Make kube_service tag collection configurable:

- add collect_service_tags option to disable collection
- add service_tag_update_freq to reduce event requests

kubernetes/check.py

@@ -32,6 +32,8 @@
 DEFAULT_COLLECT_EVENTS = False
 DEFAULT_NAMESPACES = ['default']
 
+DEFAULT_SERVICE_EVENT_FREQ = 5 * 60  # seconds
+
 NET_ERRORS = ['rx_errors', 'tx_errors', 'rx_dropped', 'tx_dropped']
 
 DEFAULT_ENABLED_GAUGES = [
@@ -106,14 +108,17 @@ def __init__(self, name, init_config, agentConfig, instances=None):
             self._collect_events = _is_affirmative(inst.get('collect_events', DEFAULT_COLLECT_EVENTS))
             if self._collect_events:
                 self.event_retriever = self.kubeutil.get_event_retriever()
+            elif self.kubeutil.collect_service_tag:
+                # Only fetch service and pod events for service mapping
+                event_delay = inst.get('service_tag_update_freq', DEFAULT_SERVICE_EVENT_FREQ)
+                self.event_retriever = self.kubeutil.get_event_retriever(kinds=['Service', 'Pod'],
+                                                                         delay=event_delay)
             else:
-                # Only fetch service and pod events for service mapping and SD
-                self.event_retriever = self.kubeutil.get_event_retriever(kinds=['Service', 'Pod'])
+                self.event_retriever = None
         else:
             self._collect_events = None
             self.event_retriever = None
 
-
     def _perform_kubelet_checks(self, url):
         service_check_base = NAMESPACE + '.kubelet.check'
         is_ok = True

kubernetes/conf.yaml.example

@@ -51,6 +51,13 @@ instances:
   #
   # collect_events: false
   #
+  # Matching the pods to Kubernetes services requires to retrieve events regularly.
+  # To reduce the traffic to the apiserver, we only query them every 5 minutes, adding a delay
+  # in pod -> service matching. You can configure it below (in seconds) or disable kube_service
+  # tagging completely
+  #
+  # collect_service_tags: true
+  # service_tag_update_freq: 300
   #
   # The namespaces for which events should be collected.
   # If not modified, the 'default' namespace will be used.