From 8adccfd39e973241e55d115efa7094742c4d6f64 Mon Sep 17 00:00:00 2001
From: Celene <celene@datadoghq.com>
Date: Thu, 10 Apr 2014 18:47:12 +0000
Subject: [PATCH 01/33] fix mysql check duplicate metric

---
 checks.d/mysql.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/checks.d/mysql.py b/checks.d/mysql.py
index 175c14d5d2..56718c24e8 100644
--- a/checks.d/mysql.py
+++ b/checks.d/mysql.py
@@ -18,7 +18,7 @@
     'Innodb_data_reads': ('mysql.innodb.data_reads', RATE),
     'Innodb_data_writes': ('mysql.innodb.data_writes', RATE),
     'Innodb_os_log_fsyncs': ('mysql.innodb.os_log_fsyncs', RATE),
-    'Innodb_data_reads': ('mysql.innodb.buffer_pool_size', RATE),
+    'Innodb_buffer_pool_size': ('mysql.innodb.buffer_pool_size', RATE),
     'Slow_queries': ('mysql.performance.slow_queries', RATE),
     'Questions': ('mysql.performance.questions', RATE),
     'Queries': ('mysql.performance.queries', RATE),

From 724000b13a124134ca2932bac67d9332dd947101 Mon Sep 17 00:00:00 2001
From: Arthur Wang <arthur@datadoghq.com>
Date: Thu, 17 Apr 2014 15:18:55 +0000
Subject: [PATCH 02/33] Fix open_file_descriptors metric name

---
 checks.d/process.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/checks.d/process.py b/checks.d/process.py
index b2a225b6e5..718fd87960 100644
--- a/checks.d/process.py
+++ b/checks.d/process.py
@@ -10,7 +10,7 @@ class ProcessCheck(AgentCheck):
         'system.processes.mem.rss',
         'system.processes.mem.vms',
         'system.processes.mem.real',
-        'system.processes.open_file_decorators',
+        'system.processes.open_file_descriptors',
         'system.processes.ioread_count',
         'system.processes.iowrite_count',
         'system.processes.ioread_bytes',

From 0ae96b11717e01de2a897f56bf0927d19413e81b Mon Sep 17 00:00:00 2001
From: Remi Hakim <remi@datadoghq.com>
Date: Thu, 17 Apr 2014 17:03:14 +0100
Subject: [PATCH 03/33] Typo

---
 checks/check_status.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/checks/check_status.py b/checks/check_status.py
index 37ccc89a2f..508a95516a 100644
--- a/checks/check_status.py
+++ b/checks/check_status.py
@@ -346,7 +346,7 @@ def body_lines(self):
             ntp_offset, ntp_styles = get_ntp_info()
             lines.append('  ' + style('NTP offset', *ntp_styles) + ': ' +  style('%s s' % round(ntp_offset, 4), *ntp_styles))
         except Exception, e:
-            lines.append('  NTP offset: Unkwown (%s)' % str(e))
+            lines.append('  NTP offset: Unknown (%s)' % str(e))
         lines.append('  System UTC time: ' + datetime.datetime.utcnow().__str__())
         lines.append('')
 

From 077ccf5e4ea6d43a02ce2763b882a5ecf5549495 Mon Sep 17 00:00:00 2001
From: Remi Hakim <remi@datadoghq.com>
Date: Mon, 21 Apr 2014 11:22:33 -0400
Subject: [PATCH 04/33] Adding release 5.0.0

---
 CHANGELOG.md | 6 ++++++
 1 file changed, 6 insertions(+)

diff --git a/CHANGELOG.md b/CHANGELOG.md
index fefc31a435..2c93ab4965 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -1,5 +1,11 @@
 Changes
 =======
+# 5.0.0 / Unreleased
+
+### Integrations affected
+
+### Changes
+
 
 # 4.2.1 / 04-09-2014
 

From 3f12340296cf8059f653823556ed3f593f96c9e8 Mon Sep 17 00:00:00 2001
From: Arthur <arthur@datadoghq.com>
Date: Mon, 21 Apr 2014 11:48:10 -0400
Subject: [PATCH 05/33] Update CHANGELOG.md

---
 CHANGELOG.md | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/CHANGELOG.md b/CHANGELOG.md
index 2c93ab4965..e2d85eb536 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -5,7 +5,7 @@ Changes
 ### Integrations affected
 
 ### Changes
-
+* [BUGFIX] Fix incorrect open file descriptors metric name in process check: See [#904]
 
 # 4.2.1 / 04-09-2014
 

From d4c53661803e732570e42d8d7784cca3786bfe4c Mon Sep 17 00:00:00 2001
From: Arthur <arthur@datadoghq.com>
Date: Mon, 21 Apr 2014 11:49:38 -0400
Subject: [PATCH 06/33] Update CHANGELOG.md

---
 CHANGELOG.md | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/CHANGELOG.md b/CHANGELOG.md
index e2d85eb536..da9d98d0d9 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -5,7 +5,7 @@ Changes
 ### Integrations affected
 
 ### Changes
-* [BUGFIX] Fix incorrect open file descriptors metric name in process check: See [#904]
+* [BUGFIX] Fix incorrect open file descriptors metric name in process check: See [#904](https://github.com/DataDog/dd-agent/pull/904)
 
 # 4.2.1 / 04-09-2014
 

From 75ccee4a770f46667f3cf230148d040b53b5bf0f Mon Sep 17 00:00:00 2001
From: Carlo Cabanilla <carlo@datadoghq.com>
Date: Tue, 22 Apr 2014 18:17:52 +0000
Subject: [PATCH 07/33] add ability for AgentCheck instances to collect service
 check data

---
 checks/__init__.py   | 31 +++++++++++++++++++++++++++++++
 tests/test_common.py | 23 +++++++++++++++++++++--
 2 files changed, 52 insertions(+), 2 deletions(-)

diff --git a/checks/__init__.py b/checks/__init__.py
index 74cfa1c038..118bec1e87 100644
--- a/checks/__init__.py
+++ b/checks/__init__.py
@@ -262,6 +262,7 @@ def get_metrics(self, expire=True):
         return metrics
 
 class AgentCheck(object):
+    OK, WARNING, CRITICAL, UNKNOWN, NONE = (0, 1, 2, 3, 4)
 
     def __init__(self, name, init_config, agentConfig, instances=None):
         """
@@ -284,6 +285,7 @@ def __init__(self, name, init_config, agentConfig, instances=None):
         self.aggregator = MetricsAggregator(self.hostname, formatter=agent_formatter, recent_point_threshold=agentConfig.get('recent_point_threshold', None))
 
         self.events = []
+        self.service_checks = []
         self.instances = instances or []
         self.warnings = []
         self.library_versions = None
@@ -392,6 +394,23 @@ def event(self, event):
             event['api_key'] = self.agentConfig['api_key']
         self.events.append(event)
 
+    def service_check(self, service_name, status, tags=None, timestamp=None):
+        """
+        Save a service check.
+
+        :param service_name: string, name of the service
+        :param status: int, describing the status. 0 for success, 1 for
+                       warning, 2 for failure
+        :param tags: list of strings, a list of tags for this run
+        :param timestamp: int, unix timestamp for when the run occurred
+        """
+        self.service_checks.append({
+            'service_name': service_name,
+            'status': status,
+            'tags': tags,
+            'timestamp': int(timestamp or time.time())
+        })
+
     def has_events(self):
         """
         Check whether the check has saved any events
@@ -421,6 +440,18 @@ def get_events(self):
         self.events = []
         return events
 
+    def get_service_checks(self):
+        """
+        Return a list of the service checks saved by the check, if any
+        and clears them out of the instance's service_checks list
+
+        @return the list of service checks saved by this check
+        @rtype list of service check dicts
+        """
+        service_checks = self.service_checks
+        self.service_checks = []
+        return service_checks
+
     def has_warnings(self):
         """
         Check whether the instance run created any warnings
diff --git a/tests/test_common.py b/tests/test_common.py
index 4ec2499d5a..079a423a59 100644
--- a/tests/test_common.py
+++ b/tests/test_common.py
@@ -2,13 +2,14 @@
 import unittest
 import logging
 logger = logging.getLogger()
-from checks import Check, CheckException, UnknownValue, CheckException, Infinity
+from checks import (Check, AgentCheck,
+    CheckException, UnknownValue, CheckException, Infinity)
 from checks.collector import Collector
 from aggregator import MetricsAggregator
 
 class TestCore(unittest.TestCase):
     "Tests to validate the core check logic"
-    
+
     def setUp(self):
         self.c = Check(logger)
         self.c.gauge("test-metric")
@@ -95,6 +96,24 @@ def test_metadata(self):
         assert "socket-fqdn" in c._get_metadata()
         assert "socket-hostname" in c._get_metadata()
 
+    def test_service_check(self):
+        service_name = 'test.service_check'
+        status = AgentCheck.CRITICAL
+        tags = ['host:test', 'other:thing']
+        timestamp = int(time.time())
+
+        check = AgentCheck('test', {}, {})
+        check.service_check(service_name, status, tags, timestamp)
+        self.assertEquals(len(check.service_checks), 1, check.service_checks)
+        val = check.get_service_checks()
+        self.assertEquals([{
+                    'service_name': service_name,
+                    'status': status,
+                    'tags': tags,
+                    'timestamp': timestamp
+                }], val)
+        self.assertEquals(len(check.service_checks), 0, check.service_checks)
+
 class TestAggregator(unittest.TestCase):
     def setUp(self):
         self.aggr = MetricsAggregator('test-aggr')

From b5e24dd53df0e5f488f5c07ce17b20be6f3b4548 Mon Sep 17 00:00:00 2001
From: Carlo Cabanilla <carlo@datadoghq.com>
Date: Tue, 22 Apr 2014 18:36:02 +0000
Subject: [PATCH 08/33] update the collector to collect service check data from
 agent checks

---
 checks/check_status.py | 36 ++++++++++++++-------------
 checks/collector.py    | 56 ++++++++++++++++++++++++------------------
 2 files changed, 51 insertions(+), 41 deletions(-)

diff --git a/checks/check_status.py b/checks/check_status.py
index 508a95516a..2f2fb543c5 100644
--- a/checks/check_status.py
+++ b/checks/check_status.py
@@ -188,7 +188,7 @@ def to_dict(self):
     def _not_running_message(cls):
         lines = cls._title_lines() + [
             style("  %s is not running." % cls.NAME, 'red'),
-            style("""  You can get more details in the logs: 
+            style("""  You can get more details in the logs:
     %s""" % logger_info(), 'red'),
             "",
             ""
@@ -264,13 +264,15 @@ def has_warnings(self):
 
 class CheckStatus(object):
 
-    def __init__(self, check_name, instance_statuses, metric_count,
-                 event_count, init_failed_error=None,
-                 init_failed_traceback=None, library_versions=None):
+    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):
         self.name = check_name
         self.instance_statuses = instance_statuses
-        self.metric_count = metric_count
-        self.event_count = event_count
+        self.metric_count = metric_count or 0
+        self.event_count = event_count or 0
+        self.service_check_count = service_check_count or 0
         self.init_failed_error = init_failed_error
         self.init_failed_traceback = init_failed_traceback
         self.library_versions = library_versions
@@ -363,7 +365,7 @@ def body_lines(self):
             confd_path = config.get_confd_path(osname)
         except config.PathNotFound:
             confd_path = 'Not found'
-        
+
         try:
             checksd_path = config.get_checksd_path(osname)
         except config.PathNotFound:
@@ -442,7 +444,7 @@ def body_lines(self):
                                            s.traceback.split('\n'))
 
                     check_lines += [
-                        "    - Collected %s metrics & %s events" % (cs.metric_count, cs.event_count),
+                        "    - Collected %s metrics, %s events & %s service checks" % (cs.metric_count, cs.event_count, cs.service_check_count),
                     ]
 
                     if cs.library_versions is not None:
@@ -537,7 +539,7 @@ def to_dict(self):
             status_info['confd_path'] = config.get_confd_path(osname)
         except config.PathNotFound:
             status_info['confd_path'] = 'Not found'
-        
+
         try:
             status_info['checksd_path'] = config.get_checksd_path(osname)
         except config.PathNotFound:
@@ -650,7 +652,7 @@ def get_jmx_status():
 
         - One generated by jmxfetch that return information about the collection of metrics
         its format is as following:
-    
+
         ###
         timestamp: 1391037347435
         checks:
@@ -679,10 +681,10 @@ def get_jmx_status():
 
             if status_age > 60:
                 check_statuses.append(CheckStatus("jmx", [InstanceStatus(
-                                                    0, 
-                                                    STATUS_ERROR, 
+                                                    0,
+                                                    STATUS_ERROR,
                                                     error="JMXfetch didn't return any metrics during the last minute"
-                                                    )], 0, 0))
+                                                    )]))
             else:
 
                 for check_name, instances in jmx_checks.get('failed_checks', {}).iteritems():
@@ -693,7 +695,7 @@ def get_jmx_status():
                         instance_name = info.get('instance_name', None)
                         check_data[check_name]['statuses'].append(get_jmx_instance_status(instance_name, status, message, metric_count))
                         check_data[check_name]['metric_count'].append(metric_count)
-               
+
                 for check_name, instances in jmx_checks.get('initialized_checks', {}).iteritems():
                     for info in instances:
                         message = info.get('message', None)
@@ -704,18 +706,18 @@ def get_jmx_status():
                         check_data[check_name]['metric_count'].append(metric_count)
 
                 for check_name, data in check_data.iteritems():
-                    check_status = CheckStatus(check_name, data['statuses'], sum(data['metric_count']), 0)
+                    check_status = CheckStatus(check_name, data['statuses'], sum(data['metric_count']))
                     check_statuses.append(check_status)
 
         if os.path.exists(python_status_path):
             python_jmx_stats = yaml.load(file(python_status_path))
             jmx_checks = python_jmx_stats.get('invalid_checks', {})
             for check_name, excep in jmx_checks.iteritems():
-                check_statuses.append(CheckStatus(check_name, [], 0, 0, init_failed_error=excep))
+                check_statuses.append(CheckStatus(check_name, [], init_failed_error=excep))
 
 
         return check_statuses
 
-    except Exception, e:
+    except Exception:
         log.exception("Couldn't load latest jmx status")
         return []
diff --git a/checks/collector.py b/checks/collector.py
index d467354c04..bf55cb6da6 100644
--- a/checks/collector.py
+++ b/checks/collector.py
@@ -43,7 +43,7 @@ def __init__(self, agentConfig, emitters, systemStats):
         # agent config is used during checks, system_stats can be accessed through the config
         self.os = get_os()
         self.plugins = None
-        self.emitters = emitters            
+        self.emitters = emitters
         self.metadata_interval = int(agentConfig.get('metadata_interval', 10 * 60))
         self.metadata_start = time.time()
         socket.setdefaulttimeout(15)
@@ -107,7 +107,7 @@ def stop(self):
         """
         Tell the collector to stop at the next logical point.
         """
-        # This is called when the process is being killed, so 
+        # This is called when the process is being killed, so
         # try to stop the collector as soon as possible.
         # Most importantly, don't try to submit to the emitters
         # because the forwarder is quite possibly already killed
@@ -116,7 +116,7 @@ def stop(self):
         self.continue_running = False
         for check in self.initialized_checks_d:
             check.stop()
-    
+
     def run(self, checksd=None, start_event=True):
         """
         Collect data from each check and submit their data.
@@ -130,6 +130,7 @@ def run(self, checksd=None, start_event=True):
         payload = self._build_payload(start_event=start_event)
         metrics = payload['metrics']
         events = payload['events']
+        service_checks = payload['service_checks']
         if checksd:
             self.initialized_checks_d = checksd['initialized_checks'] # is of type {check_name: check}
             self.init_failed_checks_d = checksd['init_failed_checks'] # is of type {check_name: {error, traceback}}
@@ -156,21 +157,21 @@ def run(self, checksd=None, start_event=True):
 
             load = sys_checks['load'].check(self.agentConfig)
             payload.update(load)
-                
+
             memory = sys_checks['memory'].check(self.agentConfig)
 
             if memory:
                 payload.update({
-                    'memPhysUsed' : memory.get('physUsed'), 
-                    'memPhysPctUsable' : memory.get('physPctUsable'), 
-                    'memPhysFree' : memory.get('physFree'), 
-                    'memPhysTotal' : memory.get('physTotal'), 
-                    'memPhysUsable' : memory.get('physUsable'), 
-                    'memSwapUsed' : memory.get('swapUsed'), 
-                    'memSwapFree' : memory.get('swapFree'), 
-                    'memSwapPctFree' : memory.get('swapPctFree'), 
-                    'memSwapTotal' : memory.get('swapTotal'), 
-                    'memCached' : memory.get('physCached'), 
+                    'memPhysUsed' : memory.get('physUsed'),
+                    'memPhysPctUsable' : memory.get('physPctUsable'),
+                    'memPhysFree' : memory.get('physFree'),
+                    'memPhysTotal' : memory.get('physTotal'),
+                    'memPhysUsable' : memory.get('physUsable'),
+                    'memSwapUsed' : memory.get('swapUsed'),
+                    'memSwapFree' : memory.get('swapFree'),
+                    'memSwapPctFree' : memory.get('swapPctFree'),
+                    'memSwapTotal' : memory.get('swapTotal'),
+                    'memCached' : memory.get('physCached'),
                     'memBuffers': memory.get('physBuffers'),
                     'memShared': memory.get('physShared')
                 })
@@ -194,7 +195,7 @@ def run(self, checksd=None, start_event=True):
 
         if gangliaData is not False and gangliaData is not None:
             payload['ganglia'] = gangliaData
-           
+
         # dogstream
         if dogstreamData:
             dogstreamEvents = dogstreamData.get('dogstreamEvents', None)
@@ -211,7 +212,7 @@ def run(self, checksd=None, start_event=True):
         if ddforwarderData:
             payload['datadog'] = ddforwarderData
 
-        # Process the event checks. 
+        # Process the event checks.
         for event_check in self._event_checks:
             event_data = event_check.check(log, self.agentConfig)
             if event_data:
@@ -226,12 +227,12 @@ def run(self, checksd=None, start_event=True):
                 if snaps:
                     has_resource = True
                     res_value = { 'snaps': snaps,
-                                  'format_version': resources_check.get_format_version() }                              
+                                  'format_version': resources_check.get_format_version() }
                     res_format = resources_check.describe_format_if_needed()
                     if res_format is not None:
                         res_value['format_description'] = res_format
                     payload['resources'][resources_check.RESOURCE_KEY] = res_value
-     
+
             if has_resource:
                 payload['resources']['meta'] = {
                             'api_key': self.agentConfig['api_key'],
@@ -250,7 +251,7 @@ def run(self, checksd=None, start_event=True):
             if not self.continue_running:
                 return
             log.info("Running check %s" % check.name)
-            instance_statuses = [] 
+            instance_statuses = []
             metric_count = 0
             event_count = 0
             try:
@@ -260,6 +261,7 @@ def run(self, checksd=None, start_event=True):
                 # Collect the metrics and events.
                 current_check_metrics = check.get_metrics()
                 current_check_events = check.get_events()
+                current_check_service_checks = check.get_service_checks()
 
                 # Save them for the payload.
                 metrics.extend(current_check_metrics)
@@ -268,20 +270,24 @@ def run(self, checksd=None, start_event=True):
                         events[check.name] = current_check_events
                     else:
                         events[check.name] += current_check_events
+                if current_check_service_checks:
+                    service_checks.extend(current_check_service_checks)
 
                 # Save the status of the check.
                 metric_count = len(current_check_metrics)
                 event_count = len(current_check_events)
-            except Exception, e:
+                service_check_count = len(current_check_service_checks)
+            except Exception:
                 log.exception("Error running check %s" % check.name)
 
-            check_status = CheckStatus(check.name, instance_statuses, metric_count, event_count, library_versions=check.get_library_info())
+            check_status = CheckStatus(check.name, instance_statuses, metric_count, event_count, service_check_count,
+                library_versions=check.get_library_info())
             check_statuses.append(check_status)
 
         for check_name, info in self.init_failed_checks_d.iteritems():
             if not self.continue_running:
                 return
-            check_status = CheckStatus(check_name, None, None, None,
+            check_status = CheckStatus(check_name, None, None, None, None,
                                        init_failed_error=info['error'],
                                        init_failed_traceback=info['traceback'])
             check_statuses.append(check_status)
@@ -290,13 +296,14 @@ def run(self, checksd=None, start_event=True):
         # Store the metrics and events in the payload.
         payload['metrics'] = metrics
         payload['events'] = events
+        payload['service_checks'] = service_checks
         collect_duration = timer.step()
 
         if self.os != 'windows':
-            payload['metrics'].extend(self._agent_metrics.check(payload, self.agentConfig, 
+            payload['metrics'].extend(self._agent_metrics.check(payload, self.agentConfig,
                 collect_duration, self.emit_duration, time.clock() - cpu_clock))
         else:
-            payload['metrics'].extend(self._agent_metrics.check(payload, self.agentConfig, 
+            payload['metrics'].extend(self._agent_metrics.check(payload, self.agentConfig,
                 collect_duration, self.emit_duration))
 
 
@@ -353,6 +360,7 @@ def _build_payload(self, start_event=True):
             'apiKey': self.agentConfig['api_key'],
             'events': {},
             'metrics': [],
+            'service_checks': [],
             'resources': {},
             'internalHostname' : get_hostname(self.agentConfig),
             'uuid' : get_uuid(),

From ce97e7b1c5933c4e2119afe5ee6870264a41cd3a Mon Sep 17 00:00:00 2001
From: Remi Hakim <remi@datadoghq.com>
Date: Tue, 22 Apr 2014 15:21:39 -0400
Subject: [PATCH 09/33] Upgrade to cassandra 2.0.7 to fix travis tests

---
 .travis.yml | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/.travis.yml b/.travis.yml
index 6e0acd4826..e68c58c981 100644
--- a/.travis.yml
+++ b/.travis.yml
@@ -31,9 +31,9 @@ install:
   - pip install . --use-mirrors
 before_script:
   - curl -L https://raw.github.com/DataDog/dd-agent/master/tests/haproxy.cfg > /tmp/haproxy.cfg
-  - curl -L http://apache.mirrors.multidist.eu/cassandra/2.0.6/apache-cassandra-2.0.6-bin.tar.gz > /tmp/cassandra2.tar.gz
+  - curl -L http://apache.mirrors.multidist.eu/cassandra/2.0.7/apache-cassandra-2.0.7-bin.tar.gz > /tmp/cassandra2.tar.gz
   - tar -xzvf /tmp/cassandra2.tar.gz -C /tmp
-  - sudo /tmp/apache-cassandra-2.0.6/bin/cassandra
+  - sudo /tmp/apache-cassandra-2.0.7/bin/cassandra
   - sudo service haproxy restart
   - sudo bash -c "curl -L https://raw.github.com/DataDog/dd-agent/master/tests/tomcat_cfg.xml > /etc/tomcat6/server.xml"
   - sudo bash -c "curl -L https://raw.github.com/DataDog/dd-agent/master/tests/tomcat6 >> /etc/default/tomcat6"

From 851dfad9059396c01d3adf297bda7cae68351b83 Mon Sep 17 00:00:00 2001
From: Carlo Cabanilla <carlo@datadoghq.com>
Date: Thu, 24 Apr 2014 15:16:22 +0000
Subject: [PATCH 10/33] rename service_name to check_name

Services could have multiple checks
---
 checks/__init__.py | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/checks/__init__.py b/checks/__init__.py
index 118bec1e87..4520c6a4e4 100644
--- a/checks/__init__.py
+++ b/checks/__init__.py
@@ -394,18 +394,18 @@ def event(self, event):
             event['api_key'] = self.agentConfig['api_key']
         self.events.append(event)
 
-    def service_check(self, service_name, status, tags=None, timestamp=None):
+    def service_check(self, check_name, status, tags=None, timestamp=None):
         """
         Save a service check.
 
-        :param service_name: string, name of the service
+        :param check_name: string, name of the service check
         :param status: int, describing the status. 0 for success, 1 for
                        warning, 2 for failure
         :param tags: list of strings, a list of tags for this run
         :param timestamp: int, unix timestamp for when the run occurred
         """
         self.service_checks.append({
-            'service_name': service_name,
+            'check': check_name,
             'status': status,
             'tags': tags,
             'timestamp': int(timestamp or time.time())

From c42f70216c6aa5d05f243e953760713cd2885227 Mon Sep 17 00:00:00 2001
From: Carlo Cabanilla <carlo@datadoghq.com>
Date: Thu, 24 Apr 2014 16:18:05 +0000
Subject: [PATCH 11/33] allow for float timestamps in service checks

---
 checks/__init__.py | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/checks/__init__.py b/checks/__init__.py
index 4520c6a4e4..2310e79df8 100644
--- a/checks/__init__.py
+++ b/checks/__init__.py
@@ -402,13 +402,13 @@ def service_check(self, check_name, status, tags=None, timestamp=None):
         :param status: int, describing the status. 0 for success, 1 for
                        warning, 2 for failure
         :param tags: list of strings, a list of tags for this run
-        :param timestamp: int, unix timestamp for when the run occurred
+        :param timestamp: float, unix timestamp for when the run occurred
         """
         self.service_checks.append({
             'check': check_name,
             'status': status,
             'tags': tags,
-            'timestamp': int(timestamp or time.time())
+            'timestamp': float(timestamp or time.time())
         })
 
     def has_events(self):

From babb7379db6cfd3a5ab19fa4c95365290245be80 Mon Sep 17 00:00:00 2001
From: Carlo Cabanilla <carlo@datadoghq.com>
Date: Thu, 24 Apr 2014 16:27:44 +0000
Subject: [PATCH 12/33] fix service check test

---
 tests/test_common.py | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/tests/test_common.py b/tests/test_common.py
index 079a423a59..8736c50643 100644
--- a/tests/test_common.py
+++ b/tests/test_common.py
@@ -100,14 +100,14 @@ def test_service_check(self):
         service_name = 'test.service_check'
         status = AgentCheck.CRITICAL
         tags = ['host:test', 'other:thing']
-        timestamp = int(time.time())
+        timestamp = time.time()
 
         check = AgentCheck('test', {}, {})
         check.service_check(service_name, status, tags, timestamp)
         self.assertEquals(len(check.service_checks), 1, check.service_checks)
         val = check.get_service_checks()
         self.assertEquals([{
-                    'service_name': service_name,
+                    'check': service_name,
                     'status': status,
                     'tags': tags,
                     'timestamp': timestamp

From a94060cfe939c97e8c053cdf7fbf73be656796b6 Mon Sep 17 00:00:00 2001
From: Carlo Cabanilla <carlo@datadoghq.com>
Date: Thu, 24 Apr 2014 16:30:22 +0000
Subject: [PATCH 13/33] add generated ids to service checks

---
 checks/__init__.py   | 6 ++++--
 tests/test_common.py | 4 ++++
 util.py              | 8 ++++++++
 3 files changed, 16 insertions(+), 2 deletions(-)

diff --git a/checks/__init__.py b/checks/__init__.py
index 2310e79df8..8e6304e973 100644
--- a/checks/__init__.py
+++ b/checks/__init__.py
@@ -15,7 +15,7 @@
 import copy
 from pprint import pprint
 
-from util import LaconicFilter, get_os, get_hostname
+from util import LaconicFilter, get_os, get_hostname, get_next_id
 from config import get_confd_path
 from checks import check_status
 
@@ -394,7 +394,8 @@ def event(self, event):
             event['api_key'] = self.agentConfig['api_key']
         self.events.append(event)
 
-    def service_check(self, check_name, status, tags=None, timestamp=None):
+    def service_check(self, check_name, status, tags=None, timestamp=None,
+                      check_id=None):
         """
         Save a service check.
 
@@ -405,6 +406,7 @@ def service_check(self, check_name, status, tags=None, timestamp=None):
         :param timestamp: float, unix timestamp for when the run occurred
         """
         self.service_checks.append({
+            'id': check_id or get_next_id('service_check'),
             'check': check_name,
             'status': status,
             'tags': tags,
diff --git a/tests/test_common.py b/tests/test_common.py
index 8736c50643..b396fd50c5 100644
--- a/tests/test_common.py
+++ b/tests/test_common.py
@@ -106,7 +106,11 @@ def test_service_check(self):
         check.service_check(service_name, status, tags, timestamp)
         self.assertEquals(len(check.service_checks), 1, check.service_checks)
         val = check.get_service_checks()
+        self.assertEquals(len(val), 1)
+        service_check_id = val[0].get('id', None)
+        self.assertNotEquals(service_check_id, None)
         self.assertEquals([{
+                    'id': service_check_id,
                     'check': service_name,
                     'status': status,
                     'tags': tags,
diff --git a/util.py b/util.py
index 4d1855d14d..845c84aa75 100644
--- a/util.py
+++ b/util.py
@@ -149,6 +149,14 @@ def cast_metric_val(val):
         raise ValueError
     return val
 
+_IDS = {}
+def get_next_id(name):
+    global _IDS
+    current_id = _IDS.get(name, 0)
+    current_id += 1
+    _IDS[name] = current_id
+    return current_id
+
 def is_valid_hostname(hostname):
     if hostname.lower() in set([
         'localhost',

From 8d9410ba33e67b597a0d13fde1a79555dde05dd9 Mon Sep 17 00:00:00 2001
From: Carlo Cabanilla <carlo@datadoghq.com>
Date: Thu, 24 Apr 2014 16:34:37 +0000
Subject: [PATCH 14/33] allow service_check host_name be overridable,
 defaulting to agent host name

---
 checks/__init__.py   | 9 +++++++--
 tests/test_common.py | 4 +++-
 2 files changed, 10 insertions(+), 3 deletions(-)

diff --git a/checks/__init__.py b/checks/__init__.py
index 8e6304e973..01b70b4142 100644
--- a/checks/__init__.py
+++ b/checks/__init__.py
@@ -395,7 +395,7 @@ def event(self, event):
         self.events.append(event)
 
     def service_check(self, check_name, status, tags=None, timestamp=None,
-                      check_id=None):
+                      host_name=None, check_id=None):
         """
         Save a service check.
 
@@ -405,10 +405,15 @@ def service_check(self, check_name, status, tags=None, timestamp=None,
         :param tags: list of strings, a list of tags for this run
         :param timestamp: float, unix timestamp for when the run occurred
         """
+        if host_name is None:
+            host_name = self.host_name
+        if check_id is None:
+            check_id = get_next_id('service_check')
         self.service_checks.append({
-            'id': check_id or get_next_id('service_check'),
+            'id': check_id,
             'check': check_name,
             'status': status,
+            'host_name': host_name,
             'tags': tags,
             'timestamp': float(timestamp or time.time())
         })
diff --git a/tests/test_common.py b/tests/test_common.py
index b396fd50c5..a1ce73e392 100644
--- a/tests/test_common.py
+++ b/tests/test_common.py
@@ -100,10 +100,11 @@ def test_service_check(self):
         service_name = 'test.service_check'
         status = AgentCheck.CRITICAL
         tags = ['host:test', 'other:thing']
+        host_name = 'foohost'
         timestamp = time.time()
 
         check = AgentCheck('test', {}, {})
-        check.service_check(service_name, status, tags, timestamp)
+        check.service_check(service_name, status, tags, timestamp, host_name)
         self.assertEquals(len(check.service_checks), 1, check.service_checks)
         val = check.get_service_checks()
         self.assertEquals(len(val), 1)
@@ -113,6 +114,7 @@ def test_service_check(self):
                     'id': service_check_id,
                     'check': service_name,
                     'status': status,
+                    'host_name': host_name,
                     'tags': tags,
                     'timestamp': timestamp
                 }], val)

From 6dc99fe4cf4e226b5fb7a3e59360cdf6903919cc Mon Sep 17 00:00:00 2001
From: Carlo Cabanilla <carlo@datadoghq.com>
Date: Thu, 24 Apr 2014 17:54:20 +0000
Subject: [PATCH 15/33] update service check docstring

---
 checks/__init__.py | 13 +++++++++----
 1 file changed, 9 insertions(+), 4 deletions(-)

diff --git a/checks/__init__.py b/checks/__init__.py
index 01b70b4142..f76bebfed0 100644
--- a/checks/__init__.py
+++ b/checks/__init__.py
@@ -400,10 +400,15 @@ def service_check(self, check_name, status, tags=None, timestamp=None,
         Save a service check.
 
         :param check_name: string, name of the service check
-        :param status: int, describing the status. 0 for success, 1 for
-                       warning, 2 for failure
-        :param tags: list of strings, a list of tags for this run
-        :param timestamp: float, unix timestamp for when the run occurred
+        :param status: int, describing the status.
+                       0 for success, 1 for warning, 2 for failure
+        :param tags: (optional) list of strings, a list of tags for this run
+        :param timestamp: (optional) float, unix timestamp for when the run occurred
+        :param host_name: (optional) str, host that generated the service
+                          check. Defaults to the host_name of the agent
+        :param check_id: (optional) int, id used for logging and tracing
+                         purposes. Don't need to be unique. If not specified,
+                         one will be generated.
         """
         if host_name is None:
             host_name = self.host_name

From bbb53e36163689bd36dd58be86b7cf7b94bc6ece Mon Sep 17 00:00:00 2001
From: Carlo Cabanilla <carlo@datadoghq.com>
Date: Thu, 24 Apr 2014 17:56:40 +0000
Subject: [PATCH 16/33] rename check_id to check_run_id for clarify

---
 checks/__init__.py   | 14 +++++++-------
 tests/test_common.py | 12 ++++++------
 2 files changed, 13 insertions(+), 13 deletions(-)

diff --git a/checks/__init__.py b/checks/__init__.py
index f76bebfed0..ad3f4bba80 100644
--- a/checks/__init__.py
+++ b/checks/__init__.py
@@ -395,7 +395,7 @@ def event(self, event):
         self.events.append(event)
 
     def service_check(self, check_name, status, tags=None, timestamp=None,
-                      host_name=None, check_id=None):
+                      host_name=None, check_run_id=None):
         """
         Save a service check.
 
@@ -406,16 +406,16 @@ def service_check(self, check_name, status, tags=None, timestamp=None,
         :param timestamp: (optional) float, unix timestamp for when the run occurred
         :param host_name: (optional) str, host that generated the service
                           check. Defaults to the host_name of the agent
-        :param check_id: (optional) int, id used for logging and tracing
-                         purposes. Don't need to be unique. If not specified,
-                         one will be generated.
+        :param check_run_id: (optional) int, id used for logging and tracing
+                             purposes. Don't need to be unique. If not
+                             specified, one will be generated.
         """
         if host_name is None:
             host_name = self.host_name
-        if check_id is None:
-            check_id = get_next_id('service_check')
+        if check_run_id is None:
+            check_run_id = get_next_id('service_check')
         self.service_checks.append({
-            'id': check_id,
+            'id': check_run_id,
             'check': check_name,
             'status': status,
             'host_name': host_name,
diff --git a/tests/test_common.py b/tests/test_common.py
index a1ce73e392..842d889390 100644
--- a/tests/test_common.py
+++ b/tests/test_common.py
@@ -97,22 +97,22 @@ def test_metadata(self):
         assert "socket-hostname" in c._get_metadata()
 
     def test_service_check(self):
-        service_name = 'test.service_check'
+        check_name = 'test.service_check'
         status = AgentCheck.CRITICAL
         tags = ['host:test', 'other:thing']
         host_name = 'foohost'
         timestamp = time.time()
 
         check = AgentCheck('test', {}, {})
-        check.service_check(service_name, status, tags, timestamp, host_name)
+        check.service_check(check_name, status, tags, timestamp, host_name)
         self.assertEquals(len(check.service_checks), 1, check.service_checks)
         val = check.get_service_checks()
         self.assertEquals(len(val), 1)
-        service_check_id = val[0].get('id', None)
-        self.assertNotEquals(service_check_id, None)
+        check_run_id = val[0].get('id', None)
+        self.assertNotEquals(check_run_id, None)
         self.assertEquals([{
-                    'id': service_check_id,
-                    'check': service_name,
+                    'id': check_run_id,
+                    'check': check_name,
                     'status': status,
                     'host_name': host_name,
                     'tags': tags,

From 2d2287d30d0caa0081dba96b405f427fc64042e4 Mon Sep 17 00:00:00 2001
From: Remi Hakim <remi@datadoghq.com>
Date: Fri, 25 Apr 2014 13:50:29 -0400
Subject: [PATCH 17/33] Support redis check on Windows

---
 setup.py     | 2 ++
 win32/gui.py | 2 +-
 2 files changed, 3 insertions(+), 1 deletion(-)

diff --git a/setup.py b/setup.py
index 5c5991ed71..08b5f23024 100644
--- a/setup.py
+++ b/setup.py
@@ -43,6 +43,7 @@
         'pycurl',
         'MySQLdb',
         'psutil',
+        'redis',
     ])
 
     # Modules to force-include in the exe
@@ -60,6 +61,7 @@
         'MySQLdb',
         'psutil',
         'psycopg2',
+        'redis',
 
         # agent
         'checks.services_checks',
diff --git a/win32/gui.py b/win32/gui.py
index 71ec1d928e..598ed02a6d 100644
--- a/win32/gui.py
+++ b/win32/gui.py
@@ -35,7 +35,7 @@
 EXCLUDED_WINDOWS_CHECKS = [
     'cacti', 'directory', 'gearmand',
     'hdfs', 'kafka_consumer', 'mcache', 'network',
-    'redis', 'postfix', 'process', 'gunicorn', 'zk',
+    'postfix', 'process', 'gunicorn', 'zk',
     ]
 
 MAIN_WINDOW_TITLE = "Datadog Agent Manager"

From be2de5480fceab13d64d5c8281473b57fdde35c8 Mon Sep 17 00:00:00 2001
From: Remi Hakim <remi@datadoghq.com>
Date: Fri, 25 Apr 2014 14:13:12 -0400
Subject: [PATCH 18/33] Fix the way we add redis in the package

---
 setup.py | 1 -
 1 file changed, 1 deletion(-)

diff --git a/setup.py b/setup.py
index 08b5f23024..f7ecb92bc0 100644
--- a/setup.py
+++ b/setup.py
@@ -37,7 +37,6 @@
         'pymongo==2.3',
         'psycopg2',
         'python-memcached==1.48',
-        'redis==2.6.2',
         'adodbapi'
         'elementtree',
         'pycurl',

From 937d37c700cad579190756290483a3b7fed3bd28 Mon Sep 17 00:00:00 2001
From: Remi Hakim <remi@datadoghq.com>
Date: Fri, 25 Apr 2014 15:30:24 -0400
Subject: [PATCH 19/33] Bump version to 4.2.2

---
 config.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/config.py b/config.py
index c79e4b3d42..e6a57f8ccd 100644
--- a/config.py
+++ b/config.py
@@ -63,7 +63,7 @@ def get_parsed_args():
 
 
 def get_version():
-    return "4.2.1"
+    return "4.2.2"
 
 def skip_leading_wsp(f):
     "Works on a file, returns a file-like object"

From 272f4ea78f35628f9340a61d4996f005e8e76919 Mon Sep 17 00:00:00 2001
From: Remi Hakim <remi@datadoghq.com>
Date: Fri, 25 Apr 2014 15:53:48 -0400
Subject: [PATCH 20/33] Update changelog for 4.2.2

---
 CHANGELOG.md | 16 ++++++++++++++--
 1 file changed, 14 insertions(+), 2 deletions(-)

diff --git a/CHANGELOG.md b/CHANGELOG.md
index da9d98d0d9..77daae9558 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -5,7 +5,17 @@ Changes
 ### Integrations affected
 
 ### Changes
-* [BUGFIX] Fix incorrect open file descriptors metric name in process check: See [#904](https://github.com/DataDog/dd-agent/pull/904)
+* [BUGFIX] Fix incorrect open file descriptors metric name in process check: See [#904]
+
+# 4.2.2 / 04-25-2014
+
+**Windows Only**
+
+### Integrations affected
+* Redis
+
+### Changes
+* [FEATURE] Support Redis check on Windows: See [#917]
 
 # 4.2.1 / 04-09-2014
 
@@ -958,6 +968,8 @@ If you use ganglia, you want this version.
 [#883]: https://github.com/DataDog/dd-agent/issues/883
 [#893]: https://github.com/DataDog/dd-agent/issues/893
 [#894]: https://github.com/DataDog/dd-agent/issues/894
+[#904]: https://github.com/DataDog/dd-agent/issues/904
+[#917]: https://github.com/DataDog/dd-agent/issues/917
 [@CaptTofu]: https://github.com/CaptTofu
 [@brettlangdon]: https://github.com/brettlangdon
 [@charles-dyfis-net]: https://github.com/charles-dyfis-net
@@ -982,4 +994,4 @@ If you use ganglia, you want this version.
 [@steeve]: https://github.com/steeve
 [@stefan-mees]: https://github.com/stefan-mees
 [@tomduckering]: https://github.com/tomduckering
-[@walkeran]: https://github.com/walkeran
+[@walkeran]: https://github.com/walkeran
\ No newline at end of file

From a995fa2aad5137a57877532367c7d55dcbb56fd0 Mon Sep 17 00:00:00 2001
From: graemej <graeme.johnson@jadedpixel.com>
Date: Mon, 28 Apr 2014 12:06:39 -0400
Subject: [PATCH 21/33] basic monitoring support for Apache Mesos

---
 checks.d/mesos.py         | 113 ++++++++++++++++++++++++++++++++++++++
 conf.d/mesos.yaml.example |   7 +++
 2 files changed, 120 insertions(+)
 create mode 100644 checks.d/mesos.py
 create mode 100644 conf.d/mesos.yaml.example

diff --git a/checks.d/mesos.py b/checks.d/mesos.py
new file mode 100644
index 0000000000..ac4bb2c426
--- /dev/null
+++ b/checks.d/mesos.py
@@ -0,0 +1,113 @@
+import time
+import requests
+
+from checks import AgentCheck
+from util import json, headers
+from hashlib import md5
+import urllib2
+
+class Mesos(AgentCheck):
+    def check(self, instance):
+        if 'url' not in instance:
+            self.log.info("Skipping instance, no url found.")
+            return
+
+        # Load values from the instance config
+        url = instance['url']
+        default_timeout = self.init_config.get('default_timeout', 5)
+        timeout = float(instance.get('timeout', default_timeout))
+
+        response = self.master_roles(url, timeout)
+        if response is not None:
+            for role in response['roles']:
+                tags = ['mesos','role:' + role['name']]
+                self.gauge('mesos.role.frameworks', len(role['frameworks']), tags=tags)
+                self.gauge('mesos.role.weight', role['weight'], tags=tags)
+                resources = role['resources']
+                for attr in ['cpus','mem']:
+                    if attr in resources:
+                        self.gauge('mesos.role.' + attr, resources[attr], tags=tags)
+
+        response = self.master_stats(url, timeout)
+        if response is not None:
+            for key in iter(response):
+                self.gauge('mesos.stats.' + key, response[key], tags=['mesos'])
+
+        response = self.master_state(url, timeout)
+        if response is not None:
+            for attr in ['deactivated_slaves','failed_tasks','finished_tasks','killed_tasks','lost_tasks','staged_tasks','started_tasks']:
+                tags = ['mesos']
+                self.gauge('mesos.state.' + attr, response[attr], tags=tags)
+
+            for framework in response['frameworks']:
+                tags = ['mesos','framework:' + framework['id']]
+                resources = framework['resources']
+                for attr in ['cpus','mem']:
+                    if attr in resources:
+                        self.gauge('mesos.state.framework.' + attr, resources[attr], tags=tags)
+
+            for slave in response['slaves']:
+                tags = ['mesos','slave:' + slave['id']]
+                resources = slave['resources']
+                for attr in ['cpus','mem','disk']:
+                    if attr in resources:
+                        self.gauge('mesos.state.slave.' + attr, resources[attr], tags=tags)
+
+    def master_roles(self, url, timeout):
+        return self.get_json(url + "/master/roles.json", timeout)
+
+    def master_stats(self, url, timeout):
+        return self.get_json(url + "/master/stats.json", timeout)
+
+    def master_state(self, url, timeout):
+        return self.get_json(url + "/master/state.json", timeout)
+
+    def get_json(self, url, timeout):
+        # Use a hash of the URL as an aggregation key
+        aggregation_key = md5(url).hexdigest()
+
+        try:
+            response = requests.get(url, timeout=timeout)
+            parsed = response.json()
+            return parsed
+        except requests.exceptions.Timeout as e:
+            # If there's a timeout
+            self.timeout_event(url, timeout, aggregation_key)
+            return None
+
+        if r.status_code != 200:
+            self.status_code_event(url, r, aggregation_key)
+            return None
+
+
+    def timeout_event(self, url, timeout, aggregation_key):
+        self.event({
+            'timestamp': int(time.time()),
+            'event_type': 'http_check',
+            'msg_title': 'URL timeout',
+            'msg_text': '%s timed out after %s seconds.' % (url, timeout),
+            'aggregation_key': aggregation_key
+        })
+
+    def status_code_event(self, url, r, aggregation_key):
+        self.event({
+            'timestamp': int(time.time()),
+            'event_type': 'http_check',
+            'msg_title': 'Invalid reponse code for %s' % url,
+            'msg_text': '%s returned a status of %s' % (url, r.status_code),
+            'aggregation_key': aggregation_key
+        })
+
+if __name__ == '__main__':
+    check, instances = Mesos.from_yaml('/etc/dd-agent/conf.d/mesos.yaml')
+    for instance in instances:
+        print "\nRunning the check against url: %s" % (instance['url'])
+        check.check(instance)
+        if check.has_events():
+            print 'Events: %s' % (check.get_events())
+
+        i = 0
+        print 'Metrics:\n'
+        for metric in check.get_metrics():
+            print "  %d: %s" % (i, metric)
+            i += 1
\ No newline at end of file
diff --git a/conf.d/mesos.yaml.example b/conf.d/mesos.yaml.example
new file mode 100644
index 0000000000..c4527bc9c6
--- /dev/null
+++ b/conf.d/mesos.yaml.example
@@ -0,0 +1,7 @@
+init_config:
+# time to wait on a Mesos API request
+#    default_timeout: 5
+
+instances:
+# url: the API endpoint of your Mesos master
+#    - url: "https://server:port"
\ No newline at end of file

From 1a30c74861ae1198c82f9974213d0e7ce313f26f Mon Sep 17 00:00:00 2001
From: Carlo Cabanilla <carlo@datadoghq.com>
Date: Mon, 28 Apr 2014 19:00:55 +0000
Subject: [PATCH 22/33] fix hostname vs host_name

---
 checks/__init__.py | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)

diff --git a/checks/__init__.py b/checks/__init__.py
index ad3f4bba80..0476c969ee 100644
--- a/checks/__init__.py
+++ b/checks/__init__.py
@@ -395,7 +395,7 @@ def event(self, event):
         self.events.append(event)
 
     def service_check(self, check_name, status, tags=None, timestamp=None,
-                      host_name=None, check_run_id=None):
+                      hostname=None, check_run_id=None):
         """
         Save a service check.
 
@@ -404,21 +404,21 @@ def service_check(self, check_name, status, tags=None, timestamp=None,
                        0 for success, 1 for warning, 2 for failure
         :param tags: (optional) list of strings, a list of tags for this run
         :param timestamp: (optional) float, unix timestamp for when the run occurred
-        :param host_name: (optional) str, host that generated the service
+        :param hostname: (optional) str, host that generated the service
                           check. Defaults to the host_name of the agent
         :param check_run_id: (optional) int, id used for logging and tracing
                              purposes. Don't need to be unique. If not
                              specified, one will be generated.
         """
-        if host_name is None:
-            host_name = self.host_name
+        if hostname is None:
+            hostname = self.hostname
         if check_run_id is None:
             check_run_id = get_next_id('service_check')
         self.service_checks.append({
             'id': check_run_id,
             'check': check_name,
             'status': status,
-            'host_name': host_name,
+            'host_name': hostname,
             'tags': tags,
             'timestamp': float(timestamp or time.time())
         })

From 74028f020100fc5e3ff4e065c4b9dfc08673d02d Mon Sep 17 00:00:00 2001
From: Remi Hakim <remi@datadoghq.com>
Date: Mon, 28 Apr 2014 16:30:32 -0400
Subject: [PATCH 23/33] Update jmx.yaml.example

---
 conf.d/jmx.yaml.example | 20 ++++++++++----------
 1 file changed, 10 insertions(+), 10 deletions(-)

diff --git a/conf.d/jmx.yaml.example b/conf.d/jmx.yaml.example
index ff1d56ccc3..2d793a0f6b 100644
--- a/conf.d/jmx.yaml.example
+++ b/conf.d/jmx.yaml.example
@@ -1,22 +1,22 @@
 init_config:
 
 instances:
-#    -   host: localhost
-#        port: 7199
-#        name: jmx_instance
-#        user: username
-#        password: password
-#        #java_bin_path: /path/to/java #Optional, should be set if the agent cannot find your java executable
-#        #trust_store_path: /path/to/trustStore.jks # Optional, should be set if ssl is enabled
-#        #trust_store_password: password
-#        tags:
+#    - host: localhost
+#      port: 7199
+#      name: jmx_instance
+#      user: username
+#      password: password
+#      #java_bin_path: /path/to/java #Optional, should be set if the agent cannot find your java executable
+#      #trust_store_path: /path/to/trustStore.jks # Optional, should be set if ssl is enabled
+#      #trust_store_password: password
+#      tags:
 #             env: stage
 #             newTag: test
 
 #
 #        # List of metrics to be collected by the integration
 #        # Read http://docs.datadoghq.com/integrations/java/ to learn how to customize it
-#        conf:
+#      conf:
 #            - include: 
 #               domain: my_domain
 #               bean: my_bean

From 887b49ae81400cd653d3398a85ae4b61e9365f9e Mon Sep 17 00:00:00 2001
From: graemej <graeme.johnson@jadedpixel.com>
Date: Tue, 29 Apr 2014 17:07:06 -0400
Subject: [PATCH 24/33] add get_ prefix and raise on missing url

---
 checks.d/mesos.py | 14 +++++++-------
 1 file changed, 7 insertions(+), 7 deletions(-)

diff --git a/checks.d/mesos.py b/checks.d/mesos.py
index ac4bb2c426..7449e629fe 100644
--- a/checks.d/mesos.py
+++ b/checks.d/mesos.py
@@ -9,7 +9,7 @@
 class Mesos(AgentCheck):
     def check(self, instance):
         if 'url' not in instance:
-            self.log.info("Skipping instance, no url found.")
+            raise Exception('Mesos instance missing "url" value.')
             return
 
         # Load values from the instance config
@@ -17,7 +17,7 @@ def check(self, instance):
         default_timeout = self.init_config.get('default_timeout', 5)
         timeout = float(instance.get('timeout', default_timeout))
 
-        response = self.master_roles(url, timeout)
+        response = self.get_master_roles(url, timeout)
         if response is not None:
             for role in response['roles']:
                 tags = ['mesos','role:' + role['name']]
@@ -28,12 +28,12 @@ def check(self, instance):
                     if attr in resources:
                         self.gauge('mesos.role.' + attr, resources[attr], tags=tags)
 
-        response = self.master_stats(url, timeout)
+        response = self.get_master_stats(url, timeout)
         if response is not None:
             for key in iter(response):
                 self.gauge('mesos.stats.' + key, response[key], tags=['mesos'])
 
-        response = self.master_state(url, timeout)
+        response = self.get_master_state(url, timeout)
         if response is not None:
             for attr in ['deactivated_slaves','failed_tasks','finished_tasks','killed_tasks','lost_tasks','staged_tasks','started_tasks']:
                 tags = ['mesos']
@@ -53,13 +53,13 @@ def check(self, instance):
                     if attr in resources:
                         self.gauge('mesos.state.slave.' + attr, resources[attr], tags=tags)
 
-    def master_roles(self, url, timeout):
+    def get_master_roles(self, url, timeout):
         return self.get_json(url + "/master/roles.json", timeout)
 
-    def master_stats(self, url, timeout):
+    def get_master_stats(self, url, timeout):
         return self.get_json(url + "/master/stats.json", timeout)
 
-    def master_state(self, url, timeout):
+    def get_master_state(self, url, timeout):
         return self.get_json(url + "/master/state.json", timeout)
 
     def get_json(self, url, timeout):

From aebea2f12e6e9d374f6ba16e57b7a16d7ed0942d Mon Sep 17 00:00:00 2001
From: graemej <graeme.johnson@jadedpixel.com>
Date: Tue, 29 Apr 2014 17:17:07 -0400
Subject: [PATCH 25/33] basic monitoring support for Mesosphere Marathon

---
 checks.d/marathon.py         | 95 ++++++++++++++++++++++++++++++++++++
 conf.d/marathon.yaml.example |  7 +++
 2 files changed, 102 insertions(+)
 create mode 100644 checks.d/marathon.py
 create mode 100644 conf.d/marathon.yaml.example

diff --git a/checks.d/marathon.py b/checks.d/marathon.py
new file mode 100644
index 0000000000..b7b531f054
--- /dev/null
+++ b/checks.d/marathon.py
@@ -0,0 +1,95 @@
+import time
+import requests
+
+from checks import AgentCheck
+from util import json, headers
+from hashlib import md5
+import urllib2
+
+class Marathon(AgentCheck):
+    def check(self, instance):
+        if 'url' not in instance:
+            raise Exception('Marathon instance missing "url" value.')
+            return
+
+        # Load values from the instance config
+        url = instance['url']
+        default_timeout = self.init_config.get('default_timeout', 5)
+        timeout = float(instance.get('timeout', default_timeout))
+
+        response = self.get_v2_apps(url, timeout)
+        if response is not None:
+            self.gauge('marathon.apps', len(response['apps']), tags=['marathon'])
+            for app in response['apps']:
+                tags = ['marathon', 'app_id:' + app['id'], 'version:' + app['version']]
+                for attr in ['taskRateLimit','instances','cpus','mem','tasksStaged','tasksRunning']:
+                    self.gauge('marathon.' + attr, app[attr], tags=tags)
+                versions_reply = self.get_v2_app_versions(url, app['id'], timeout)
+                if versions_reply is not None:
+                    self.gauge('marathon.versions', len(versions_reply['versions']), tags=tags)
+
+    def get_v2_apps(self, url, timeout):
+        # Use a hash of the URL as an aggregation key
+        aggregation_key = md5(url).hexdigest()
+
+        try:
+            response = requests.get(url + "/v2/apps", timeout=timeout)
+            apps = response.json()
+            return apps
+        except requests.exceptions.Timeout as e:
+            # If there's a timeout
+            self.timeout_event(url, timeout, aggregation_key)
+            return None
+
+        if r.status_code != 200:
+            self.status_code_event(url, r, aggregation_key)
+            return None
+
+    def get_v2_app_versions(self, url, app_id, timeout):
+        # Use a hash of the URL as an aggregation key
+        aggregation_key = md5(url).hexdigest()
+
+        try:
+            response = requests.get(url + "/v2/apps/" + app_id + "/versions", timeout=timeout)
+            apps = response.json()
+            return apps
+        except requests.exceptions.Timeout as e:
+            # If there's a timeout
+            self.timeout_event(url, timeout, aggregation_key)
+            return None
+
+        if r.status_code != 200:
+            self.status_code_event(url, r, aggregation_key)
+            return None
+
+    def timeout_event(self, url, timeout, aggregation_key):
+        self.event({
+            'timestamp': int(time.time()),
+            'event_type': 'http_check',
+            'msg_title': 'URL timeout',
+            'msg_text': '%s timed out after %s seconds.' % (url, timeout),
+            'aggregation_key': aggregation_key
+        })
+
+    def status_code_event(self, url, r, aggregation_key):
+        self.event({
+            'timestamp': int(time.time()),
+            'event_type': 'http_check',
+            'msg_title': 'Invalid reponse code for %s' % url,
+            'msg_text': '%s returned a status of %s' % (url, r.status_code),
+            'aggregation_key': aggregation_key
+        })
+
+if __name__ == '__main__':
+    check, instances = Marathon.from_yaml('/etc/dd-agent/conf.d/marathon.yaml')
+    for instance in instances:
+        print "\nRunning the check against url: %s" % (instance['url'])
+        check.check(instance)
+        if check.has_events():
+            print 'Events: %s' % (check.get_events())
+
+        i = 0
+        print 'Metrics:\n'
+        for metric in check.get_metrics():
+            print "  %d: %s" % (i, metric)
+            i += 1
\ No newline at end of file
diff --git a/conf.d/marathon.yaml.example b/conf.d/marathon.yaml.example
new file mode 100644
index 0000000000..b230b0b6cd
--- /dev/null
+++ b/conf.d/marathon.yaml.example
@@ -0,0 +1,7 @@
+init_config:
+# time to wait on a Marathon API request
+#    default_timeout: 5
+
+instances:
+# url: the API endpoint of your Marathon master
+#    - url: "https://server:port"
\ No newline at end of file

From 418435534ffdac7838db68087f822adac0facc55 Mon Sep 17 00:00:00 2001
From: graemej <graeme.johnson@jadedpixel.com>
Date: Wed, 30 Apr 2014 13:23:06 -0400
Subject: [PATCH 26/33] add instance tags

---
 checks.d/mesos.py | 12 +++++++-----
 1 file changed, 7 insertions(+), 5 deletions(-)

diff --git a/checks.d/mesos.py b/checks.d/mesos.py
index 7449e629fe..19b0eaf6f8 100644
--- a/checks.d/mesos.py
+++ b/checks.d/mesos.py
@@ -14,13 +14,14 @@ def check(self, instance):
 
         # Load values from the instance config
         url = instance['url']
+        instance_tags = instance.get('tags', [])
         default_timeout = self.init_config.get('default_timeout', 5)
         timeout = float(instance.get('timeout', default_timeout))
 
         response = self.get_master_roles(url, timeout)
         if response is not None:
             for role in response['roles']:
-                tags = ['mesos','role:' + role['name']]
+                tags = ['role:' + role['name']] + instance_tags
                 self.gauge('mesos.role.frameworks', len(role['frameworks']), tags=tags)
                 self.gauge('mesos.role.weight', role['weight'], tags=tags)
                 resources = role['resources']
@@ -30,24 +31,25 @@ def check(self, instance):
 
         response = self.get_master_stats(url, timeout)
         if response is not None:
+            tags = instance_tags
             for key in iter(response):
-                self.gauge('mesos.stats.' + key, response[key], tags=['mesos'])
+                self.gauge('mesos.stats.' + key, response[key], tags=tags)
 
         response = self.get_master_state(url, timeout)
         if response is not None:
+            tags = instance_tags
             for attr in ['deactivated_slaves','failed_tasks','finished_tasks','killed_tasks','lost_tasks','staged_tasks','started_tasks']:
-                tags = ['mesos']
                 self.gauge('mesos.state.' + attr, response[attr], tags=tags)
 
             for framework in response['frameworks']:
-                tags = ['mesos','framework:' + framework['id']]
+                tags = ['framework:' + framework['id']] + instance_tags
                 resources = framework['resources']
                 for attr in ['cpus','mem']:
                     if attr in resources:
                         self.gauge('mesos.state.framework.' + attr, resources[attr], tags=tags)
 
             for slave in response['slaves']:
-                tags = ['mesos','slave:' + slave['id']]
+                tags = ['mesos','slave:' + slave['id']] + instance_tags
                 resources = slave['resources']
                 for attr in ['cpus','mem','disk']:
                     if attr in resources:

From 87623ac676dc1115b721d89289ccbc905ffa1097 Mon Sep 17 00:00:00 2001
From: graemej <graeme.johnson@jadedpixel.com>
Date: Wed, 30 Apr 2014 13:31:50 -0400
Subject: [PATCH 27/33] add instance tags

---
 checks.d/marathon.py | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)

diff --git a/checks.d/marathon.py b/checks.d/marathon.py
index b7b531f054..340e5aa4e1 100644
--- a/checks.d/marathon.py
+++ b/checks.d/marathon.py
@@ -14,14 +14,15 @@ def check(self, instance):
 
         # Load values from the instance config
         url = instance['url']
+        instance_tags = instance.get('tags', [])
         default_timeout = self.init_config.get('default_timeout', 5)
         timeout = float(instance.get('timeout', default_timeout))
 
         response = self.get_v2_apps(url, timeout)
         if response is not None:
-            self.gauge('marathon.apps', len(response['apps']), tags=['marathon'])
+            self.gauge('marathon.apps', len(response['apps']), tags=instance_tags)
             for app in response['apps']:
-                tags = ['marathon', 'app_id:' + app['id'], 'version:' + app['version']]
+                tags = ['app_id:' + app['id'], 'version:' + app['version']] + instance_tags
                 for attr in ['taskRateLimit','instances','cpus','mem','tasksStaged','tasksRunning']:
                     self.gauge('marathon.' + attr, app[attr], tags=tags)
                 versions_reply = self.get_v2_app_versions(url, app['id'], timeout)

From 7b0dea036dcda5ccdcd3883b6adfbdaad86fd3dd Mon Sep 17 00:00:00 2001
From: Rudy <rudy@datadoghq.com>
Date: Thu, 1 May 2014 09:40:56 -0400
Subject: [PATCH 28/33] CI| Remove Unused Code making pylint fail

---
 checks/datadog.py | 14 --------------
 1 file changed, 14 deletions(-)

diff --git a/checks/datadog.py b/checks/datadog.py
index 46ec5efe90..fa5886a367 100644
--- a/checks/datadog.py
+++ b/checks/datadog.py
@@ -565,20 +565,6 @@ def check(self, agentConfig, move_end=True):
             return {}
 
 
-def testDogStream():
-    import logging
-
-    logger = logging.getLogger("ddagent.checks.datadog")
-    logger.setLevel(logging.DEBUG)
-    logger.addHandler(logging.StreamHandler())
-    dogstream = Dogstream(logger)
-
-    while True:
-        events = dogstream.check({'api_key':'my_apikey', 'dogstream_log': sys.argv[1]}, move_end=True)
-        for e in events:
-            print "Event:", e
-        time.sleep(5)
-
 def testddForwarder():
     import logging
 

From e331e03356545f1dc66d15e45448da7bd9448b76 Mon Sep 17 00:00:00 2001
From: Rudy <rudy@datadoghq.com>
Date: Thu, 1 May 2014 10:22:31 -0400
Subject: [PATCH 29/33] CI| pylint errors in graphite.py

---
 graphite.py | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)

diff --git a/graphite.py b/graphite.py
index 7066f1cedd..fb77ad4212 100644
--- a/graphite.py
+++ b/graphite.py
@@ -82,7 +82,8 @@ def _postMetric(self, name, host, device, datapoint):
 
         ts = datapoint[0]
         value = datapoint[1]
-        self.app.appendMetric("graphite", name, host, device, ts, value)        
+        if self.app is not None:
+            self.app.appendMetric("graphite", name, host, device, ts, value)
 
     def _processMetric(self, metric, datapoint):
         """Parse the metric name to fetch (host, metric, device) and
@@ -114,7 +115,8 @@ def _decode(self, data):
         self.stream.read_bytes(4, self._on_read_header)
 
 def start_graphite_listener(port):
-    echo_server = GraphiteServer()
+    from util import get_hostname
+    echo_server = GraphiteServer(None,get_hostname(None))
     echo_server.listen(port)
     IOLoop.instance().start()
 

From 7e2f50764b0b3e9ac885146611966c34c69b6e12 Mon Sep 17 00:00:00 2001
From: Rudy <rudy@datadoghq.com>
Date: Mon, 5 May 2014 12:07:57 -0400
Subject: [PATCH 30/33] Style| Forgot space after comma

---
 graphite.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/graphite.py b/graphite.py
index fb77ad4212..44187e0881 100644
--- a/graphite.py
+++ b/graphite.py
@@ -116,7 +116,7 @@ def _decode(self, data):
 
 def start_graphite_listener(port):
     from util import get_hostname
-    echo_server = GraphiteServer(None,get_hostname(None))
+    echo_server = GraphiteServer(None, get_hostname(None))
     echo_server.listen(port)
     IOLoop.instance().start()
 

From 924e46eb3bd2b6ffdc060a901c0f168069525c47 Mon Sep 17 00:00:00 2001
From: Arthur Neves <arthurnn@gmail.com>
Date: Mon, 5 May 2014 12:33:39 -0400
Subject: [PATCH 31/33] Add delete_misses/delete_hits to memcache

See https://github.com/memcached/memcached/blob/master/doc/protocol.txt,
there is delete_* stats we could use to report to datadog.
---
 checks.d/mcache.py | 27 ++++++++++++++++-----------
 1 file changed, 16 insertions(+), 11 deletions(-)

diff --git a/checks.d/mcache.py b/checks.d/mcache.py
index 4732212b22..57231e81c5 100644
--- a/checks.d/mcache.py
+++ b/checks.d/mcache.py
@@ -9,34 +9,37 @@
 # 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
+# delete_misses     64u      Number of deletions reqs for missing keys
+# delete_hits       64u      Number of deletion reqs resulting in
+#                            an item being removed.
 # 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()
@@ -77,6 +80,8 @@ class Memcache(AgentCheck):
         "cmd_flush",
         "get_hits",
         "get_misses",
+        "delete_misses",
+        "delete_hits",
         "evictions",
         "bytes_read",
         "bytes_written",

From 33d774f36bcbf507807272bd7497bd693f13ef60 Mon Sep 17 00:00:00 2001
From: Remi Hakim <remi@datadoghq.com>
Date: Mon, 5 May 2014 14:11:08 -0400
Subject: [PATCH 32/33] Fix memcache test since new metrics have been added

---
 tests/test_mcache.py | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/tests/test_mcache.py b/tests/test_mcache.py
index f68c1a40fd..311976c98e 100644
--- a/tests/test_mcache.py
+++ b/tests/test_mcache.py
@@ -46,8 +46,8 @@ def testMetrics(self):
         # Check that we got metrics from 3 hosts (aka all but the dummy host)
         self.assertEquals(len([t for t in r if t[0] == "memcache.total_items"]), 3, r)
 
-        # Check that we got 21 metrics for a specific host
-        self.assertEquals(len([t for t in r if t[3].get('tags') == ["instance:mythirdtag"]]), 21, r)
+        # Check that we got 23 metrics for a specific host
+        self.assertEquals(len([t for t in r if t[3].get('tags') == ["instance:mythirdtag"]]), 23, r)
 
     def testTagging(self):
         instance = {
@@ -64,7 +64,7 @@ def testTagging(self):
         r = self.c.get_metrics()
 
         # Check the tags
-        self.assertEquals(len([t for t in r if t[3].get('tags') == ["regular_old_tag"]]), 21, r)
+        self.assertEquals(len([t for t in r if t[3].get('tags') == ["regular_old_tag"]]), 23, r)
 
         conf = {
             'memcache_server': 'localhost',
@@ -80,7 +80,7 @@ def testTagging(self):
         r = self.c.get_metrics()
 
         # Check the tags
-        self.assertEquals(len([t for t in r if t[3].get('tags') == ["instance:localhost_11211"]]), 21, r)
+        self.assertEquals(len([t for t in r if t[3].get('tags') == ["instance:localhost_11211"]]), 23, r)
 
     def testDummyHost(self):
         new_conf = self.c.parse_agent_config({"memcache_instance_1": "dummy:11211:myothertag"})

From 05fe34d741e1a52f5427387fd63ed560c867038d Mon Sep 17 00:00:00 2001
From: Remi Hakim <remi@datadoghq.com>
Date: Mon, 5 May 2014 15:32:23 -0400
Subject: [PATCH 33/33] Make docker integration compatible with latest versions
 of docker

---
 checks.d/docker.py | 28 +++++++++++++++++++++++-----
 1 file changed, 23 insertions(+), 5 deletions(-)

diff --git a/checks.d/docker.py b/checks.d/docker.py
index 241d631bdc..22c6c058fa 100644
--- a/checks.d/docker.py
+++ b/checks.d/docker.py
@@ -13,7 +13,7 @@
 LXC_METRICS = [
     {
         "cgroup": "memory",
-        "file": "lxc/%s/memory.stat",
+        "file": "%s/%s/memory.stat",
         "metrics": {
             "active_anon": ("docker.mem.active_anon", "gauge"),
             "active_file": ("docker.mem.active_file", "gauge"),
@@ -47,7 +47,7 @@
     },
     {
         "cgroup": "cpuacct",
-        "file": "lxc/%s/cpuacct.stat",
+        "file": "%s/%s/cpuacct.stat",
         "metrics": {
             "user": ("docker.cpu.user", "gauge"),
             "system": ("docker.cpu.system", "gauge"),
@@ -103,12 +103,30 @@ def unix_open(self, req):
 class Docker(AgentCheck):
     def __init__(self, *args, **kwargs):
         super(Docker, self).__init__(*args, **kwargs)
-        urllib2.install_opener(urllib2.build_opener(UnixSocketHandler()))
         self._mounpoints = {}
+        self.cgroup_path_prefix = None # Depending on the version 
         for metric in LXC_METRICS:
             self._mounpoints[metric["cgroup"]] = self._find_cgroup(metric["cgroup"])
+        self._path_prefix = None
+
+    @property
+    def path_prefix(self):
+        if self._path_prefix is None:
+            metric = LXC_METRICS[0]
+            mountpoint = self._mounpoints[metric["cgroup"]]
+            stat_file_lxc = os.path.join(mountpoint, "lxc")
+            stat_file_docker = os.path.join(mountpoint, "docker")
+
+            if os.path.exists(stat_file_lxc):
+                self._path_prefix = "lxc"
+            elif os.path.exists(stat_file_docker):
+                self._path_prefix = "docker"
+            else:
+                raise Exception("Cannot find Docker cgroup file. If you are using Docker 0.9 or 0.10, it is a known bug in Docker fixed in Docker 0.10.1")
+        return self._path_prefix
 
     def check(self, instance):
+        urllib2.install_opener(urllib2.build_opener(UnixSocketHandler())) # We need to reinstall the opener every time as it gets uninstalled
         tags = instance.get("tags") or []
         containers = self._get_containers(instance)
         if not containers:
@@ -143,7 +161,7 @@ def check(self, instance):
                     getattr(self, metric_type)(dd_key, int(container[key]), tags=container_tags)
             for metric in LXC_METRICS:
                 mountpoint = self._mounpoints[metric["cgroup"]]
-                stat_file = os.path.join(mountpoint, metric["file"] % container["Id"])
+                stat_file = os.path.join(mountpoint, metric["file"] % (self.path_prefix, container["Id"]))
                 stats = self._parse_cgroup_file(stat_file)
                 for key, (dd_key, metric_type) in metric["metrics"].items():
                     if key in stats:
@@ -213,7 +231,7 @@ def _parse_cgroup_file(self, file_):
             try:
                 fp = open(file_)
             except IOError:
-                raise IOError("Can't open %s. If you are using Docker 0.9.0 or higher, the Datadog agent is not yet compatible with these versions. Please get in touch with Datadog Support for more information" % file_)
+                raise IOError("Can't open %s. If you are using Docker 0.9 or 0.10, it is a known bug in Docker fixed in Docker 0.10.1" % file_)
             return dict(map(lambda x: x.split(), fp.read().splitlines()))
 
         finally: