Merge branch 'master' into quentin/windows-omnibus-5.0

* master: (53 commits)
  [nginx] Update example config
  [service_discovery] Add a Zookeeper service discovery implementation.
  [aggregator] if sample rate is bad, fix it but still parse tags. (#3073)
  [yarn] whitelist authorized application_tags
  Alex poe/update jmx with refresh beans (#3068)
  [config] Fix `_is_affirmative` when passed argument is `None` (#3063)
  Send all configured tags with process checks. (#2976)
  fix flake8 errors
  [flare] ignore whitespace before proxy credentials
  [core] add a switch to disable profiling, but still use developer mode (#2898)
  [tests] allow tests to use the additional_checksd parameter (#3056)
  [service_discovery][jmx] trying to pick-up JMX changes with SD. (#3010)
  [install_script] Make `dd-agent` group of `datadog.conf` (#3036)
  [postgres] Allow disable postgresql.database_size (#3035)
  [core] Fixes IndexError for process lookup (#3043)
  remove warning message leaking password strings (#3053)
  trap psutil.NoSuchProcess exception (#3052)
  Fix grammar and casing in exception text (#3050)
  allow override of kubelet host with KUBERNETES_KUBELET_HOST env var
  [service discovery] properly handle config reload for removed containers
  ...

Author: degemer

.gitignore

@@ -36,3 +36,6 @@ embedded/*
 dump.rdb
 tests/core/fixtures/flare/dd*
 .python-version
+.ropeproject
+.bundle
+tags

.rubocop.yml

@@ -23,3 +23,6 @@ Style/Documentation:
 # Configuration parameters: Methods.
 Style/SingleLineBlockParams:
   Enabled: false
+
+BlockLength:
+  Max: 110

CHANGELOG.md

@@ -1,6 +1,33 @@
 Changes
 =======
 
+# 5.10.1 / 11-21-2016
+**Linux, Windows, Docker and Source Install**
+
+### Details
+https://github.com/DataDog/dd-agent/compare/5.10.0...5.10.1
+
+### Updated Integrations
+
+* RiakCS
+* Mongo
+
+### Changes
+* [IMPROVEMENT] Core/Forwarder: stop flushing after 10s. See [#3018][].
+* [IMPROVEMENT] Core: isolate system checks. See [#3011][].
+* [IMPROVEMENT] RiakCS: support Riak CS 2.1+ stats format. See [#2920][]. (Thanks [@millerdev][])
+* [IMPROVEMENT] Status: Silence requests exception. See [#3023][].
+* [IMPROVEMENT] SpooledTemporaryFile for subprocess output. See [#3002][].
+
+* [BUGFIX] Core: fix unintended subprocess_output empty output errors. See [#3024][].
+* [BUGFIX] Core/Multiple Checks: Only set `psutil.PROCFS_PATH` once in the collector. See [#3013][].
+* [BUGFIX] Core: use proxy for API key status check in info page. See [#3012][]. (Thanks [@2rs2ts][])
+* [BUGFIX] Mongo: use db.current_op instead of manually querying. See [#3016][] (Thanks [@ebroder][])
+* [BUGFIX] Mongo: use `currentOp` for monodb 3.1+. See [#3015][] (Thanks [@lattwood][])
+
+* [DEPRECATE] Process: `procfs_path` is now deprecated, should be set in `datadog.conf`. See [#3013][].
+
+
 # 5.10.0 / 11-09-2016
 **Linux, Windows, Docker and Source Install**
 
@@ -3609,6 +3636,7 @@ https://github.com/DataDog/dd-agent/compare/2.2.9...2.2.10
 [#2908]: https://github.com/DataDog/dd-agent/issues/2908
 [#2910]: https://github.com/DataDog/dd-agent/issues/2910
 [#2915]: https://github.com/DataDog/dd-agent/issues/2915
+[#2920]: https://github.com/DataDog/dd-agent/issues/2920
 [#2921]: https://github.com/DataDog/dd-agent/issues/2921
 [#2926]: https://github.com/DataDog/dd-agent/issues/2926
 [#2928]: https://github.com/DataDog/dd-agent/issues/2928
@@ -3644,8 +3672,18 @@ https://github.com/DataDog/dd-agent/compare/2.2.9...2.2.10
 [#2984]: https://github.com/DataDog/dd-agent/issues/2984
 [#2989]: https://github.com/DataDog/dd-agent/issues/2989
 [#2991]: https://github.com/DataDog/dd-agent/issues/2991
+[#3002]: https://github.com/DataDog/dd-agent/issues/3002
 [#3006]: https://github.com/DataDog/dd-agent/issues/3006
+[#3011]: https://github.com/DataDog/dd-agent/issues/3011
+[#3012]: https://github.com/DataDog/dd-agent/issues/3012
+[#3013]: https://github.com/DataDog/dd-agent/issues/3013
+[#3015]: https://github.com/DataDog/dd-agent/issues/3015
+[#3016]: https://github.com/DataDog/dd-agent/issues/3016
+[#3018]: https://github.com/DataDog/dd-agent/issues/3018
+[#3023]: https://github.com/DataDog/dd-agent/issues/3023
+[#3024]: https://github.com/DataDog/dd-agent/issues/3024
 [#3399]: https://github.com/DataDog/dd-agent/issues/3399
+[@2rs2ts]: https://github.com/2rs2ts
 [@AirbornePorcine]: https://github.com/AirbornePorcine
 [@AntoCard]: https://github.com/AntoCard
 [@CaptTofu]: https://github.com/CaptTofu
@@ -3736,6 +3774,7 @@ https://github.com/DataDog/dd-agent/compare/2.2.9...2.2.10
 [@jslatts]: https://github.com/jslatts
 [@jzoldak]: https://github.com/jzoldak
 [@kzw]: https://github.com/kzw
+[@lattwood]: https://github.com/lattwood
 [@leifwalsh]: https://github.com/leifwalsh
 [@leucos]: https://github.com/leucos
 [@loris]: https://github.com/loris
@@ -3751,6 +3790,7 @@ https://github.com/DataDog/dd-agent/compare/2.2.9...2.2.10
 [@micktwomey]: https://github.com/micktwomey
 [@mike-lerch]: https://github.com/mike-lerch
 [@mikekap]: https://github.com/mikekap
+[@millerdev]: https://github.com/millerdev
 [@mms-gianni]: https://github.com/mms-gianni
 [@mooney6023]: https://github.com/mooney6023
 [@morskoyzmey]: https://github.com/morskoyzmey

agent.py

@@ -19,6 +19,8 @@
 import signal
 import sys
 import time
+import supervisor.xmlrpc
+import xmlrpclib
 from copy import copy
 
 # For pickle & PID files, see issue 293
@@ -29,13 +31,19 @@
 from checks.collector import Collector
 from config import (
     get_config,
+    get_jmx_pipe_path,
     get_parsed_args,
     get_system_stats,
     load_check_directory,
-    load_check
+    load_check,
+    generate_jmx_configs,
+    _is_affirmative,
+    SD_PIPE_NAME
+
 )
 from daemon import AgentSupervisor, Daemon
 from emitter import http_emitter
+from utils.platform import Platform
 
 # utils
 from utils.cloud_metadata import EC2
@@ -51,11 +59,18 @@
 from utils.watchdog import new_watchdog
 
 # Constants
+from jmxfetch import JMX_CHECKS
 PID_NAME = "dd-agent"
 PID_DIR = None
 WATCHDOG_MULTIPLIER = 10
 RESTART_INTERVAL = 4 * 24 * 60 * 60  # Defaults to 4 days
 
+JMX_SUPERVISOR_ENTRY = 'datadog-agent:jmxfetch'
+JMX_GRACE_SECS = 2
+SERVICE_DISCOVERY_PREFIX = 'SD-'
+SD_CONFIG_SEP = "#### SERVICE-DISCOVERY ####\n"
+
+DEFAULT_SUPERVISOR_SOCKET = '/opt/datadog-agent/run/datadog-supervisor.sock'
 DEFAULT_COLLECTOR_PROFILE_INTERVAL = 20
 
 # Globals
@@ -80,6 +95,9 @@ def __init__(self, pidfile, autorestart, start_event=True, in_developer_mode=Fal
         # this flag can be set to True, False, or a list of checks (for partial reload)
         self.reload_configs_flag = False
         self.sd_backend = None
+        self.supervisor_proxy = None
+        self.sd_pipe = None
+
 
     def _handle_sigterm(self, signum, frame):
         """Handles SIGTERM and SIGINT, which gracefully stops the agent."""
@@ -105,6 +123,7 @@ def reload_configs(self, checks_to_reload=set()):
            Can also reload only an explicit set of checks."""
         log.info("Attempting a configuration reload...")
         hostname = get_hostname(self._agentConfig)
+        jmx_sd_configs = None
 
         # if no check was given, reload them all
         if not checks_to_reload:
@@ -114,13 +133,23 @@ def reload_configs(self, checks_to_reload=set()):
                 check.stop()
 
             self._checksd = load_check_directory(self._agentConfig, hostname)
+            if self._jmx_service_discovery_enabled:
+                jmx_sd_configs = generate_jmx_configs(self._agentConfig, hostname)
         else:
             new_checksd = copy(self._checksd)
 
-            self.refresh_specific_checks(hostname, new_checksd, checks_to_reload)
+            jmx_checks = [check for check in checks_to_reload if check in JMX_CHECKS]
+            py_checks = set(checks_to_reload) - set(jmx_checks)
+            self.refresh_specific_checks(hostname, new_checksd, py_checks)
+            if self._jmx_service_discovery_enabled:
+                jmx_sd_configs = generate_jmx_configs(self._agentConfig, hostname, jmx_checks)
+
             # once the reload is done, replace existing checks with the new ones
             self._checksd = new_checksd
 
+        if jmx_sd_configs:
+            self._submit_jmx_service_discovery(jmx_sd_configs)
+
         # Logging
         num_checks = len(self._checksd['initialized_checks'])
         if num_checks > 0:
@@ -218,9 +247,32 @@ def run(self, config=None):
         if self._agentConfig.get('service_discovery'):
             self.sd_backend = get_sd_backend(self._agentConfig)
 
+        if _is_affirmative(self._agentConfig.get('sd_jmx_enable')):
+            pipe_path = get_jmx_pipe_path()
+            if Platform.is_windows():
+                pipe_name = pipe_path.format(pipename=SD_PIPE_NAME)
+            else:
+                pipe_name = os.path.join(pipe_path, SD_PIPE_NAME)
+
+            if os.access(pipe_path, os.W_OK):
+                if not os.path.exists(pipe_name):
+                    os.mkfifo(pipe_name)
+                self.sd_pipe = os.open(pipe_name, os.O_RDWR) # RW to avoid blocking (will only W)
+
+                # Initialize Supervisor proxy
+                self.supervisor_proxy = self._get_supervisor_socket(self._agentConfig)
+            else:
+                log.debug('Unable to create pipe in temporary directory. JMX service discovery disabled.')
+
         # Load the checks.d checks
         self._checksd = load_check_directory(self._agentConfig, hostname)
 
+        # Load JMX configs if available
+        if self._jmx_service_discovery_enabled:
+            jmx_sd_configs = generate_jmx_configs(self._agentConfig, hostname)
+            if jmx_sd_configs:
+                self._submit_jmx_service_discovery(jmx_sd_configs)
+
         # Initialize the Collector
         self.collector = Collector(self._agentConfig, emitters, systemStats, hostname)
 
@@ -241,13 +293,15 @@ def run(self, config=None):
         self.restart_interval = int(self._agentConfig.get('restart_interval', RESTART_INTERVAL))
         self.agent_start = time.time()
 
+        self.allow_profiling = self._agentConfig.get('allow_profiling', True)
+
         profiled = False
         collector_profiled_runs = 0
 
         # Run the main loop.
         while self.run_forever:
             # Setup profiling if necessary
-            if self.in_developer_mode and not profiled:
+            if self.allow_profiling and self.in_developer_mode and not profiled:
                 try:
                     profiler = AgentProfiler()
                     profiler.enable_profiling()
@@ -344,6 +398,52 @@ def _set_agent_config_hostname(self, agentConfig):
                 log.info('Not running on EC2, using hostname to identify this server')
         return agentConfig
 
+    def _get_supervisor_socket(self, agentConfig):
+        if Platform.is_windows():
+            return None
+
+        sockfile = agentConfig.get('supervisor_socket', DEFAULT_SUPERVISOR_SOCKET)
+        supervisor_proxy = xmlrpclib.ServerProxy(
+            'http://127.0.0.1',
+            transport=supervisor.xmlrpc.SupervisorTransport(
+                None, None, serverurl="unix://{socket}".format(socket=sockfile))
+        )
+
+        return supervisor_proxy
+
+    @property
+    def _jmx_service_discovery_enabled(self):
+        return self.sd_pipe is not None
+
+    def _submit_jmx_service_discovery(self, jmx_sd_configs):
+
+        if not jmx_sd_configs or not self.sd_pipe:
+            return
+
+        if self.supervisor_proxy is not None:
+            jmx_state = self.supervisor_proxy.supervisor.getProcessInfo(JMX_SUPERVISOR_ENTRY)
+            log.debug("Current JMX check state: %s", jmx_state['statename'])
+            # restart jmx if stopped
+            if jmx_state['statename'] in ['STOPPED', 'EXITED', 'FATAL'] and self._agentConfig.get('sd_jmx_enable'):
+                self.supervisor_proxy.supervisor.startProcess(JMX_SUPERVISOR_ENTRY)
+                time.sleep(JMX_GRACE_SECS)
+        else:
+            log.debug("Unable to automatically start jmxfetch on Windows via supervisor.")
+
+        buffer = ""
+        for name, yaml in jmx_sd_configs.iteritems():
+            try:
+                buffer += SD_CONFIG_SEP
+                buffer += "# {}\n".format(name)
+                buffer += yaml
+            except Exception as e:
+                log.exception("unable to submit YAML via RPC: %s", e)
+            else:
+                log.info("JMX SD Config via named pip %s successfully.", name)
+
+        if buffer:
+            os.write(self.sd_pipe, buffer)
+
     def _should_restart(self):
         if time.time() - self.agent_start > self.restart_interval:
             return True

aggregator.py

@@ -492,13 +492,14 @@ def parse_metric_packet(self, packet):
                     # Parse the sample rate
                     if m[0] == '@':
                         sample_rate = float(m[1:])
-                        assert 0 <= sample_rate <= 1
+                        # in case it's in a bad state
+                        sample_rate = 1 if sample_rate < 0 or sample_rate > 1 else sample_rate
                     elif m[0] == '#':
                         tags = tuple(sorted(m[1:].split(',')))
-            except (IndexError, AssertionError):
+            except IndexError:
                 log.warning(u'Incorrect metric metadata: metric_name:%s, metadata:%s',
                             name, u' '.join(value_and_metadata[2:]))
-                sample_rate = 1  # In case it's in a bad state
+
             parsed_packets.append((name, value, metric_type, tags, sample_rate))
 
         return parsed_packets