[yarn] whitelist authorized application_tags

Risk of context spam.
DataDog · Dec 12, 2016 · 0358b58 · 0358b58
1 parent 325f1e6
commit 0358b58
Show file tree

Hide file tree

Showing 3 changed files with 25 additions and 11 deletions.
diff --git a/checks.d/yarn.py b/checks.d/yarn.py
@@ -143,6 +143,13 @@ class YarnCheck(AgentCheck):
     '''
     Extract statistics from YARN's ResourceManger REST API
     '''
+    _ALLOWED_APPLICATION_TAGS = [
+        'applicationTags',
+        'applicationType',
+        'name',
+        'queue',
+        'user'
+    ]
 
     def check(self, instance):
 
@@ -154,9 +161,16 @@ def check(self, instance):
             self.log.error('application_tags is incorrect: %s is not a dictionary', app_tags)
             app_tags = {}
 
+        filtered_app_tags = {}
+        for dd_prefix, yarn_key in app_tags.iteritems():
+            if yarn_key in self._ALLOWED_APPLICATION_TAGS:
+                filtered_app_tags[dd_prefix] = yarn_key
+        app_tags = filtered_app_tags
+
         # Collected by default
         app_tags['app_name'] = 'name'
 
+
         # Get additional tags from the conf file
         tags = instance.get('tags', [])
         if tags is None:
@@ -201,7 +215,7 @@ def _yarn_app_metrics(self, rm_address, app_tags, addl_tags):
         )
 
         if (metrics_json and metrics_json['apps'] is not None and
-            metrics_json['apps']['app'] is not None):
+                metrics_json['apps']['app'] is not None):
 
             for app_json in metrics_json['apps']['app']:
 
@@ -225,7 +239,7 @@ def _yarn_node_metrics(self, rm_address, addl_tags):
         metrics_json = self._rest_request_to_json(rm_address, YARN_NODES_PATH)
 
         if (metrics_json and metrics_json['nodes'] is not None and
-            metrics_json['nodes']['node'] is not None):
+                metrics_json['nodes']['node'] is not None):
 
             for node_json in metrics_json['nodes']['node']:
                 node_id = node_json['id']

diff --git a/conf.d/yarn.yaml.example b/conf.d/yarn.yaml.example
@@ -19,17 +19,16 @@ instances:
 
     # Optional tags to be applied to every emitted metric.
     # tags:
-    #   - key:value
-    #   - instance:production
+    #   - "key:value"
+    #   - "instance:production"
 
     # Optional tags retrieved from the application data to be applied to the
     # application metrics.
     # application_tags:
-    #   # tag_prefix: yarn_key
-    #   - app_queue: queue
+    # # tag_prefix: yarn_key
+    #   app_queue: queue
     # This will add a tag 'app_queue:name_of_the_queue' to the app metrics,
-    # app_queue being the tag_prefix and queue the actual YARN key. Whole list
-    # of keys on yarn api reference:
-    # https://hadoop.apache.org/docs/current/hadoop-yarn/hadoop-yarn-site/ResourceManagerRest.html#Cluster_Applications_API
+    # app_queue being the tag_prefix and queue the actual YARN key.
+    # Allowed yarn keys: applicationType, applicationTags, name, queue, user
     # By default, the application name is collected with the prefix app_name.
 
diff --git a/tests/checks/mock/test_yarn.py b/tests/checks/mock/test_yarn.py
@@ -58,7 +58,8 @@ class YARNCheck(AgentCheckTest):
             'opt_key:opt_value'
         ],
         'application_tags': {
-            'app_id': 'id'
+            'app_id': 'id',
+            'app_queue': 'queue'
         }
     }
 
@@ -108,7 +109,7 @@ class YARNCheck(AgentCheckTest):
     YARN_APP_METRICS_TAGS = [
         'cluster_name:%s' % CLUSTER_NAME,
         'app_name:word count',
-        'app_id:application_1326815542473_0001',
+        'app_queue:default',
         'opt_key:opt_value'
     ]