@ def __init__(
             gcs_address: The GCS address in the {address}:{port} format.
             log_dir: The log directory. E.g., /tmp/session_latest/logs.
             temp_dir: The temp directory. E.g., /tmp.
-            session_dir: The session directory. E.g., tmp/session_latest.
+            session_dir: The session directory. E.g., /tmp/session_latest.
             minimal: Whether or not it will load the minimal modules.
             serve_frontend: If configured, frontend HTML is
                 served from the dashboard.

doc/source/ray-core/configure.rst

@@ -95,12 +95,15 @@ If using the command line, connect to the Ray cluster as follow:
 Logging and Debugging
 ---------------------
 
-Each Ray session will have a unique name. By default, the name is
+Each Ray session will have a unique name. By default, the name is given automatically like 
 ``session_{timestamp}_{pid}``. The format of ``timestamp`` is
 ``%Y-%m-%d_%H-%M-%S_%f`` (See `Python time format <strftime.org>`__ for details);
 the pid belongs to the startup process (the process calling ``ray.init()`` or
 the Ray process executed by a shell in ``ray start``).
 
+Change the *session name* by passing ``--session-name={unique session name}`` to ``ray start``.
+If the given session name is not unique, the old data with the same session name may be overwritten.
+
 For each session, Ray will place all its temporary files under the
 *session directory*. A *session directory* is a subdirectory of the
 *root temporary path* (``/tmp/ray`` by default),
@@ -109,7 +112,7 @@ You can sort by their names to find the latest session.
 
 Change the *root temporary directory* by passing ``--temp-dir={your temp path}`` to ``ray start``.
 
-(There is not currently a stable way to change the root temporary directory when calling ``ray.init()``, but if you need to, you can provide the ``_temp_dir`` argument to ``ray.init()``.)
+(There is not currently a stable way to change the root temporary directory when calling ``ray.init()``, but if you need to, you can provide the ``_temp_dir`` argument to ``ray.init()``. The session name is also changable when calling ``ray.init()`` by providing ``_session_name`` argument.)
 
 Look :ref:`Logging Directory Structure <logging-directory-structure>` for more details.
 

python/ray/_private/node.py

@@ -459,6 +459,9 @@ def _init_temp(self):
 
         try_to_create_directory(self._temp_dir)
 
+        if self._ray_params.session_name is not None:
+            self._session_name = self._ray_params.session_name
+
         if self.head:
             self._session_dir = os.path.join(self._temp_dir, self._session_name)
         else:

python/ray/_private/worker.py

@@ -1373,6 +1373,8 @@ def init(
         _temp_dir: If provided, specifies the root temporary
             directory for the Ray process. Must be an absolute path. Defaults to an
             OS-specific conventional location, e.g., "/tmp/ray".
+        _session_name: The session name of the Ray process, only works when --head is 
+            specified. If it is not unique, the old data will be overwritten.
         _metrics_export_port: Port number Ray exposes system metrics
             through a Prometheus endpoint. It is currently under active
             development, and the API is subject to change.
@@ -1425,6 +1427,7 @@ def init(
         "_redis_password", ray_constants.REDIS_DEFAULT_PASSWORD
     )
     _temp_dir: Optional[str] = kwargs.pop("_temp_dir", None)
+    _session_name: Optional[str] = kwargs.pop("_session_name", None)
     _metrics_export_port: Optional[int] = kwargs.pop("_metrics_export_port", None)
     _system_config: Optional[Dict[str, str]] = kwargs.pop("_system_config", None)
     _tracing_startup_hook: Optional[Callable] = kwargs.pop(
@@ -1658,6 +1661,7 @@ def sigterm_handler(signum, frame):
             redis_max_memory=_redis_max_memory,
             plasma_store_socket_name=None,
             temp_dir=_temp_dir,
+            session_name=_session_name,
             storage=storage,
             _system_config=_system_config,
             enable_object_reconstruction=_enable_object_reconstruction,
@@ -1727,6 +1731,7 @@ def sigterm_handler(signum, frame):
             redis_password=_redis_password,
             object_ref_seed=None,
             temp_dir=_temp_dir,
+            session_name=_session_name,
             _system_config=_system_config,
             enable_object_reconstruction=_enable_object_reconstruction,
             metrics_export_port=_metrics_export_port,

python/ray/scripts/scripts.py

@@ -504,6 +504,12 @@ def debug(address):
     help="manually specify the root temporary dir of the Ray process, only "
     "works when --head is specified",
 )
+@click.option(
+    "--session-name",
+    default=None,
+    help="The session name of the Ray process, only works when --head "
+    "is specified. If it is not unique, the old data will be overwritten.",
+)
 @click.option(
     "--storage",
     default=None,
@@ -603,6 +609,7 @@ def start(
     plasma_store_socket_name,
     raylet_socket_name,
     temp_dir,
+    session_name,
     storage,
     system_config,
     enable_object_reconstruction,
@@ -653,6 +660,13 @@ def start(
             "All the worker nodes will use the same temp_dir as a head node. "
         )
         temp_dir = None
+    if session_name and not head:
+        cli_logger.warning(
+            f"`--session-name={session_name}` option will be ignored. "
+            "`--head` is a required flag to use `--session-name`. "
+            "All the worker nodes will use the same session_name as a head node. "
+        )
+        session_name = None
 
     redirect_output = None if not no_redirect_output else True
 
@@ -687,6 +701,7 @@ def start(
         plasma_store_socket_name=plasma_store_socket_name,
         raylet_socket_name=raylet_socket_name,
         temp_dir=temp_dir,
+        session_name=session_name,
         storage=storage,
         include_dashboard=include_dashboard,
         dashboard_host=dashboard_host,