Add a web dashboard for monitoring node resource usage (#4066)

python/ray/node.py

@@ -10,6 +10,7 @@ import json
 import os
 import logging
 import signal
+import sys
 import tempfile
 import threading
 import time
@@ -24,6 +25,8 @@ from ray.utils import try_to_create_directory
 # using logging.basicConfig in its entry/init points.
 logger = logging.getLogger(__name__)
 
+PY3 = sys.version_info.major >= 3
+
 
 class Node(object):
     """An encapsulation of the Ray processes on a single node.
@@ -81,6 +84,7 @@ class Node(object):
             self._plasma_store_socket_name = None
             self._raylet_socket_name = None
             self._webui_url = None
+            self._dashboard_url = None
         else:
             self._plasma_store_socket_name = (
                 ray_params.plasma_store_socket_name)
@@ -284,6 +288,35 @@ class Node(object):
             process_info
         ]
 
+    def start_reporter(self):
+        """Start the reporter."""
+        stdout_file, stderr_file = self.new_log_files("reporter", True)
+        process_info = ray.services.start_reporter(
+            self.redis_address,
+            stdout_file=stdout_file,
+            stderr_file=stderr_file,
+            redis_password=self._ray_params.redis_password)
+        assert ray_constants.PROCESS_TYPE_REPORTER not in self.all_processes
+        if process_info is not None:
+            self.all_processes[ray_constants.PROCESS_TYPE_REPORTER] = [
+                process_info
+            ]
+
+    def start_dashboard(self):
+        """Start the dashboard."""
+        stdout_file, stderr_file = self.new_log_files("dashboard", True)
+        self._dashboard_url, process_info = ray.services.start_dashboard(
+            self.redis_address,
+            self._temp_dir,
+            stdout_file=stdout_file,
+            stderr_file=stderr_file,
+            redis_password=self._ray_params.redis_password)
+        assert ray_constants.PROCESS_TYPE_DASHBOARD not in self.all_processes
+        if process_info is not None:
+            self.all_processes[ray_constants.PROCESS_TYPE_DASHBOARD] = [
+                process_info
+            ]
+
     def start_ui(self):
         """Start the web UI."""
         stdout_file, stderr_file = self.new_log_files("webui")
@@ -408,14 +441,16 @@ class Node(object):
             self.start_redis()
             self.start_monitor()
             self.start_raylet_monitor()
+            if PY3 and self._ray_params.include_webui:
+                self.start_dashboard()
 
         self.start_plasma_store()
         self.start_raylet()
+        if PY3 and self._ray_params.include_webui:
+            self.start_reporter()
 
         if self._ray_params.include_log_monitor:
             self.start_log_monitor()
-        if self._ray_params.include_webui:
-            self.start_ui()
 
     def _kill_process_type(self,
                            process_type,
@@ -545,6 +580,26 @@ class Node(object):
         self._kill_process_type(
             ray_constants.PROCESS_TYPE_LOG_MONITOR, check_alive=check_alive)
 
+    def kill_reporter(self, check_alive=True):
+        """Kill the reporter.
+
+        Args:
+            check_alive (bool): Raise an exception if the process was already
+                dead.
+        """
+        self._kill_process_type(
+            ray_constants.PROCESS_TYPE_REPORTER, check_alive=check_alive)
+
+    def kill_dashboard(self, check_alive=True):
+        """Kill the dashboard.
+
+        Args:
+            check_alive (bool): Raise an exception if the process was already
+                dead.
+        """
+        self._kill_process_type(
+            ray_constants.PROCESS_TYPE_DASHBOARD, check_alive=check_alive)
+
     def kill_monitor(self, check_alive=True):
         """Kill the monitor.