rework AD image name extractor to support new swarm format

swarm can use org/image:tag@sha256:xxxxxxxx format, that throws off
the parser. reworking it to handle all known formats and unit test it

Author: xvello

tests/core/test_service_discovery.py

@@ -174,6 +174,26 @@ class TestServiceDiscovery(unittest.TestCase):
         ])),
     }
 
+    image_formats = {
+        # Don't crash on empty string or None
+        '': '',
+        None: '',
+        # Shortest possibility
+        'alpine': 'alpine',
+        # Historical docker format
+        'nginx:latest': 'nginx',
+        # Org prefix to be removed
+        'datadog/docker-dd-agent:latest-jmx': 'docker-dd-agent',
+        # Sha-pinning used by many orchestrators
+        'redis@sha256:5bef08742407efd622d243692b79ba0055383bbce12900324f75e56f589aedb0': 'redis',
+        # Quirky pinning used by swarm
+        'org/redis:latest@sha256:5bef08742407efd622d243692b79ba0055383bbce12900324f75e56f589aedb0': 'redis',
+        # Custom registry, simple form
+        'myregistry.local:5000/testing/test-image:version': 'test-image',
+        # Custom registry, most insane form possible
+        'myregistry.local:5000/testing/test-image:version@sha256:5bef08742407efd622d243692b79ba0055383bbce12900324f75e56f589aedb0': 'test-image',
+    }
+
     def setUp(self):
         self.etcd_agentConfig = {
             'service_discovery': True,
@@ -328,6 +348,16 @@ def test_get_config_templates(self, *args):
                 self.assertEquals(sd_backend._get_config_templates(image), None)
             clear_singletons(agentConfig)
 
+    @mock.patch('config.get_auto_confd_path', return_value=os.path.join(
+        os.path.dirname(__file__), 'fixtures/auto_conf/'))
+    @mock.patch('utils.dockerutil.DockerUtil.client', return_value=None)
+    #@mock.patch.object(AbstractConfigStore, 'get_check_tpls', side_effect=_get_check_tpls)
+    def test_get_image_ident(self, *args):
+        sd_backend = get_sd_backend(agentConfig=self.auto_conf_agentConfig)
+        # normal cases
+        for image, ident in self.image_formats.iteritems():
+            self.assertEquals(ident, sd_backend.config_store._get_image_ident(image))
+
     @mock.patch('config.get_auto_confd_path', return_value=os.path.join(
         os.path.dirname(__file__), 'fixtures/auto_conf/'))
     def test_render_template(self, mock_get_auto_confd_path):

utils/service_discovery/abstract_config_store.py

@@ -366,20 +366,25 @@ def read_config_from_store(self, identifier):
 
         return res
 
-    def _get_image_ident(self, ident):
+    def _get_image_ident(self, image_name):
         """Extract an identifier from the image"""
+        # See image_formats in test_service_discovery.py for supported formats
+
         # handle exceptionnal empty ident case (docker bug)
-        if not ident:
+        if not image_name:
             return ""
-        # handle the 'redis@sha256:...' format
-        if '@' in ident:
-            return ident.split('@')[0].split('/')[-1]
-        # if a custom image store is used there can be a port which adds a colon
-        elif ident.count(':') > 1:
-            return ident.split(':')[1].split('/')[-1]
-        # otherwise we just strip the tag and keep the image name
-        else:
-            return ident.split(':')[0].split('/')[-1]
+        ident = image_name
+        # remove the @sha256: suffix if present
+        if '@sha' in ident:
+            ident = ident.split('@sha')[0]
+        # remove image org / store prefix, we keep the last part after '/''
+        if '/' in ident:
+            ident = ident.split('/')[-1]
+        # remove the image tag after :
+        if ':' in ident:
+            ident = ident.split(':')[0]
+
+        return ident
 
     def crawl_config_template(self):
         """Return whether or not configuration templates have changed since the previous crawl"""