Set the process title in workers and actors (#3219)

.travis/format.sh

@@ -51,6 +51,11 @@ format_changed() {
     if ! git diff --diff-filter=ACM --quiet --exit-code "$MERGEBASE" -- '*.py' &>/dev/null; then
         git diff --name-only --diff-filter=ACM "$MERGEBASE" -- '*.py' | xargs -P 5 \
              yapf --in-place "${YAPF_EXCLUDES[@]}" "${YAPF_FLAGS[@]}"
+        if which flake8 >/dev/null; then
+            git diff --name-only --diff-filter=ACM "$MERGEBASE" -- '*.py' | xargs -P 5 \
+                 flake8 --exclude=python/ray/core/generated/,doc/source/conf.py,python/ray/cloudpickle/ \
+                    --ignore=C408,E121,E123,E126,E226,E24,E704,W503,W504,W605
+        fi
     fi
 
     if which clang-format >/dev/null; then

.travis/install-dependencies.sh

@@ -25,7 +25,7 @@ if [[ "$PYTHON" == "2.7" ]] && [[ "$platform" == "linux" ]]; then
   bash miniconda.sh -b -p $HOME/miniconda
   export PATH="$HOME/miniconda/bin:$PATH"
   pip install -q cython==0.27.3 cmake tensorflow gym opencv-python pyyaml pandas==0.22 requests \
-    feather-format lxml openpyxl xlrd py-spy
+    feather-format lxml openpyxl xlrd py-spy setproctitle
 elif [[ "$PYTHON" == "3.5" ]] && [[ "$platform" == "linux" ]]; then
   sudo apt-get update
   sudo apt-get install -y cmake pkg-config python-dev python-numpy build-essential autoconf curl libtool unzip
@@ -34,7 +34,7 @@ elif [[ "$PYTHON" == "3.5" ]] && [[ "$platform" == "linux" ]]; then
   bash miniconda.sh -b -p $HOME/miniconda
   export PATH="$HOME/miniconda/bin:$PATH"
   pip install -q cython==0.27.3 cmake tensorflow gym opencv-python pyyaml pandas==0.22 requests \
-    feather-format lxml openpyxl xlrd py-spy
+    feather-format lxml openpyxl xlrd py-spy setproctitle
 elif [[ "$PYTHON" == "2.7" ]] && [[ "$platform" == "macosx" ]]; then
   # check that brew is installed
   which -s brew
@@ -51,7 +51,7 @@ elif [[ "$PYTHON" == "2.7" ]] && [[ "$platform" == "macosx" ]]; then
   bash miniconda.sh -b -p $HOME/miniconda
   export PATH="$HOME/miniconda/bin:$PATH"
   pip install -q cython==0.27.3 cmake tensorflow gym opencv-python pyyaml pandas==0.22 requests \
-    feather-format lxml openpyxl xlrd py-spy
+    feather-format lxml openpyxl xlrd py-spy setproctitle
 elif [[ "$PYTHON" == "3.5" ]] && [[ "$platform" == "macosx" ]]; then
   # check that brew is installed
   which -s brew
@@ -68,7 +68,7 @@ elif [[ "$PYTHON" == "3.5" ]] && [[ "$platform" == "macosx" ]]; then
   bash miniconda.sh -b -p $HOME/miniconda
   export PATH="$HOME/miniconda/bin:$PATH"
   pip install -q cython==0.27.3 cmake tensorflow gym opencv-python pyyaml pandas==0.22 requests \
-    feather-format lxml openpyxl xlrd py-spy
+    feather-format lxml openpyxl xlrd py-spy setproctitle
 elif [[ "$LINT" == "1" ]]; then
   sudo apt-get update
   sudo apt-get install -y cmake build-essential autoconf curl libtool unzip

doc/requirements-doc.txt

@@ -9,6 +9,7 @@ pyarrow
 pyyaml
 recommonmark
 redis
+setproctitle
 sphinx
 sphinx-click
 sphinx_rtd_theme

python/ray/rllib/test/test_policy_evaluator.py

@@ -165,11 +165,14 @@ class TestPolicyEvaluator(unittest.TestCase):
                 },
             })
         pg.train()
-        self.assertEqual(counts["sample"], 1)
+        pg.train()
+        pg.train()
+        pg.train()
+        self.assertEqual(counts["sample"], 4)
         self.assertGreater(counts["start"], 0)
         self.assertGreater(counts["end"], 0)
-        self.assertGreater(counts["step"], 50)
-        self.assertLess(counts["step"], 100)
+        self.assertGreater(counts["step"], 200)
+        self.assertLess(counts["step"], 400)
 
     def testQueryEvaluators(self):
         register_env("test", lambda _: gym.make("CartPole-v0"))

python/ray/scripts/scripts.py

@@ -388,6 +388,12 @@ def stop():
             "grep -v grep | awk '{ print $2 }') 2> /dev/null"
         ],
         shell=True)
+    subprocess.call(
+        [
+            "kill -9 $(ps aux | grep ' ray_' | "
+            "grep -v grep | awk '{ print $2 }') 2> /dev/null"
+        ],
+        shell=True)
 
     # Find the PID of the Ray log monitor process and kill it.
     subprocess.call(
@@ -591,7 +597,7 @@ export IFS="
 # Call sudo to prompt for password before anything has been printed.
 sudo true
 workers=$(
-    ps aux | grep default_worker.py | grep -v grep | grep -v raylet/raylet
+    ps aux | grep ' ray_' | grep -v grep
 )
 for worker in $workers; do
     echo "Stack dump for $worker";

python/ray/worker.py

@@ -2,6 +2,7 @@ from __future__ import absolute_import
 from __future__ import division
 from __future__ import print_function
 
+from contextlib import contextmanager
 import atexit
 import colorama
 import hashlib
@@ -10,6 +11,7 @@ import logging
 import numpy as np
 import os
 import redis
+import setproctitle
 import signal
 import sys
 import threading
@@ -899,8 +901,15 @@ class Worker(object):
                 "name": function_name,
                 "task_id": task.task_id().hex()
             }
+            if task.actor_id().id() == NIL_ACTOR_ID:
+                title = "ray_worker:{}()".format(function_name)
+            else:
+                actor = self.actors[task.actor_id().id()]
+                title = "ray_{}:{}()".format(actor.__class__.__name__,
+                                             function_name)
             with profiling.profile("task", extra_data=extra_data, worker=self):
-                self._process_task(task, execution_info)
+                with _changeproctitle(title):
+                    self._process_task(task, execution_info)
 
         # Increase the task execution counter.
         self.function_actor_manager.increase_task_counter(
@@ -1844,6 +1853,7 @@ def connect(info,
     # Initialize some fields.
     if mode is WORKER_MODE:
         worker.worker_id = random_string()
+        setproctitle.setproctitle("ray_worker")
     else:
         # This is the code path of driver mode.
         if driver_id is None:
@@ -2087,6 +2097,14 @@ def disconnect(worker=global_worker):
     worker.serialization_context_map.clear()
 
 
+@contextmanager
+def _changeproctitle(title):
+    old_title = setproctitle.getproctitle()
+    setproctitle.setproctitle(title)
+    yield
+    setproctitle.setproctitle(old_title)
+
+
 def _try_to_compute_deterministic_class_id(cls, depth=5):
     """Attempt to produce a deterministic class ID for a given class.
 

python/setup.py

@@ -151,6 +151,7 @@ setup(
         "pytest",
         "pyyaml",
         "redis",
+        "setproctitle",
         # The six module is required by pyarrow.
         "six >= 1.0.0",
         "flatbuffers"

test/runtest.py

@@ -4,6 +4,7 @@ from __future__ import print_function
 
 import os
 import re
+import setproctitle
 import string
 import subprocess
 import sys
@@ -2299,6 +2300,26 @@ def test_wait_reconstruction(shutdown_only):
     assert len(ready_ids) == 1
 
 
+def test_ray_setproctitle(shutdown_only):
+    ray.init(num_cpus=2)
+
+    @ray.remote
+    class UniqueName(object):
+        def __init__(self):
+            assert setproctitle.getproctitle() == "ray_UniqueName:__init__()"
+
+        def f(self):
+            assert setproctitle.getproctitle() == "ray_UniqueName:f()"
+
+    @ray.remote
+    def unique_1():
+        assert setproctitle.getproctitle() == "ray_worker:runtest.unique_1()"
+
+    actor = UniqueName.remote()
+    ray.get(actor.f.remote())
+    ray.get(unique_1.remote())
+
+
 @pytest.mark.skipif(
     os.getenv("TRAVIS") is None,
     reason="This test should only be run on Travis.")