Use events instead of sleep for result reading

Instead of creating a one-time use thread for every host/task

lib/ansible/executor/process/threading.py

@@ -0,0 +1,116 @@
+# (c) 2012-2014, Michael DeHaan <michael.dehaan@gmail.com>
+#
+# This file is part of Ansible
+#
+# Ansible is free software: you can redistribute it and/or modify
+# it under the terms of the GNU General Public License as published by
+# the Free Software Foundation, either version 3 of the License, or
+# (at your option) any later version.
+#
+# Ansible is distributed in the hope that it will be useful,
+# but WITHOUT ANY WARRANTY; without even the implied warranty of
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+# GNU General Public License for more details.
+#
+# You should have received a copy of the GNU General Public License
+# along with Ansible.  If not, see <http://www.gnu.org/licenses/>.
+
+# Make coding more python3-ish
+from __future__ import (absolute_import, division, print_function)
+__metaclass__ = type
+
+import os
+import sys
+import time
+import traceback
+
+from jinja2.exceptions import TemplateNotFound
+
+from ansible.errors import AnsibleConnectionFailure
+from ansible.executor.task_executor import TaskExecutor
+from ansible.executor.task_result import TaskResult
+from ansible.module_utils._text import to_text
+
+try:
+    from __main__ import display
+except ImportError:
+    from ansible.utils.display import Display
+    display = Display()
+
+__all__ = ['WorkerProcess']
+
+
+def run_worker(tqm, shared_loader_obj):
+    '''
+    The worker thread class, which uses TaskExecutor to run tasks
+    read from a job queue and pushes results into a results queue
+    for reading later.
+    '''
+
+    # import cProfile, pstats, StringIO
+    # pr = cProfile.Profile()
+    # pr.enable()
+
+    display.debug("STARTING WORKER")
+    while not tqm._terminated:
+        job = tqm.get_job()
+        if job is None:
+            time.sleep(0.0001)
+            continue
+
+        display.debug("WORKER GOT A JOB")
+        (host, task, play_context, task_vars) = job
+
+        try:
+            # execute the task and build a TaskResult from the result
+            display.debug("running TaskExecutor() for %s/%s" % (host, task))
+            executor_result = TaskExecutor(
+                host,
+                task,
+                task_vars,
+                play_context,
+                None, #new_stdin
+                tqm._loader,
+                shared_loader_obj,
+                tqm, #rslt_q
+            ).run()
+
+            display.debug("done running TaskExecutor() for %s/%s" % (host, task))
+
+            # put the result on the result queue
+            display.debug("sending task result")
+            tqm.put_result(TaskResult(
+                host,
+                task,
+                executor_result,
+            ))
+            display.debug("done task result")
+
+        except AnsibleConnectionFailure:
+            tqm.put_result(TaskResult(
+                host,
+                task,
+                dict(unreachable=True),
+            ))
+
+        except Exception as e:
+            if not isinstance(e, (IOError, EOFError, KeyboardInterrupt, SystemExit)) or isinstance(e, TemplateNotFound):
+                try:
+                    tqm.put_result(TaskResult(
+                        host,
+                        task,
+                        dict(failed=True, exception=to_text(traceback.format_exc()), stdout=''),
+                    ))
+                except:
+                    display.debug(u"WORKER EXCEPTION: %s" % to_text(e))
+                    display.debug(u"WORKER TRACEBACK: %s" % to_text(traceback.format_exc()))
+
+    # pr.disable()
+    # s = StringIO.StringIO()
+    # sortby = 'time'
+    # ps = pstats.Stats(pr, stream=s).sort_stats(sortby)
+    # ps.print_stats()
+    # with open('worker_%06d.stats' % os.getpid(), 'w') as f:
+    #     f.write(s.getvalue())
+
+    display.debug("WORKER PROCESS EXITING")

lib/ansible/executor/task_executor.py

@@ -299,14 +299,13 @@ class TaskExecutor:
                 templar = Templar(loader=self._loader, shared_loader_obj=self._shared_loader_obj, variables=self._job_vars)
                 res['_ansible_item_label'] = templar.template(label)
 
-            self._rslt_q.put(
+            self._rslt_q.put_result(
                 TaskResult(
-                    self._host.name,
+                    self._host,
-                    self._task._uuid,
+                    self._task,
                     res,
                     task_fields=self._task.dump_attrs(),
                 ),
-                block=False,
             )
             results.append(res)
             del task_vars[loop_var]
@@ -591,7 +590,7 @@ class TaskExecutor:
                         result['_ansible_retry'] = True
                         result['retries'] = retries
                         display.debug('Retrying task, attempt %d of %d' % (attempt, retries))
-                        self._rslt_q.put(TaskResult(self._host.name, self._task._uuid, result, task_fields=self._task.dump_attrs()), block=False)
+                        self._rslt_q.put_result(TaskResult(self._host, self._task, result, task_fields=self._task.dump_attrs()))
                         time.sleep(delay)
         else:
             if retries > 1:

lib/ansible/executor/task_queue_manager.py

@@ -19,19 +19,22 @@
 from __future__ import (absolute_import, division, print_function)
 __metaclass__ = type
 
-import multiprocessing
+import threading
 import os
 import tempfile
 
+from collections import deque
+
 from ansible import constants as C
 from ansible.errors import AnsibleError
 from ansible.executor.play_iterator import PlayIterator
+from ansible.executor.process.threading import run_worker
 from ansible.executor.stats import AggregateStats
 from ansible.module_utils.six import string_types
 from ansible.module_utils._text import to_text
 from ansible.playbook.block import Block
 from ansible.playbook.play_context import PlayContext
-from ansible.plugins.loader import callback_loader, strategy_loader, module_loader
+from ansible.plugins.loader import action_loader, connection_loader, filter_loader, test_loader, lookup_loader, callback_loader, strategy_loader, module_loader
 from ansible.plugins.callback import CallbackBase
 from ansible.template import Templar
 from ansible.utils.helpers import pct_to_int
@@ -48,6 +51,23 @@ except ImportError:
 __all__ = ['TaskQueueManager']
 
 
+# TODO: this should probably be in the plugins/__init__.py, with
+#       a smarter mechanism to set all of the attributes based on
+#       the loaders created there
+class SharedPluginLoaderObj:
+    '''
+    A simple object to make pass the various plugin loaders to
+    the forked processes over the queue easier
+    '''
+    def __init__(self):
+        self.action_loader = action_loader
+        self.connection_loader = connection_loader
+        self.filter_loader = filter_loader
+        self.test_loader   = test_loader
+        self.lookup_loader = lookup_loader
+        self.module_loader = module_loader
+
+
 class TaskQueueManager:
 
     '''
@@ -100,18 +120,59 @@ class TaskQueueManager:
         self._failed_hosts = dict()
         self._unreachable_hosts = dict()
 
-        self._final_q = multiprocessing.Queue()
-
         # A temporary file (opened pre-fork) used by connection
         # plugins for inter-process locking.
         self._connection_lockfile = tempfile.TemporaryFile()
 
+        self._job_queue = deque()
+        self._job_queue_lock = threading.Lock()
+
+        self._res_queue = deque()
+        self._res_queue_lock = threading.Lock()
+        self._res_ready = threading.Event()
+
+    def _put_in_queue(self, data, queue, lock):
+        lock.acquire()
+        queue.appendleft(data)
+        lock.release()
+
+    def _pop_off_queue(self, queue, lock):
+        try:
+            data = None
+            lock.acquire()
+            data = queue.pop()
+        except:
+            pass
+        finally:
+            lock.release()
+        return data
+
+    def put_job(self, data):
+        self._put_in_queue(data, self._job_queue, self._job_queue_lock)
+
+    def get_job(self):
+        return self._pop_off_queue(self._job_queue, self._job_queue_lock)
+
+    def put_result(self, data):
+        self._put_in_queue(data, self._res_queue, self._res_queue_lock)
+        self._res_ready.set()
+
+    def get_result(self):
+        return self._pop_off_queue(self._res_queue, self._res_queue_lock)
+
     def _initialize_processes(self, num):
+        # FIXME: do we need a global lock for workers here instead of a per-worker?
         self._workers = []
 
+        # create a dummy object with plugin loaders set as an easier
+        # way to share them with the forked processes
+        shared_loader_obj = SharedPluginLoaderObj()
+
         for i in range(num):
-            rslt_q = multiprocessing.Queue()
+            w_thread = threading.Thread(target=run_worker, args=(self, shared_loader_obj))
-            self._workers.append([None, rslt_q])
+            w_thread.start()
+            w_lock = threading.Lock()
+            self._workers.append([w_thread, w_lock])
 
     def _initialize_notified_handlers(self, play):
         '''
@@ -312,18 +373,13 @@ class TaskQueueManager:
     def cleanup(self):
         display.debug("RUNNING CLEANUP")
         self.terminate()
-        self._final_q.close()
         self._cleanup_processes()
 
