Changeset 244571 in webkit

Timestamp:

Apr 23, 2019 4:52:43 PM (5 years ago)

Author:

Jonathan Bedard

Message:

webkitpy: Kill unresponsive worker processes
​https://bugs.webkit.org/show_bug.cgi?id=197205
<rdar://problem/50104388>

Reviewed by Lucas Forschler.

• Scripts/webkitpy/common/message_pool.py:

(_MessagePool.init): Add close timeout.
(_MessagePool._close): Send termination before joining workers, kill unresponsive workers.

Location:

trunk/Tools

Files:

• ChangeLog (modified) (1 diff)
• Scripts/webkitpy/common/message_pool.py (modified) (4 diffs)

trunk/Tools/ChangeLog

@@ +1 @@
+2019-04-23  Jonathan Bedard  <jbedard@apple.com>
+
+        webkitpy: Kill unresponsive worker processes
+        https://bugs.webkit.org/show_bug.cgi?id=197205
+        <rdar://problem/50104388>
+
+        Reviewed by Lucas Forschler.
+
+        * Scripts/webkitpy/common/message_pool.py:
+        (_MessagePool.__init__): Add close timeout.
+        (_MessagePool._close): Send termination before joining workers, kill unresponsive workers.
+
 2019-04-23  Alex Christensen  <achristensen@webkit.org>
 

trunk/Tools/Scripts/webkitpy/common/message_pool.py

@@ -44 +44 @@
 import logging
 import multiprocessing
+import os
 import Queue
+import signal
 import sys
 import time
@@ -63 +65 @@
 
 class _MessagePool(object):
-    def __init__(self, caller, worker_factory, num_workers, worker_startup_delay_secs=0.0, host=None):
+    def __init__(self, caller, worker_factory, num_workers, worker_startup_delay_secs=0.0, host=None, timeout=30):
         self._caller = caller
         self._worker_factory = worker_factory
@@ -73 +75 @@
         self._name = 'manager'
         self._running_inline = (self._num_workers == 1)
+        self._timeout = timeout
         if self._running_inline:
             self._messages_to_worker = Queue.Queue()
@@ -138 +141 @@
             if worker.is_alive():
                 worker.terminate()
-                worker.join()
+
+        deadline = time.time() + self._timeout
+        for worker in self._workers:
+            if worker.is_alive():
+                worker.join(timeout=max(deadline - time.time(), 0))
+            if worker.is_alive():
+                _log.error('{} failed to terminate, killing it'.format(worker.name))
+                os.kill(worker.ident, signal.SIGKILL)
+
         self._workers = []
         if not self._running_inline: