Tristan/agent checks ui

checks.d/couch.py

@@ -7,14 +7,16 @@ class CouchDb(AgentCheck):
     """Extracts stats from CouchDB via its REST API
     http://wiki.apache.org/couchdb/Runtime_Statistics
     """
+    SOURCE_TYPE_NAME = 'couchdb'
+
     def _create_metric(self, data, tags=None):
         overall_stats = data.get('stats', {})
         for key, stats in overall_stats.items():
             for metric, val in stats.items():
                 if val['current'] is not None:
                     metric_name = '.'.join(['couchdb', key, metric])
                     self.gauge(metric_name, val['current'], tags=tags)
-        
+
         for db_name, db_stats in data.get('databases', {}).items():
             for name, val in db_stats.items():
                 if name in ['doc_count', 'disk_size'] and val is not None:
@@ -23,7 +25,7 @@ def _create_metric(self, data, tags=None):
                     metric_tags.append('db:%s' % db_name)
                     self.gauge(metric_name, val, tags=metric_tags, device_name=db_name)
 
-        
+
     def _get_stats(self, url):
         "Hit a given URL and return the parsed json"
         self.log.debug('Fetching Couchdb stats at url: %s' % url)
@@ -79,9 +81,9 @@ def parse_agent_config(agentConfig):
         if not agentConfig.get('couchdb_server'):
             return False
 
-        
+
         return {
             'instances': [{
                 'server': agentConfig.get('couchdb_server'),
             }]
-        }
+        }

checks.d/elastic.py

@@ -110,6 +110,8 @@ class ElasticSearch(AgentCheck):
         "elasticsearch.cluster_status": ("gauge", "status", lambda v: {"red":0,"yellow":1,"green":2}.get(v, -1)),
     }
 
+    SOURCE_TYPE_NAME = 'elasticsearch'
+
     def __init__(self, name, init_config, agentConfig):
         AgentCheck.__init__(self, name, init_config, agentConfig)
 

checks.d/gearmand.py

@@ -2,6 +2,8 @@
 
 class Gearman(AgentCheck):
 
+    SOURCE_TYPE_NAME = 'germand'
+
     def get_library_versions(self):
         try:
             import gearman
@@ -28,7 +30,7 @@ def _get_metrics(self, client, tags):
         running = 0
         queued = 0
         workers = 0
-        
+
         for stat in data:
             running += stat['running']
             queued += stat['queued']

checks.d/kafka_consumer.py

@@ -21,6 +21,9 @@
 import random
 
 class KafkaCheck(AgentCheck):
+
+    SOURCE_TYPE_NAME = 'kafka'
+
     def check(self, instance):
         consumer_groups = self.read_config(instance, 'consumer_groups',
                                            cast=self._validate_consumer_groups)

checks.d/kyototycoon.py

@@ -14,6 +14,7 @@ class KyotoTycoonCheck(AgentCheck):
     """Report statistics about the Kyoto Tycoon DBM-style
     database server (http://fallabs.com/kyototycoon/)
     """
+    SOURCE_TYPE_NAME = 'kyoto tycoon'
 
     GAUGES = {
         'repl_delay':         'replication.delay',
@@ -68,7 +69,7 @@ def check(self, instance):
             if key in self.GAUGES:
                 name = self.GAUGES[key]
                 self.gauge('kyototycoon.%s' % name, float(value), tags=tags)
-            
+
             elif key in self.RATES:
                 name = self.RATES[key]
                 self.rate('kyototycoon.%s_per_s' % name, float(value), tags=tags)

checks.d/mcache.py

@@ -9,34 +9,34 @@
 # version           string   Version string of this server
 # pointer_size      32       Default size of pointers on the host OS
 #                            (generally 32 or 64)
-# rusage_user       32u:32u  Accumulated user time for this process 
+# rusage_user       32u:32u  Accumulated user time for this process
 #                            (seconds:microseconds)
-# rusage_system     32u:32u  Accumulated system time for this process 
+# rusage_system     32u:32u  Accumulated system time for this process
 #                            (seconds:microseconds)
 # curr_items        32u      Current number of items stored by the server
-# total_items       32u      Total number of items stored by this server 
+# total_items       32u      Total number of items stored by this server
 #                            ever since it started
-# bytes             64u      Current number of bytes used by this server 
+# bytes             64u      Current number of bytes used by this server
 #                            to store items
 # curr_connections  32u      Number of open connections
-# total_connections 32u      Total number of connections opened since 
+# total_connections 32u      Total number of connections opened since
 #                            the server started running
-# connection_structures 32u  Number of connection structures allocated 
+# connection_structures 32u  Number of connection structures allocated
 #                            by the server
 # cmd_get           64u      Cumulative number of retrieval requests
 # cmd_set           64u      Cumulative number of storage requests
-# get_hits          64u      Number of keys that have been requested and 
+# get_hits          64u      Number of keys that have been requested and
 #                            found present
-# get_misses        64u      Number of items that have been requested 
+# get_misses        64u      Number of items that have been requested
 #                            and not found
 # evictions         64u      Number of valid items removed from cache
 #                            to free memory for new items
-# bytes_read        64u      Total number of bytes read by this server 
+# bytes_read        64u      Total number of bytes read by this server
 #                            from network
-# bytes_written     64u      Total number of bytes sent by this server to 
+# bytes_written     64u      Total number of bytes sent by this server to
 #                            network
 # limit_maxbytes    32u      Number of bytes this server is allowed to
-#                            use for storage. 
+#                            use for storage.
 # threads           32u      Number of worker threads requested.
 #                            (see doc/threads.txt)
 #     >>> mc.get_stats()
@@ -55,6 +55,9 @@
 # https://github.com/membase/ep-engine/blob/master/docs/stats.org
 
 class Memcache(AgentCheck):
+
+    SOURCE_TYPE_NAME = 'memcached'
+
     DEFAULT_PORT = 11211
 
     GAUGES = [

checks.d/mongo.py

@@ -14,6 +14,8 @@
 
 class MongoDb(AgentCheck):
 
+    SOURCE_TYPE_NAME = 'mongodb'
+
     GAUGES = [
         "indexCounters.btree.missRatio",
         "globalLock.ratio",

checks.d/network.py

@@ -88,7 +88,7 @@ def _submit_devicemetrics(self, iface, vals_by_metric):
 
         # For reasons i don't understand only these metrics are skipped if a
         # particular interface is in the `excluded_interfaces` config list.
-        # Not sure why the others aren't included. Until I understand why, I'm 
+        # Not sure why the others aren't included. Until I understand why, I'm
         # going to keep the same behaviour.
         exclude_iface_metrics = [
             'packets_in.count',
@@ -238,7 +238,7 @@ def _check_bsd(self, instance):
                     'bytes_rcvd': self._parse_value(x[-5]),
                     'bytes_sent': self._parse_value(x[-2]),
                     'packets_in.count': self._parse_value(x[-7]),
-                    'packets_in.error': self._parse_value(x[-6]), 
+                    'packets_in.error': self._parse_value(x[-6]),
                     'packets_out.count': self._parse_value(x[-4]),
                     'packets_out.error':self._parse_value(x[-3]),
                 }

checks.d/redisdb.py

@@ -8,6 +8,9 @@
 class Redis(AgentCheck):
     db_key_pattern = re.compile(r'^db\d+')
     subkeys = ['keys', 'expires']
+
+    SOURCE_TYPE_NAME = 'redis'
+
     GAUGE_KEYS = {
         # Append-only metrics
         'aof_last_rewrite_time_sec':    'redis.aof.last_rewrite_time',
@@ -106,7 +109,7 @@ def _get_conn(self, instance):
         key = self._generate_instance_key(instance)
         if key not in self.connections:
             try:
-                
+
                 # Only send useful parameters to the redis client constructor
                 list_params = ['host', 'port', 'db', 'password', 'socket_timeout',
                     'connection_pool', 'charset', 'errors', 'unix_socket_path']
@@ -139,12 +142,12 @@ def _check_db(self, instance, custom_tags=None):
         try:
             info = conn.info()
         except ValueError, e:
-            # This is likely a know issue with redis library 2.0.0 
+            # This is likely a know issue with redis library 2.0.0
             # See https://github.com/DataDog/dd-agent/issues/374 for details
             import redis
             raise Exception("""Unable to run the info command. This is probably an issue with your version of the python-redis library.
                 Minimum required version: 2.4.11
-                Your current version: %s 
+                Your current version: %s
                 Please upgrade to a newer version by running sudo easy_install redis""" % redis.__version__)
 
         latency_ms = round((time.time() - start) * 1000, 2)

checks.d/sqlserver.py

@@ -8,6 +8,9 @@
 VALID_METRIC_TYPES = ('gauge', 'rate', 'histogram')
 
 class SQLServer(AgentCheck):
+
+    SOURCE_TYPE_NAME = 'sql server'
+
     METRICS = [
         ('sqlserver.buffer.cache_hit_ratio', 'gauge', 'Buffer cache hit ratio'),
         ('sqlserver.buffer.page_life_expectancy', 'gauge', 'Page life expectancy'),

checks.d/wmi_check.py

@@ -29,7 +29,7 @@ def _get_wmi_conn(self, host, user, password):
     def check(self, instance):
         if wmi is None:
             raise Exception("Missing 'wmi' module")
-        
+
         host = instance.get('host', None)
         user = instance.get('username', None)
         password = instance.get('password', None)

checks/__init__.py

@@ -264,6 +264,8 @@ def get_metrics(self, expire=True):
 class AgentCheck(object):
     OK, WARNING, CRITICAL, UNKNOWN, NONE = (0, 1, 2, 3, 4)
 
+    SOURCE_TYPE_NAME = None
+
     def __init__(self, name, init_config, agentConfig, instances=None):
         """
         Initialize a new check.
@@ -275,7 +277,6 @@ def __init__(self, name, init_config, agentConfig, instances=None):
         """
         from aggregator import MetricsAggregator
 
-
         self.name = name
         self.init_config = init_config
         self.agentConfig = agentConfig

checks/check_status.py

@@ -267,8 +267,9 @@ class CheckStatus(object):
     def __init__(self, check_name, instance_statuses, metric_count=None,
                  event_count=None, service_check_count=None,
                  init_failed_error=None, init_failed_traceback=None,
-                 library_versions=None):
+                 library_versions=None, check_source_type_name=None):
         self.name = check_name
+        self.source_type_name = check_source_type_name
         self.instance_statuses = instance_statuses
         self.metric_count = metric_count or 0
         self.event_count = event_count or 0

checks/collector.py

@@ -257,7 +257,6 @@ def run(self, checksd=None, start_event=True):
             try:
                 # Run the check.
                 instance_statuses = check.run()
-
                 # Collect the metrics and events.
                 current_check_metrics = check.get_metrics()
                 current_check_events = check.get_events()
@@ -281,7 +280,8 @@ def run(self, checksd=None, start_event=True):
                 log.exception("Error running check %s" % check.name)
 
             check_status = CheckStatus(check.name, instance_statuses, metric_count, event_count, service_check_count,
-                library_versions=check.get_library_info())
+                library_versions=check.get_library_info(),
+                check_source_type_name=check.SOURCE_TYPE_NAME or check.name)
             check_statuses.append(check_status)
 
         for check_name, info in self.init_failed_checks_d.iteritems():
@@ -297,6 +297,21 @@ def run(self, checksd=None, start_event=True):
         payload['metrics'] = metrics
         payload['events'] = events
         payload['service_checks'] = service_checks
+
+        # Add agent_checks if needed
+        if self._should_send_metadata():
+            agent_checks = []
+            for check in check_statuses:
+                for instance_status in check.instance_statuses:
+                    agent_checks.append(
+                        (
+                            check.name, check.source_type_name,
+                            instance_status.instance_id,
+                            instance_status.status, instance_status.error
+                        )
+                    )
+            payload['agent_checks'] = agent_checks
+
         collect_duration = timer.step()
 
         if self.os != 'windows':
@@ -379,7 +394,7 @@ def _build_payload(self, start_event=True):
                                  }]
 
         # Periodically send the host metadata.
-        if self._is_first_run() or self._should_send_metadata():
+        if self._should_send_metadata():
             payload['systemStats'] = get_system_stats()
             payload['meta'] = self._get_metadata()
             self.metadata_cache = payload['meta']
@@ -427,6 +442,8 @@ def _get_metadata(self):
         return metadata
 
     def _should_send_metadata(self):
+        if self._is_first_run():
+            return True
         # If the interval has passed, send the metadata again
         now = time.time()
         if now - self.metadata_start >= self.metadata_interval: