Keep current and previous snapshot images

The previous logic around how to keep images was not accomplishing
anything particularly useful.

Instead:
  * Delete images that are not configured or have no corresponding
    base image.
  * Keep the current and previous READY images.
  * Otherwise, delete any images that have been in their current
    state for more than 8 hours.

Also, correct the image-update command which no longer needs to
join a thread.  Also, fix up some poorly exercised parts of the
fake provider.

Change-Id: Iba921f26d971e56692b9104f9d7c531d955d17b4

nodepool/cmd/nodepoolcmd.py

@@ -129,8 +129,7 @@ class NodePoolCmd(object):
         image = provider.images[self.args.image]
 
         with self.pool.getDB().getSession() as session:
-            t = self.pool.updateImage(session, provider, image)
+            self.pool.updateImage(session, provider, image)
-            t.join()
 
     def alien_list(self):
         self.pool.reconfigureManagers(self.pool.config)

nodepool/fakeprovider.py

@@ -54,11 +54,15 @@ class FakeList(object):
         time.sleep(delay)
         obj.status = status
 
-    def delete(self, obj):
+    def delete(self, *args, **kw):
-        if hasattr(obj, 'id'):
+        if 'image' in kw:
-            self._list.remove(obj)
+            self._list.remove(self.get(kw['image']))
         else:
-            self._list.remove(self.get(obj))
+            obj = args[0]
+            if hasattr(obj, 'id'):
+                self._list.remove(obj)
+            else:
+                self._list.remove(self.get(obj))
 
     def create(self, **kw):
         s = Dummy(id=uuid.uuid4().hex,

nodepool/nodedb.py

@@ -250,12 +250,16 @@ class NodeDatabaseSession(object):
             return None
         return images[0]
 
-    def getCurrentSnapshotImage(self, provider_name, image_name):
+    def getOrderedReadySnapshotImages(self, provider_name, image_name):
         images = self.session().query(SnapshotImage).filter(
             snapshot_image_table.c.provider_name == provider_name,
             snapshot_image_table.c.image_name == image_name,
             snapshot_image_table.c.state == READY).order_by(
             snapshot_image_table.c.version.desc()).all()
+        return images
+
+    def getCurrentSnapshotImage(self, provider_name, image_name):
+        images = self.getOrderedReadySnapshotImages(provider_name, image_name)
         if not images:
             return None
         return images[0]

nodepool/nodepool.py

@@ -44,7 +44,8 @@ CONNECT_TIMEOUT = 10 * MINS  # How long to try to connect after a server
 NODE_CLEANUP = 8 * HOURS     # When to start deleting a node that is not
                              # READY or HOLD
 TEST_CLEANUP = 5 * MINS      # When to start deleting a node that is in TEST
-KEEP_OLD_IMAGE = 24 * HOURS  # How long to keep an old (good) image
+IMAGE_CLEANUP = 8 * HOURS    # When to start deleting an image that is not
+                             # READY or is not the current or previous image
 DELETE_DELAY = 1 * MINS      # Delay before deleting a node that has completed
                              # its job.
 
@@ -1283,7 +1284,8 @@ class NodePool(threading.Thread):
 
     def cleanupOneImage(self, session, image):
         # Normally, reap images that have sat in their current state
-        # for 24 hours, unless the image is the current snapshot
+        # for 8 hours, unless the image is the current or previous
+        # snapshot.
         delete = False
         now = time.time()
         if image.provider_name not in self.config.providers:
@@ -1296,12 +1298,17 @@ class NodePool(threading.Thread):
             self.log.info("Deleting image id: %s which has no current "
                           "base image" % image.id)
         else:
-            current = session.getCurrentSnapshotImage(image.provider_name,
+            images = session.getOrderedReadySnapshotImages(
-                                                      image.image_name)
+                image.provider_name, image.image_name)
-            if (current and image != current and
+            current = previous = None
-                (now - image.state_time) > KEEP_OLD_IMAGE):
+            if len(images) > 0:
-                self.log.info("Deleting non-current image id: %s because "
+                current = images[0]
-                              "the image is %s hours old" %
+            if len(images) > 1:
+                previous = images[1]
+            if (image != current and image != previous and
+                (now - image.state_time) > IMAGE_CLEANUP):
+                self.log.info("Deleting image id: %s which is "
+                              "%s hours old" %
                               (image.id,
                                (now - image.state_time) / (60 * 60)))
                 delete = True

tools/fake-servers.py

@@ -68,7 +68,8 @@ def main():
     zsocket = context.socket(zmq.PUB)
     zsocket.bind("tcp://*:8881")
 
-    geard = MyGearmanServer()
+    geard = MyGearmanServer(statsd_host='localhost', statsd_port=8125,
+                            statsd_prefix='zuul.geard')
     geard._count = 0
 
     statsd = FakeStatsd()