drain runner queue and undo cancel_join_thread

Sometimes the runners can hang.  Initially
debugging lead to the queue join thread, so I thought
we could cancel all the join threads and everything would be okay.

But it turns out canceling the queue join threads can lead
to corruption of the queues, so when we go to drain the queues
the task hangs.

But it also turns out that we were not properly draining
the queues in the task process.  We were waiting for all
the runners to exit, then draining the queues.

This is bad and will cause the queues to fill up and hang
and/or drop data or corrupt the queues.

The proper fix seems to be to draining the queues in a
loop before calling join with a timeout.

Also modified the queue drain loops to no block on queue.get()

Revert "cancel all queue join threads"

This reverts commit 75c0e3a54b8f6e8fd77c7d9d95decab830159929.

Revert "duration runner: add teardown and cancel all queue join threads"

This reverts commit 7eb6abb6931b24e085b139cc3500f4497cdde57d.

Change-Id: Ic4f8e814cf23615621c1250535967716b425ac18
Signed-off-by: Ross Brattain <ross.b.brattain@intel.com>

6 files changed, 95 insertions, 16 deletions

diff --git a/yardstick/benchmark/runners/arithmetic.py b/yardstick/benchmark/runners/arithmetic.py
index 7898ae2bc..3ff064ae1 100755
--- a/yardstick/benchmark/runners/arithmetic.py
+++ b/yardstick/benchmark/runners/arithmetic.py
@@ -138,8 +138,18 @@ def _worker_process(queue, cls, method_name, scenario_cfg,
         if errors and sla_action is None:
             break
 
-    benchmark.teardown()
+    try:
+        benchmark.teardown()
+    except Exception:
+        # catch any exception in teardown and convert to simple exception
+        # never pass exceptions back to multiprocessing, because some exceptions can
+        # be unpicklable
+        # https://bugs.python.org/issue9400
+        LOG.exception("")
+        raise SystemExit(1)
     LOG.info("worker END")
+    LOG.debug("queue.qsize() = %s", queue.qsize())
+    LOG.debug("output_queue.qsize() = %s", output_queue.qsize())
 
 
 class ArithmeticRunner(base.Runner):
diff --git a/yardstick/benchmark/runners/base.py b/yardstick/benchmark/runners/base.py
index a69811f8a..3ecf67736 100755
--- a/yardstick/benchmark/runners/base.py
+++ b/yardstick/benchmark/runners/base.py
@@ -24,6 +24,9 @@ import subprocess
 import time
 import traceback
 
+
+from six.moves.queue import Empty
+
 import yardstick.common.utils as utils
 from yardstick.benchmark.scenarios import base as base_scenario
 
@@ -116,7 +119,7 @@ class Runner(object):
     @staticmethod
     def terminate_all():
         """Terminate all runners (subprocesses)"""
-        log.debug("Terminating all runners")
+        log.debug("Terminating all runners", exc_info=True)
 
         # release dumper process as some errors before any runner is created
         if not Runner.runners:
@@ -205,9 +208,21 @@ class Runner(object):
         """Abort the execution of a scenario"""
         self.aborted.set()
 
-    def join(self, timeout=None):
-        self.process.join(timeout)
+    QUEUE_JOIN_INTERVAL = 5
+
+    def join(self, outputs, result, interval=QUEUE_JOIN_INTERVAL):
+        while self.process.exitcode is None:
+            # drain the queue while we are running otherwise we won't terminate
+            outputs.update(self.get_output())
+            result.extend(self.get_result())
+            self.process.join(interval)
+        # drain after the process has exited
+        outputs.update(self.get_output())
+        result.extend(self.get_result())
+
+        self.process.terminate()
         if self.periodic_action_process:
+            self.periodic_action_process.join(1)
             self.periodic_action_process.terminate()
             self.periodic_action_process = None
 
@@ -217,11 +232,19 @@ class Runner(object):
     def get_output(self):
         result = {}
         while not self.output_queue.empty():
-            result.update(self.output_queue.get())
+            log.debug("output_queue size %s", self.output_queue.qsize())
+            try:
+                result.update(self.output_queue.get(True, 1))
+            except Empty:
+                pass
         return result
 
     def get_result(self):
         result = []
         while not self.result_queue.empty():
-            result.append(self.result_queue.get())
+            log.debug("result_queue size %s", self.result_queue.qsize())
+            try:
+                result.append(self.result_queue.get(True, 1))
+            except Empty:
+                pass
         return result
diff --git a/yardstick/benchmark/runners/duration.py b/yardstick/benchmark/runners/duration.py
index c2c6a8f19..75942766d 100644
--- a/yardstick/benchmark/runners/duration.py
+++ b/yardstick/benchmark/runners/duration.py
@@ -54,6 +54,7 @@ def _worker_process(queue, cls, method_name, scenario_cfg,
         sla_action = scenario_cfg["sla"].get("action", "assert")
 
     start = time.time()
+    timeout = start + duration
     while True:
 
