X-Git-Url: https://git.arvados.org/arvados.git/blobdiff_plain/fe3f19bb8f0a684de9c5a7d1f7c619943d0daca0..a20c1480b7ed827d02511d1630e0894696814107:/services/nodemanager/arvnodeman/baseactor.py

diff --git a/services/nodemanager/arvnodeman/baseactor.py b/services/nodemanager/arvnodeman/baseactor.py
index 9591b42ca3..bdfe5d45a7 100644
--- a/services/nodemanager/arvnodeman/baseactor.py
+++ b/services/nodemanager/arvnodeman/baseactor.py
@@ -1,13 +1,21 @@
+# Copyright (C) The Arvados Authors. All rights reserved.
+#
+# SPDX-License-Identifier: AGPL-3.0
+
 from __future__ import absolute_import, print_function
 
 import errno
 import logging
 import os
+import signal
+import time
 import threading
 import traceback
 
 import pykka
 
+from .status import tracker
+
 class _TellCallableProxy(object):
     """Internal helper class for proxying callables."""
 
@@ -76,10 +84,46 @@ class BaseNodeManagerActor(pykka.ThreadingActor):
     def __init__(self, *args, **kwargs):
          super(pykka.ThreadingActor, self).__init__(*args, **kwargs)
          self.actor_ref = TellableActorRef(self)
+         self._killfunc = kwargs.get("killfunc", os.kill)
 
     def on_failure(self, exception_type, exception_value, tb):
         lg = getattr(self, "_logger", logging)
         if (exception_type in (threading.ThreadError, MemoryError) or
             exception_type is OSError and exception_value.errno == errno.ENOMEM):
             lg.critical("Unhandled exception is a fatal error, killing Node Manager")
-            os.killpg(os.getpgid(0), 9)
+            self._killfunc(os.getpid(), signal.SIGKILL)
+        tracker.counter_add('actor_exceptions')
+
+    def ping(self):
+        return True
+
+    def get_thread(self):
+        return threading.current_thread()
+
+class WatchdogActor(pykka.ThreadingActor):
+    def __init__(self, timeout, *args, **kwargs):
+         super(pykka.ThreadingActor, self).__init__(*args, **kwargs)
+         self.timeout = timeout
+         self.actors = [a.proxy() for a in args]
+         self.actor_ref = TellableActorRef(self)
+         self._later = self.actor_ref.tell_proxy()
+         self._killfunc = kwargs.get("killfunc", os.kill)
+
+    def kill_self(self, e, act):
+        lg = getattr(self, "_logger", logging)
+        lg.critical("Watchdog exception", exc_info=e)
+        lg.critical("Actor %s watchdog ping time out, killing Node Manager", act)
+        self._killfunc(os.getpid(), signal.SIGKILL)
+
+    def on_start(self):
+        self._later.run()
+
+    def run(self):
+        a = None
+        try:
+            for a in self.actors:
+                a.ping().get(self.timeout)
+            time.sleep(20)
+            self._later.run()
+        except Exception as e:
+            self.kill_self(e, a)