     def _cleanup_processes(self):
         if hasattr(self, '_workers'):
-            for (worker_prc, rslt_q) in self._workers:
+            for (w_thread, w_lock) in self._workers:
-                rslt_q.close()
+                if w_thread and not w_thread.is_alive():
-                if worker_prc and worker_prc.is_alive():
+                    w_thread.join()
-                    try:
-                        worker_prc.terminate()
-                    except AttributeError:
-                        pass
 
     def clear_failed_hosts(self):
         self._failed_hosts = dict()

lib/ansible/plugins/loader.py

@@ -15,6 +15,7 @@ import sys
 import warnings
 
 from collections import defaultdict
+from threading import Lock
 
 from ansible import constants as C
 from ansible.plugins import get_plugin_class, MODULE_CACHE, PATH_CACHE, PLUGIN_PATH_CACHE
@@ -73,6 +74,7 @@ class PluginLoader:
 
         self._extra_dirs = []
         self._searched_paths = set()
+        self._lock = Lock()
 
     def __setstate__(self, data):
         '''
@@ -360,11 +362,14 @@ class PluginLoader:
         if path is None:
             return None
 
+        self._lock.acquire()
         if path not in self._module_cache:
             self._module_cache[path] = self._load_module_source(name, path)
             found_in_cache = False
 
         obj = getattr(self._module_cache[path], self.class_name)
+        self._lock.release()
+
         if self.base_class:
             # The import path is hardcoded and should be the right place,
             # so we are not expecting an ImportError.
@@ -423,15 +428,18 @@ class PluginLoader:
                 yield path
                 continue
 
-            if path not in self._module_cache:
-                self._module_cache[path] = self._load_module_source(name, path)
-                found_in_cache = False
-
             try:
+                self._lock.acquire()
+                if path not in self._module_cache:
+                    self._module_cache[path] = self._load_module_source(name, path)
+                    found_in_cache = False
+
                 obj = getattr(self._module_cache[path], self.class_name)
             except AttributeError as e:
                 display.warning("Skipping plugin (%s) as it seems to be invalid: %s" % (path, to_text(e)))
                 continue
+            finally:
+                self._lock.release()
 
             if self.base_class:
                 # The import path is hardcoded and should be the right place,

lib/ansible/plugins/strategy/__init__.py

@@ -24,13 +24,11 @@ import threading
 import time
 
 from collections import deque
-from multiprocessing import Lock
 from jinja2.exceptions import UndefinedError
 
 from ansible import constants as C
 from ansible.errors import AnsibleError, AnsibleParserError, AnsibleUndefinedVariable
 from ansible.executor import action_write_locks
-from ansible.executor.process.worker import WorkerProcess
 from ansible.executor.task_result import TaskResult
 from ansible.inventory.host import Host
 from ansible.module_utils.six.moves import queue as Queue
@@ -55,45 +53,6 @@ except ImportError:
 __all__ = ['StrategyBase']
 
 
-class StrategySentinel:
-    pass
-
-
-# TODO: this should probably be in the plugins/__init__.py, with
-#       a smarter mechanism to set all of the attributes based on
-#       the loaders created there
-class SharedPluginLoaderObj:
-    '''
-    A simple object to make pass the various plugin loaders to
-    the forked processes over the queue easier
-    '''
-    def __init__(self):
-        self.action_loader = action_loader
-        self.connection_loader = connection_loader
-        self.filter_loader = filter_loader
-        self.test_loader = test_loader
-        self.lookup_loader = lookup_loader
-        self.module_loader = module_loader
-
-_sentinel = StrategySentinel()
-
-
-def results_thread_main(strategy):
-    while True:
-        try:
-            result = strategy._final_q.get()
-            if isinstance(result, StrategySentinel):
-                break
-            else:
-                strategy._results_lock.acquire()
-                strategy._results.append(result)
-                strategy._results_lock.release()
-        except (IOError, EOFError):
-            break
-        except Queue.Empty:
-            pass
-
-
 class StrategyBase:
 
     '''
@@ -102,16 +61,15 @@ class StrategyBase:
     '''
 
     def __init__(self, tqm):
-        self._tqm = tqm
+        self._tqm                = tqm
-        self._inventory = tqm.get_inventory()
+        self._inventory          = tqm.get_inventory()
-        self._workers = tqm.get_workers()
+        self._workers            = tqm._workers
-        self._notified_handlers = tqm._notified_handlers
+        self._notified_handlers  = tqm._notified_handlers
         self._listening_handlers = tqm._listening_handlers
-        self._variable_manager = tqm.get_variable_manager()
+        self._variable_manager   = tqm.get_variable_manager()
-        self._loader = tqm.get_loader()
+        self._loader             = tqm.get_loader()
-        self._final_q = tqm._final_q
+        self._step               = getattr(tqm._options, 'step', False)
-        self._step = getattr(tqm._options, 'step', False)
+        self._diff               = getattr(tqm._options, 'diff', False)
-        self._diff = getattr(tqm._options, 'diff', False)
 
         # Backwards compat: self._display isn't really needed, just import the global display and use that.
         self._display = display
@@ -124,17 +82,8 @@ class StrategyBase:
         # outstanding tasks still in queue
         self._blocked_hosts = dict()
 
-        self._results = deque()
-        self._results_lock = threading.Condition(threading.Lock())
-
-        # create the result processing thread for reading results in the background
-        self._results_thread = threading.Thread(target=results_thread_main, args=(self,))
-        self._results_thread.daemon = True
-        self._results_thread.start()
-
     def cleanup(self):
-        self._final_q.put(_sentinel)
+        self._tqm.terminate()
-        self._results_thread.join()
 
     def run(self, iterator, play_context, result=0):
         # execute one more pass through the iterator without peeking, to
@@ -203,38 +152,11 @@ class StrategyBase:
 
         if task.action not in action_write_locks.action_write_locks:
             display.debug('Creating lock for %s' % task.action)
-            action_write_locks.action_write_locks[task.action] = Lock()
+            action_write_locks.action_write_locks[task.action] = threading.Lock()
 
-        # and then queue the new task
+        self._tqm.put_job((host, task, play_context, task_vars))
-        try:
+        self._pending_results += 1
 
-            # create a dummy object with plugin loaders set as an easier
-            # way to share them with the forked processes
-            shared_loader_obj = SharedPluginLoaderObj()
-
-            queued = False
-            starting_worker = self._cur_worker
-            while True:
-                (worker_prc, rslt_q) = self._workers[self._cur_worker]
-                if worker_prc is None or not worker_prc.is_alive():
-                    worker_prc = WorkerProcess(self._final_q, task_vars, host, task, play_context, self._loader, self._variable_manager, shared_loader_obj)
-                    self._workers[self._cur_worker][0] = worker_prc
-                    worker_prc.start()
-                    display.debug("worker is %d (out of %d available)" % (self._cur_worker + 1, len(self._workers)))
-                    queued = True
-                self._cur_worker += 1
-                if self._cur_worker >= len(self._workers):
-                    self._cur_worker = 0
-                if queued:
-                    break
-                elif self._cur_worker == starting_worker:
-                    time.sleep(0.0001)
-
-            self._pending_results += 1
-        except (EOFError, IOError, AssertionError) as e:
-            # most likely an abort
-            display.debug("got an error while queuing: %s" % e)
-            return
         display.debug("exiting _queue_task() for %s/%s" % (host.name, task.action))
 
     def get_task_hosts(self, iterator, task_host, task):
@@ -325,23 +247,21 @@ class StrategyBase:
 
         cur_pass = 0
         while True:
-            try:
+            task_result = self._tqm.get_result()
-                self._results_lock.acquire()
+            if task_result is None:
-                task_result = self._results.popleft()
-            except IndexError:
                 break
-            finally:
-                self._results_lock.release()
 
             # get the original host and task. We then assign them to the TaskResult for use in callbacks/etc.
-            original_host = get_original_host(task_result._host)
+            #original_host = get_original_host(task_result._host)
-            found_task = iterator.get_original_task(original_host, task_result._task)
+            #found_task = iterator.get_original_task(original_host, task_result._task)
-            original_task = found_task.copy(exclude_parent=True, exclude_tasks=True)
+            #original_task = found_task.copy(exclude_parent=True, exclude_tasks=True)
-            original_task._parent = found_task._parent
+            #original_task._parent = found_task._parent
-            original_task.from_attrs(task_result._task_fields)
+            #original_task.from_attrs(task_result._task_fields)
 
-            task_result._host = original_host
+            #task_result._host = original_host
-            task_result._task = original_task
+            #task_result._task = original_task
+            original_host = task_result._host
+            original_task = task_result._task
 
             # get the correct loop var for use later
             if original_task.loop_control:
@@ -586,7 +506,9 @@ class StrategyBase:
             results = self._process_pending_results(iterator)
             ret_results.extend(results)
             if self._pending_results > 0:
-                time.sleep(C.DEFAULT_INTERNAL_POLL_INTERVAL)
+                #time.sleep(C.DEFAULT_INTERNAL_POLL_INTERVAL)
+                self._tqm._res_ready.wait()
+                self._tqm._res_ready.clear()
 
         display.debug("no more pending results, returning what we have")