         LOG.debug("runner=%(runner)s seq=%(sequence)s START",
@@ -71,9 +72,11 @@ def _worker_process(queue, cls, method_name, scenario_cfg,
             elif sla_action == "monitor":
                 LOG.warning("SLA validation failed: %s", assertion.args)
                 errors = assertion.args
-        except Exception as e:
+        # catch all exceptions because with multiprocessing we can have un-picklable exception
+        # problems  https://bugs.python.org/issue9400
+        except Exception:
             errors = traceback.format_exc()
-            LOG.exception(e)
+            LOG.exception("")
         else:
             if result:
                 output_queue.put(result)
@@ -94,12 +97,22 @@ def _worker_process(queue, cls, method_name, scenario_cfg,
 
         sequence += 1
 
-        if (errors and sla_action is None) or \
-                (time.time() - start > duration or aborted.is_set()):
+        if (errors and sla_action is None) or time.time() > timeout or aborted.is_set():
             LOG.info("Worker END")
             break
 
-    benchmark.teardown()
+    try:
+        benchmark.teardown()
+    except Exception:
+        # catch any exception in teardown and convert to simple exception
+        # never pass exceptions back to multiprocessing, because some exceptions can
+        # be unpicklable
+        # https://bugs.python.org/issue9400
+        LOG.exception("")
+        raise SystemExit(1)
+
+    LOG.debug("queue.qsize() = %s", queue.qsize())
+    LOG.debug("output_queue.qsize() = %s", output_queue.qsize())
 
 
 class DurationRunner(base.Runner):
diff --git a/yardstick/benchmark/runners/dynamictp.py b/yardstick/benchmark/runners/dynamictp.py
index afff27d75..01e76c6f4 100755
--- a/yardstick/benchmark/runners/dynamictp.py
+++ b/yardstick/benchmark/runners/dynamictp.py
@@ -141,7 +141,17 @@ def _worker_process(queue, cls, method_name, scenario_cfg,
             LOG.debug("iterator: %s iterations: %s", iterator, iterations)
 
     if "teardown" in run_step:
-        benchmark.teardown()
+        try:
+            benchmark.teardown()
+        except Exception:
+            # catch any exception in teardown and convert to simple exception
+            # never pass exceptions back to multiprocessing, because some exceptions can
+            # be unpicklable
+            # https://bugs.python.org/issue9400
+            LOG.exception("")
+            raise SystemExit(1)
+
+    LOG.debug("queue.qsize() = %s", queue.qsize())
 
 
 class IterationRunner(base.Runner):
diff --git a/yardstick/benchmark/runners/iteration.py b/yardstick/benchmark/runners/iteration.py
index 50fe106bd..4a7439588 100644
--- a/yardstick/benchmark/runners/iteration.py
+++ b/yardstick/benchmark/runners/iteration.py
@@ -90,7 +90,8 @@ def _worker_process(queue, cls, method_name, scenario_cfg,
                 LOG.exception(e)
             else:
                 if result:
-                    output_queue.put(result)
+                    LOG.debug("output_queue.put %s", result)
+                    output_queue.put(result, True, 1)
 
             time.sleep(interval)
 
@@ -101,7 +102,8 @@ def _worker_process(queue, cls, method_name, scenario_cfg,
                 'errors': errors
             }
 
-            queue.put(benchmark_output)
+            LOG.debug("queue.put, %s", benchmark_output)
+            queue.put(benchmark_output, True, 1)
 
             LOG.debug("runner=%(runner)s seq=%(sequence)s END",
                       {"runner": runner_cfg["runner_id"],
@@ -114,7 +116,18 @@ def _worker_process(queue, cls, method_name, scenario_cfg,
                 LOG.info("worker END")
                 break
     if "teardown" in run_step:
-        benchmark.teardown()
+        try:
+            benchmark.teardown()
+        except Exception:
+            # catch any exception in teardown and convert to simple exception
+            # never pass exceptions back to multiprocessing, because some exceptions can
+            # be unpicklable
+            # https://bugs.python.org/issue9400
+            LOG.exception("")
+            raise SystemExit(1)
+
+    LOG.debug("queue.qsize() = %s", queue.qsize())
+    LOG.debug("output_queue.qsize() = %s", output_queue.qsize())
 
 
 class IterationRunner(base.Runner):
diff --git a/yardstick/benchmark/runners/sequence.py b/yardstick/benchmark/runners/sequence.py
index 68e272c57..f08ca5dde 100644
--- a/yardstick/benchmark/runners/sequence.py
+++ b/yardstick/benchmark/runners/sequence.py
@@ -105,8 +105,18 @@ def _worker_process(queue, cls, method_name, scenario_cfg,
         if (errors and sla_action is None) or aborted.is_set():
             break
 
-    benchmark.teardown()
+    try:
+        benchmark.teardown()
+    except Exception:
+        # catch any exception in teardown and convert to simple exception
+        # never pass exceptions back to multiprocessing, because some exceptions can
+        # be unpicklable
+        # https://bugs.python.org/issue9400
+        LOG.exception("")
+        raise SystemExit(1)
     LOG.info("worker END")
+    LOG.debug("queue.qsize() = %s", queue.qsize())
+    LOG.debug("output_queue.qsize() = %s", output_queue.qsize())
 
 
 class SequenceRunner(base.Runner):