X-Git-Url: https://git.arvados.org/arvados.git/blobdiff_plain/2ec2c8ed2c5db174f3a83dc257fa4c4b3190f47b..de985c815dd4a81dcb5b03475452e2c93d3dadb4:/services/nodemanager/arvnodeman/computenode/dispatch/__init__.py

diff --git a/services/nodemanager/arvnodeman/computenode/dispatch/__init__.py b/services/nodemanager/arvnodeman/computenode/dispatch/__init__.py
index a239f1f252..42960efeba 100644
--- a/services/nodemanager/arvnodeman/computenode/dispatch/__init__.py
+++ b/services/nodemanager/arvnodeman/computenode/dispatch/__init__.py
@@ -46,7 +46,7 @@ class ComputeNodeStateChangeBase(config.actor_class, RetryMixin):
     def subscribe(self, subscriber):
         if self.subscribers is None:
             try:
-                subscriber(self._later)
+                subscriber(self.actor_ref.proxy())
             except pykka.ActorDeadError:
                 pass
         else:
@@ -256,12 +256,6 @@ class ComputeNodeUpdateActor(config.actor_class):
     this to perform maintenance tasks on themselves.  Having a
     dedicated actor for this gives us the opportunity to control the
     flow of requests; e.g., by backing off when errors occur.
-
-    This actor is most like a "traditional" Pykka actor: there's no
-    subscribing, but instead methods return real driver results.  If
-    you're interested in those results, you should get them from the
-    Future that the proxy method returns.  Be prepared to handle exceptions
-    from the cloud driver when you do.
     """
     def __init__(self, cloud_factory, max_retry_wait=180):
         super(ComputeNodeUpdateActor, self).__init__()
@@ -270,6 +264,12 @@ class ComputeNodeUpdateActor(config.actor_class):
         self.error_streak = 0
         self.next_request_time = time.time()
 
+    def _set_logger(self):
+        self._logger = logging.getLogger("%s.%s" % (self.__class__.__name__, self.actor_urn[33:]))
+
+    def on_start(self):
+        self._set_logger()
+
     def _throttle_errors(orig_func):
         @functools.wraps(orig_func)
         def throttle_wrapper(self, *args, **kwargs):
@@ -283,7 +283,9 @@ class ComputeNodeUpdateActor(config.actor_class):
                 self.error_streak += 1
                 self.next_request_time += min(2 ** self.error_streak,
                                               self.max_retry_wait)
-                raise
+                self._logger.error(
+                    "Caught unknown error (no retry): %s",
+                    error, exc_info=error)
             else:
                 self.error_streak = 0
                 return result
@@ -406,7 +408,7 @@ class ComputeNodeMonitorActor(config.actor_class):
         first_ping_s = arvados_node.get('first_ping_at')
         if (self.arvados_node is not None) or (not first_ping_s):
             return None
-        elif ((arvados_node['ip_address'] in self.cloud_node.private_ips) and
+        elif ((arvados_node['info'].get('ec2_instance_id') == self._cloud.node_id(self.cloud_node)) and
               (arvados_timestamp(first_ping_s) >= self.cloud_node_start_time)):
             self._later.update_arvados_node(arvados_node)
             return self.cloud_node.id