From: Steve Baker <sbaker@redhat.com>
Date: Mon, 24 Jun 2013 02:19:36 +0000 (+1200)
Subject: Emit alarm actions if a rule remains in ALARM state.
X-Git-Tag: 2014.1~445^2
X-Git-Url: https://review.fuel-infra.org/gitweb?a=commitdiff_plain;h=bdeb33afdc52f37c2270a72c2221b051024f3f3f;p=openstack-build%2Fheat-build.git

Emit alarm actions if a rule remains in ALARM state.

It looks like the original intent of this edge-triggering was to
prevent actions being emitted too frequently, causing overshoot of
scaling resources.

However this causes an issue when one scaling resource change is not
enough to bring the alarm state under the threshold.

Proper use of Period, EvaluationPeriods and Threshold should be sufficient
to prevent scaling resource overshoot.

Fixes bug: #1193970

Change-Id: I5b38eb82cd1ad568fe4af77999e9d52f9eb99328
---

diff --git a/heat/engine/watchrule.py b/heat/engine/watchrule.py
index e6db474b..0a70ac93 100644
--- a/heat/engine/watchrule.py
+++ b/heat/engine/watchrule.py
@@ -216,11 +216,8 @@ class WatchRule(object):
 
     def run_rule(self):
         new_state = self.get_alarm_state()
-
-        actions = []
-        if new_state != self.state:
-            actions = self.rule_actions(new_state)
-            self.state = new_state
+        actions = self.rule_actions(new_state)
+        self.state = new_state
 
         self.last_evaluated = self.now
         self.store()
diff --git a/heat/tests/test_watch.py b/heat/tests/test_watch.py
index 5066cd08..1c342dd7 100644
--- a/heat/tests/test_watch.py
+++ b/heat/tests/test_watch.py
@@ -439,14 +439,12 @@ class WatchRuleTest(HeatTestCase):
         self.assertEqual(watcher.state, 'ALARM')
         self.assertEqual(actions, ['DummyAction'])
 
-        # re-set last_evaluated so the rule will be evaluated again,
-        # but since we're already in ALARM state, we should not generate
-        # any additional actions
+        # re-set last_evaluated so the rule will be evaluated again.
         last = now - datetime.timedelta(seconds=300)
         watcher.last_evaluated = last
         actions = watcher.evaluate()
         self.assertEqual(watcher.state, 'ALARM')
-        self.assertEqual(actions, [])
+        self.assertEqual(actions, ['DummyAction'])
         self.m.VerifyAll()
 
     def test_rule_actions_alarm_two_actions(self):