From: Miguel Angel Ajo <mangelajo@redhat.com>
Date: Wed, 10 Dec 2014 11:21:52 +0000 (+0100)
Subject: Fix race condition in ProcessMonitor
X-Git-Url: https://review.fuel-infra.org/gitweb?a=commitdiff_plain;h=b25915f1d94e19756c932ae18119bdf872cc6316;p=openstack-build%2Fneutron-build.git

Fix race condition in ProcessMonitor

If any new external process was enabled/disabled during
the process monitoring loop time, a RuntimeError: dictionary
changed size during iteration was thrown. This is fixed
by pre-building the service list from the dictionary
for the iteration.

Closes-Bug: #1401042

Change-Id: Id0a3af22fb85992bde35d8c691bee3ddf435639d
---

diff --git a/neutron/agent/linux/external_process.py b/neutron/agent/linux/external_process.py
index b6b3688e8..35053d41a 100644
--- a/neutron/agent/linux/external_process.py
+++ b/neutron/agent/linux/external_process.py
@@ -220,7 +220,10 @@ class ProcessMonitor(object):
 
     @lockutils.synchronized("_check_child_processes")
     def _check_child_processes(self):
-        for service_id in self._process_managers:
+        # we build the list of keys before iterating in the loop to cover
+        # the case where other threads add or remove items from the
+        # dictionary which otherwise will cause a RuntimeError
+        for service_id in list(self._process_managers):
             pm = self._process_managers.get(service_id)
 
             if pm and not pm.active: