def reload_configs(self, checks_to_reload=set()):
"""Reload the agent configuration and checksd configurations.
Can also reload only an explicit set of checks."""
log.info("Attempting a configuration reload...")
hostname = get_hostname(self._agentConfig)
# if no check was given, reload them all
if not checks_to_reload:
log.debug("No check list was passed, reloading every check")
# stop checks
for check in self._checksd.get('initialized_checks', []):
check.stop()
self._checksd = load_check_directory(self._agentConfig, hostname)
else:
new_checksd = copy(self._checksd)
self.refresh_specific_checks(hostname, new_checksd, checks_to_reload)
# once the reload is done, replace existing checks with the new ones
self._checksd = new_checksd
# Logging
num_checks = len(self._checksd['initialized_checks'])
if num_checks > 0:
opt_msg = " (refreshed %s checks)" % len(checks_to_reload) if checks_to_reload else ''
msg = "Check reload was successful. Running {num_checks} checks{opt_msg}.".format(
num_checks=num_checks, opt_msg=opt_msg)
log.info(msg)
else:
log.info("No checksd configs found")
def testConfigETC3rdParty(self, *args):
copyfile("%s/valid_conf.yaml" % FIXTURE_PATH, "%s/test_check.yaml" % TEMP_ETC_CONF_DIR)
copyfile("%s/valid_check_2.py" % FIXTURE_PATH, "%s/test_check/check.py" % TEMP_SDK_INTEGRATIONS_CHECKS_DIR)
copyfile("%s/valid_check_1.py" % FIXTURE_PATH, "%s/test_check.py" % TEMP_ETC_CHECKS_DIR)
checks = load_check_directory({"additional_checksd": TEMP_ETC_CHECKS_DIR}, "foo")
self.assertEquals(1, len(checks["initialized_checks"]))
self.assertEquals("valid_check_1", checks["initialized_checks"][0].check(None))
def testConfigCheckImportError(self, *args):
copyfile('%s/valid_conf.yaml' % FIXTURE_PATH,
'%s/test_check.yaml' % TEMP_ETC_CONF_DIR)
copyfile('%s/invalid_check_2.py' % FIXTURE_PATH,
'%s/test_check.py' % TEMP_AGENT_CHECK_DIR)
checks = load_check_directory({"additional_checksd": TEMP_ETC_CHECKS_DIR}, "foo")
self.assertEquals(1, len(checks['init_failed_checks']))
def testConfigDefault(self, *args):
copyfile('%s/valid_conf.yaml' % FIXTURE_PATH,
'%s/test_check.yaml.default' % TEMP_ETC_CONF_DIR)
copyfile('%s/valid_check_1.py' % FIXTURE_PATH,
'%s/test_check.py' % TEMP_ETC_CHECKS_DIR)
checks = load_check_directory({"additional_checksd": TEMP_ETC_CHECKS_DIR}, "foo")
self.assertEquals(1, len(checks['initialized_checks']))
def testConfigInheritedCheck(self, *args):
copyfile('%s/valid_conf.yaml' % FIXTURE_PATH,
'%s/test_check.yaml' % TEMP_ETC_CONF_DIR)
copyfile('%s/valid_sub_check.py' % FIXTURE_PATH,
'%s/test_check.py' % TEMP_ETC_CHECKS_DIR)
checks = load_check_directory({"additional_checksd": TEMP_ETC_CHECKS_DIR}, "foo")
self.assertEquals(1, len(checks['initialized_checks']))
self.assertEquals('valid_check_1', checks['initialized_checks'][0].check(None))
def testConfigCustomOverDefault(self, *args):
copyfile("%s/valid_conf.yaml" % FIXTURE_PATH, "%s/test_check.yaml.default" % TEMP_ETC_CONF_DIR)
# a 2nd valid conf file, slightly different so that we can test which one has been picked up
# (with 2 instances for instance)
copyfile("%s/valid_conf_2.yaml" % FIXTURE_PATH, "%s/test_check.yaml" % TEMP_ETC_CONF_DIR)
copyfile("%s/valid_check_1.py" % FIXTURE_PATH, "%s/test_check.py" % TEMP_ETC_CHECKS_DIR)
checks = load_check_directory({"additional_checksd": TEMP_ETC_CHECKS_DIR}, "foo")
self.assertEquals(1, len(checks["initialized_checks"]))
self.assertEquals(2, checks["initialized_checks"][0].instance_count()) # check that we picked the right conf
def testConfig3rdPartyAgent(self, *args):
copyfile('%s/valid_conf.yaml' % FIXTURE_PATH,
'%s/test_check.yaml' % TEMP_ETC_CONF_DIR)
copyfile('%s/valid_check_2.py' % FIXTURE_PATH,
'%s/test_check.py' % TEMP_AGENT_CHECK_DIR)
copyfile('%s/valid_check_1.py' % FIXTURE_PATH,
'%s/test_check/check.py' % TEMP_SDK_INTEGRATIONS_CHECKS_DIR)
checks = load_check_directory({"additional_checksd": TEMP_ETC_CHECKS_DIR}, "foo")
self.assertEquals(1, len(checks['initialized_checks']))
self.assertEquals('valid_check_1', checks['initialized_checks'][0].check(None))
def run(self):
emitters = self.get_emitters()
systemStats = get_system_stats()
collector = Collector(self.config, emitters, systemStats)
# Load the checks.d checks
checksd = load_check_directory(self.config)
# Main agent loop will run until interrupted
while self.running:
collector.run(checksd=checksd)
time.sleep(self.config['check_freq'])
def sd_configcheck(agentConfig):
if agentConfig.get('service_discovery', False):
# set the TRACE_CONFIG flag to True to make load_check_directory return
# the source of config objects.
# Then call load_check_directory here and pass the result to get_sd_configcheck
# to avoid circular imports
agentConfig[TRACE_CONFIG] = True
configs = {
# check_name: (config_source, config)
}
print("\nLoading check configurations...\n\n")
configs = load_check_directory(agentConfig, get_hostname(agentConfig))
get_sd_configcheck(agentConfig, configs)
def run(self):
log.debug("Windows Service - Starting collector")
emitters = self.get_emitters()
systemStats = get_system_stats()
self.collector = Collector(self.config, emitters, systemStats)
# Load the checks.d checks
checksd = load_check_directory(self.config)
# Main agent loop will run until interrupted
while self.running:
self.collector.run(checksd=checksd, start_event=self.start_event)
time.sleep(self.config['check_freq'])
def run(self):
emitters = self.get_emitters();
chk = checks(self.config, emitters)
# Load the checks.d checks
checksd = load_check_directory(self.config)
# Main agent loop will run until interrupted
systemStats = get_system_stats()
chk.doChecks(True, systemStats, checksd)
while self.running:
chk.doChecks(checksd=checksd)
time.sleep(self.config['check_freq'])
def run(self):
from config import initialize_logging; initialize_logging('windows_collector')
log.debug("Windows Service - Starting collector")
emitters = self.get_emitters()
systemStats = get_system_stats()
self.collector = Collector(self.config, emitters, systemStats, self.hostname)
# Load the checks.d checks
checksd = load_check_directory(self.config, self.hostname)
# Main agent loop will run until interrupted
while self.running:
self.collector.run(checksd=checksd, start_event=self.start_event)
time.sleep(self.config['check_freq'])
def reload_configs(self):
"""Reloads the agent configuration and checksd configurations."""
log.info("Attempting a configuration reload...")
# Reload checksd configs
hostname = get_hostname(self._agentConfig)
self._checksd = load_check_directory(self._agentConfig, hostname)
# Logging
num_checks = len(self._checksd["initialized_checks"])
if num_checks > 0:
log.info("Successfully reloaded {num_checks} checks".format(num_checks=num_checks))
else:
log.info("No checksd configs found")
def run(self, agentConfig=None, run_forever=True):
"""Main loop of the collector"""
agentLogger = logging.getLogger('agent')
systemStats = get_system_stats()
agentLogger.debug('System Properties: ' + str(systemStats))
if agentConfig is None:
agentConfig = get_config()
# Load the checks.d checks
checksd = load_check_directory(agentConfig)
# Try to fetch instance Id from EC2 if not hostname has been set
# in the config file.
# DEPRECATED
if agentConfig.get('hostname') is None and agentConfig.get('use_ec2_instance_id'):
instanceId = EC2.get_instance_id()
if instanceId is not None:
agentLogger.info("Running on EC2, instanceId: %s" % instanceId)
agentConfig['hostname'] = instanceId
else:
agentLogger.info('Not running on EC2, using hostname to identify this server')
emitters = [http_emitter]
for emitter_spec in [s.strip() for s in agentConfig.get('custom_emitters', '').split(',')]:
if len(emitter_spec) == 0: continue
emitters.append(modules.load(emitter_spec, 'emitter'))
check_freq = int(agentConfig['check_freq'])
# Checks instance
c = checks(agentConfig, emitters)
# Watchdog
watchdog = None
if agentConfig.get("watchdog", True):
watchdog = Watchdog(check_freq * WATCHDOG_MULTIPLIER)
watchdog.reset()
# Run checks once, to get once-in-a-run data
c.doChecks(True, systemStats, checksd)
# Main loop
while run_forever:
if watchdog is not None:
watchdog.reset()
time.sleep(check_freq)
c.doChecks(checksd=checksd)
def run(self):
from config import initialize_logging
initialize_logging('windows_collector')
log.debug("Windows Service - Starting collector")
set_win32_requests_ca_bundle_path()
emitters = self.get_emitters()
systemStats = get_system_stats()
self.collector = Collector(self.config, emitters, systemStats, self.hostname)
in_developer_mode = self.config.get('developer_mode')
# In developer mode, the number of runs to be included in a single collector profile
collector_profile_interval = self.config.get('collector_profile_interval',
DEFAULT_COLLECTOR_PROFILE_INTERVAL)
profiled = False
collector_profiled_runs = 0
# Load the checks.d checks
checksd = load_check_directory(self.config, self.hostname)
# Main agent loop will run until interrupted
while self.running:
if self._heartbeat:
self._heartbeat.send(0)
if in_developer_mode and not profiled:
try:
profiler = AgentProfiler()
profiler.enable_profiling()
profiled = True
except Exception as e:
log.warn("Cannot enable profiler: %s" % str(e))
self.collector.run(checksd=checksd)
if profiled:
if collector_profiled_runs >= collector_profile_interval:
try:
profiler.disable_profiling()
profiled = False
collector_profiled_runs = 0
except Exception as e:
log.warn("Cannot disable profiler: %s" % str(e))
else:
collector_profiled_runs += 1
time.sleep(self.config['check_freq'])
def run(self, config=None):
"""Main loop of the collector"""
# Gracefully exit on sigterm.
signal.signal(signal.SIGTERM, self._handle_sigterm)
# Save the agent start-up stats.
CollectorStatus().persist()
# Intialize the collector.
if not config:
config = get_config(parse_args=True)
agentConfig = self._set_agent_config_hostname(config)
systemStats = get_system_stats()
emitters = self._get_emitters(agentConfig)
self.collector = Collector(agentConfig, emitters, systemStats)
# Load the checks.d checks
checksd = load_check_directory(agentConfig)
# Configure the watchdog.
check_frequency = int(agentConfig['check_freq'])
watchdog = self._get_watchdog(check_frequency, agentConfig)
# Run the main loop.
while self.run_forever:
# Do the work.
self.collector.run(checksd=checksd)
# Only plan for the next loop if we will continue,
# otherwise just exit quickly.
if self.run_forever:
if watchdog:
watchdog.reset()
time.sleep(check_frequency)
# Now clean-up.
try:
CollectorStatus.remove_latest_status()
except:
pass
# Explicitly kill the process, because it might be running
# as a daemon.
log.info("Exiting. Bye bye.")
sys.exit(0)
def testConfigDeprecatedNagiosConfig(self, *args):
copyfile('%s/valid_check_1.py' % FIXTURE_PATH,
'%s/nagios.py' % TEMP_ETC_CHECKS_DIR)
checks = load_check_directory({"nagios_perf_cfg": None, "additional_checksd": TEMP_ETC_CHECKS_DIR}, "foo")
self.assertEquals(1, len(checks['initialized_checks']))
self.assertEquals('valid_check_1', checks['initialized_checks'][0].check(None))
def run(self, config=None):
"""Main loop of the collector"""
# Gracefully exit on sigterm.
signal.signal(signal.SIGTERM, self._handle_sigterm)
if not Platform.is_windows():
# A SIGUSR1 signals an exit with an autorestart
signal.signal(signal.SIGUSR1, self._handle_sigusr1)
# Handle Keyboard Interrupt
signal.signal(signal.SIGINT, self._handle_sigterm)
# A SIGHUP signals a configuration reload
signal.signal(signal.SIGHUP, self._handle_sighup)
else:
sdk_integrations = get_sdk_integration_paths()
for name, path in sdk_integrations.iteritems():
lib_path = os.path.join(path, 'lib')
if os.path.exists(lib_path):
sys.path.append(lib_path)
# Save the agent start-up stats.
CollectorStatus().persist()
# Intialize the collector.
if not config:
config = get_config(parse_args=True)
self._agentConfig = self._set_agent_config_hostname(config)
hostname = get_hostname(self._agentConfig)
systemStats = get_system_stats(proc_path=self._agentConfig.get(
'procfs_path', '/proc').rstrip('/'))
emitters = self._get_emitters()
# Initialize service discovery
if self._agentConfig.get('service_discovery'):
self.sd_backend = get_sd_backend(self._agentConfig)
if self.sd_backend and _is_affirmative(
self._agentConfig.get('sd_jmx_enable', False)):
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:
self.sd_pipe_jmx_configs(hostname)
# Initialize the Collector
self.collector = Collector(self._agentConfig, emitters, systemStats,
hostname)
# In developer mode, the number of runs to be included in a single collector profile
try:
self.collector_profile_interval = int(
self._agentConfig.get('collector_profile_interval',
DEFAULT_COLLECTOR_PROFILE_INTERVAL))
except ValueError:
log.warn('collector_profile_interval is invalid. '
'Using default value instead (%s).' %
DEFAULT_COLLECTOR_PROFILE_INTERVAL)
self.collector_profile_interval = DEFAULT_COLLECTOR_PROFILE_INTERVAL
# Configure the watchdog.
self.check_frequency = int(self._agentConfig['check_freq'])
watchdog = self._get_watchdog(self.check_frequency)
# Initialize the auto-restarter
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.allow_profiling and self.in_developer_mode and not profiled:
try:
profiler = AgentProfiler()
profiler.enable_profiling()
profiled = True
except Exception as e:
log.warn("Cannot enable profiler: %s" % str(e))
if self.reload_configs_flag:
if isinstance(self.reload_configs_flag, set):
self.reload_configs(
checks_to_reload=self.reload_configs_flag)
else:
self.reload_configs()
# JMXFetch restarts should prompt re-piping *all* JMX configs
if self._jmx_service_discovery_enabled and \
(not self.reload_configs_flag or isinstance(self.reload_configs_flag, set)):
try:
jmx_launch = JMXFetch._get_jmx_launchtime()
if self.last_jmx_piped and self.last_jmx_piped < jmx_launch:
self.sd_pipe_jmx_configs(hostname)
except Exception as e:
log.debug("could not stat JMX lunch file: %s", e)
# Do the work. Pass `configs_reloaded` to let the collector know if it needs to
# look for the AgentMetrics check and pop it out.
_, continue_immediately = self.collector.run(
checksd=self._checksd,
start_event=self.start_event,
configs_reloaded=True if self.reload_configs_flag else False)
self.reload_configs_flag = False
# Look for change in the config template store.
# The self.sd_backend.reload_check_configs flag is set
# to True if a config reload is needed.
if self._agentConfig.get('service_discovery') and self.sd_backend and \
not self.sd_backend.reload_check_configs:
try:
self.sd_backend.reload_check_configs = get_config_store(
self._agentConfig).crawl_config_template()
except Exception as e:
log.warn(
'Something went wrong while looking for config template changes: %s'
% str(e))
# Check if we should run service discovery
# The `reload_check_configs` flag can be set through the docker_daemon check or
# using ConfigStore.crawl_config_template
if self._agentConfig.get('service_discovery') and self.sd_backend and \
self.sd_backend.reload_check_configs:
self.reload_configs_flag = self.sd_backend.reload_check_configs
self.sd_backend.reload_check_configs = False
if profiled:
if collector_profiled_runs >= self.collector_profile_interval:
try:
profiler.disable_profiling()
profiled = False
collector_profiled_runs = 0
except Exception as e:
log.warn("Cannot disable profiler: %s" % str(e))
# Check if we should restart.
if self.autorestart and self._should_restart():
self._do_restart()
# Only plan for next loop if we will continue, otherwise exit quickly.
if self.run_forever:
if watchdog:
watchdog.reset()
if profiled:
collector_profiled_runs += 1
if not continue_immediately:
log.debug("Sleeping for {0} seconds".format(
self.check_frequency))
time.sleep(self.check_frequency)
else:
log.debug("Continuing immediately")
# Now clean-up.
try:
CollectorStatus.remove_latest_status()
except Exception:
pass
# Explicitly kill the process, because it might be running as a daemon.
log.info("Exiting. Bye bye.")
sys.exit(0)
def main():
options, args = get_parsed_args()
agentConfig = get_config(options=options)
autorestart = agentConfig.get('autorestart', False)
hostname = get_hostname(agentConfig)
COMMANDS = [
'start',
'stop',
'restart',
'foreground',
'status',
'info',
'check',
'configcheck',
'jmx',
]
if len(args) < 1:
sys.stderr.write("Usage: %s %s\n" % (sys.argv[0], "|".join(COMMANDS)))
return 2
command = args[0]
if command not in COMMANDS:
sys.stderr.write("Unknown command: %s\n" % command)
return 3
pid_file = PidFile('dd-agent')
if options.clean:
pid_file.clean()
agent = Agent(pid_file.get_path(), autorestart)
if command in START_COMMANDS:
log.info('Agent version %s' % get_version())
if 'start' == command:
log.info('Start daemon')
agent.start()
elif 'stop' == command:
log.info('Stop daemon')
agent.stop()
elif 'restart' == command:
log.info('Restart daemon')
agent.restart()
elif 'status' == command:
agent.status()
elif 'info' == command:
return agent.info(verbose=options.verbose)
elif 'foreground' == command:
logging.info('Running in foreground')
if autorestart:
# Set-up the supervisor callbacks and fork it.
logging.info('Running Agent with auto-restart ON')
def child_func(): agent.run()
def parent_func(): agent.start_event = False
AgentSupervisor.start(parent_func, child_func)
else:
# Run in the standard foreground.
agent.run(config=agentConfig)
elif 'check' == command:
if len(args) < 2:
sys.stderr.write(
"Usage: %s check <check_name> [check_rate]\n"
"Add check_rate as last argument to compute rates\n"
% sys.argv[0]
)
return 1
check_name = args[1]
try:
import checks.collector
# Try the old-style check first
print getattr(checks.collector, check_name)(log).check(agentConfig)
except Exception:
# If not an old-style check, try checks.d
checks = load_check_directory(agentConfig, hostname)
for check in checks['initialized_checks']:
if check.name == check_name:
check.run()
print check.get_metrics()
print check.get_events()
if len(args) == 3 and args[2] == 'check_rate':
print "Running 2nd iteration to capture rate metrics"
time.sleep(1)
check.run()
print check.get_metrics()
print check.get_events()
check.stop()
elif 'configcheck' == command or 'configtest' == command:
osname = get_os()
all_valid = True
for conf_path in glob.glob(os.path.join(get_confd_path(osname), "*.yaml")):
basename = os.path.basename(conf_path)
try:
check_yaml(conf_path)
except Exception, e:
all_valid = False
print "%s contains errors:\n %s" % (basename, e)
else:
print "%s is valid" % basename
if all_valid:
print "All yaml files passed. You can now run the Datadog agent."
return 0
else:
print("Fix the invalid yaml files above in order to start the Datadog agent. "
"A useful external tool for yaml parsing can be found at "
"http://yaml-online-parser.appspot.com/")
return 1
def testConfigCheckNotAgentCheck(self, *args):
copyfile("%s/valid_conf.yaml" % FIXTURE_PATH, "%s/test_check.yaml" % TEMP_ETC_CONF_DIR)
copyfile("%s/invalid_check_1.py" % FIXTURE_PATH, "%s/test_check.py" % TEMP_AGENT_CHECK_DIR)
checks = load_check_directory({"additional_checksd": TEMP_ETC_CHECKS_DIR}, "foo")
self.assertEquals(0, len(checks["init_failed_checks"]))
self.assertEquals(0, len(checks["initialized_checks"]))
def run(self, config=None):
"""Main loop of the collector"""
# Gracefully exit on sigterm.
signal.signal(signal.SIGTERM, self._handle_sigterm)
# A SIGUSR1 signals an exit with an autorestart
signal.signal(signal.SIGUSR1, self._handle_sigusr1)
# Handle Keyboard Interrupt
signal.signal(signal.SIGINT, self._handle_sigterm)
# Save the agent start-up stats.
CollectorStatus().persist()
# Intialize the collector.
if not config:
config = get_config(parse_args=True)
agentConfig = self._set_agent_config_hostname(config)
systemStats = get_system_stats()
emitters = self._get_emitters(agentConfig)
# Load the checks.d checks
checksd = load_check_directory(agentConfig)
self.collector = Collector(agentConfig, emitters, systemStats)
# Configure the watchdog.
check_frequency = int(agentConfig['check_freq'])
watchdog = self._get_watchdog(check_frequency, agentConfig)
# Initialize the auto-restarter
self.restart_interval = int(agentConfig.get('restart_interval', RESTART_INTERVAL))
self.agent_start = time.time()
# Run the main loop.
while self.run_forever:
# Do the work.
self.collector.run(checksd=checksd, start_event=self.start_event)
# Check if we should restart.
if self.autorestart and self._should_restart():
self._do_restart()
# Only plan for the next loop if we will continue,
# otherwise just exit quickly.
if self.run_forever:
if watchdog:
watchdog.reset()
time.sleep(check_frequency)
# Now clean-up.
try:
CollectorStatus.remove_latest_status()
except Exception:
pass
# Explicitly kill the process, because it might be running
# as a daemon.
log.info("Exiting. Bye bye.")
sys.exit(0)
def testDefaultChecks(self):
checks = load_check_directory({"additional_checksd": "/etc/dd-agent/checks.d/"}, "foo")
init_checks_names = [c.name for c in checks['initialized_checks']]
for c in DEFAULT_CHECKS:
self.assertTrue(c in init_checks_names)
def testConfigDeprecatedNagiosConfig(self, *args):
copyfile('%s/valid_check_1.py' % FIXTURE_PATH,
'%s/nagios.py' % TEMP_ETC_CHECKS_DIR)
checks = load_check_directory({"nagios_perf_cfg": None, "additional_checksd": TEMP_ETC_CHECKS_DIR}, "foo")
self.assertEquals(1, len(checks['initialized_checks']))
self.assertEquals('valid_check_1', checks['initialized_checks'][0].check(None))
def run(self, config=None):
"""Main loop of the collector"""
# Gracefully exit on sigterm.
signal.signal(signal.SIGTERM, self._handle_sigterm)
# A SIGUSR1 signals an exit with an autorestart
signal.signal(signal.SIGUSR1, self._handle_sigusr1)
# Handle Keyboard Interrupt
signal.signal(signal.SIGINT, self._handle_sigterm)
# Save the agent start-up stats.
CollectorStatus().persist()
# Intialize the collector.
if not config:
config = get_config(parse_args=True)
agentConfig = self._set_agent_config_hostname(config)
hostname = get_hostname(agentConfig)
systemStats = get_system_stats()
emitters = self._get_emitters(agentConfig)
# Load the checks.d checks
checksd = load_check_directory(agentConfig, hostname)
self.collector = Collector(agentConfig, emitters, systemStats, hostname)
# Configure the watchdog.
check_frequency = int(agentConfig['check_freq'])
watchdog = self._get_watchdog(check_frequency, agentConfig)
# Initialize the auto-restarter
self.restart_interval = int(agentConfig.get('restart_interval', RESTART_INTERVAL))
self.agent_start = time.time()
# Run the main loop.
while self.run_forever:
# enable profiler if needed
profiled = False
if agentConfig.get('profile', False) and agentConfig.get('profile').lower() == 'yes':
try:
import cProfile
profiler = cProfile.Profile()
profiled = True
profiler.enable()
log.debug("Agent profiling is enabled")
except Exception:
log.warn("Cannot enable profiler")
# Do the work.
self.collector.run(checksd=checksd, start_event=self.start_event)
# disable profiler and printout stats to stdout
if agentConfig.get('profile', False) and agentConfig.get('profile').lower() == 'yes' and profiled:
try:
profiler.disable()
import pstats
from cStringIO import StringIO
s = StringIO()
ps = pstats.Stats(profiler, stream=s).sort_stats("cumulative")
ps.print_stats()
log.debug(s.getvalue())
except Exception:
log.warn("Cannot disable profiler")
# Check if we should restart.
if self.autorestart and self._should_restart():
self._do_restart()
# Only plan for the next loop if we will continue,
# otherwise just exit quickly.
if self.run_forever:
if watchdog:
watchdog.reset()
time.sleep(check_frequency)
# Now clean-up.
try:
CollectorStatus.remove_latest_status()
except Exception:
pass
# Explicitly kill the process, because it might be running
# as a daemon.
log.info("Exiting. Bye bye.")
sys.exit(0)
def main():
options, args = get_parsed_args()
agentConfig = get_config(options=options)
autorestart = agentConfig.get('autorestart', False)
COMMANDS = [
'start',
'stop',
'restart',
'foreground',
'status',
'info',
'check',
]
if len(args) < 1:
sys.stderr.write("Usage: %s %s\n" % (sys.argv[0], "|".join(COMMANDS)))
return 2
command = args[0]
if command not in COMMANDS:
sys.stderr.write("Unknown command: %s\n" % command)
return 3
pid_file = PidFile('dd-agent')
if options.clean:
pid_file.clean()
agent = Agent(pid_file.get_path(), autorestart)
if 'start' == command:
log.info('Start daemon')
agent.start()
elif 'stop' == command:
log.info('Stop daemon')
agent.stop()
elif 'restart' == command:
log.info('Restart daemon')
agent.restart()
elif 'status' == command:
agent.status()
elif 'info' == command:
return agent.info(verbose=options.verbose)
elif 'foreground' == command:
logging.info('Running in foreground')
if autorestart:
# Set-up the supervisor callbacks and fork it.
logging.info('Running Agent with auto-restart ON')
def child_func(): agent.run()
def parent_func(): agent.start_event = False
AgentSupervisor.start(parent_func, child_func)
else:
# Run in the standard foreground.
agent.run(config=agentConfig)
elif 'check' == command:
check_name = args[1]
try:
import checks.collector
# Try the old-style check first
print getattr(checks.collector, check_name)(log).check(agentConfig)
except Exception:
# If not an old-style check, try checks.d
checks = load_check_directory(agentConfig)
for check in checks['initialized_checks']:
if check.name == check_name:
check.run()
print check.get_metrics()
print check.get_events()
if len(args) == 3 and args[2] == 'check_rate':
print "Running 2nd iteration to capture rate metrics"
time.sleep(1)
check.run()
print check.get_metrics()
print check.get_events()
return 0
def main():
options, args = get_parsed_args()
agentConfig = get_config(options=options)
autorestart = agentConfig.get('autorestart', False)
hostname = get_hostname(agentConfig)
in_developer_mode = agentConfig.get('developer_mode')
COMMANDS_AGENT = [
'start',
'stop',
'restart',
'status',
'foreground',
]
COMMANDS_NO_AGENT = [
'info',
'check',
'configcheck',
'jmx',
'flare',
]
COMMANDS = COMMANDS_AGENT + COMMANDS_NO_AGENT
if len(args) < 1:
sys.stderr.write("Usage: %s %s\n" % (sys.argv[0], "|".join(COMMANDS)))
return 2
command = args[0]
if command not in COMMANDS:
sys.stderr.write("Unknown command: %s\n" % command)
return 3
# Deprecation notice
if command not in DD_AGENT_COMMANDS:
# Will become an error message and exit after deprecation period
from utils.deprecations import deprecate_old_command_line_tools
deprecate_old_command_line_tools()
if command in COMMANDS_AGENT:
agent = Agent(PidFile('dd-agent').get_path(), autorestart, in_developer_mode=in_developer_mode)
if command in START_COMMANDS:
log.info('Agent version %s' % get_version())
if 'start' == command:
log.info('Start daemon')
agent.start()
elif 'stop' == command:
log.info('Stop daemon')
agent.stop()
elif 'restart' == command:
log.info('Restart daemon')
agent.restart()
elif 'status' == command:
agent.status()
elif 'info' == command:
return Agent.info(verbose=options.verbose)
elif 'foreground' == command:
logging.info('Running in foreground')
if autorestart:
# Set-up the supervisor callbacks and fork it.
logging.info('Running Agent with auto-restart ON')
def child_func():
agent.start(foreground=True)
def parent_func():
agent.start_event = False
AgentSupervisor.start(parent_func, child_func)
else:
# Run in the standard foreground.
agent.start(foreground=True)
elif 'check' == command:
if len(args) < 2:
sys.stderr.write(
"Usage: %s check <check_name> [check_rate]\n"
"Add check_rate as last argument to compute rates\n"
% sys.argv[0]
)
return 1
check_name = args[1]
try:
import checks.collector
# Try the old-style check first
print getattr(checks.collector, check_name)(log).check(agentConfig)
except Exception:
# If not an old-style check, try checks.d
checks = load_check_directory(agentConfig, hostname)
for check in checks['initialized_checks']:
if check.name == check_name:
if in_developer_mode:
check.run = AgentProfiler.wrap_profiling(check.run)
cs = Collector.run_single_check(check, verbose=True)
print CollectorStatus.render_check_status(cs)
if len(args) == 3 and args[2] == 'check_rate':
print "Running 2nd iteration to capture rate metrics"
time.sleep(1)
cs = Collector.run_single_check(check, verbose=True)
print CollectorStatus.render_check_status(cs)
check.stop()
elif 'configcheck' == command or 'configtest' == command:
configcheck()
elif 'jmx' == command:
if len(args) < 2 or args[1] not in JMX_LIST_COMMANDS.keys():
print "#" * 80
print "JMX tool to be used to help configuring your JMX checks."
print "See http://docs.datadoghq.com/integrations/java/ for more information"
print "#" * 80
print "\n"
print "You have to specify one of the following commands:"
for command, desc in JMX_LIST_COMMANDS.iteritems():
print " - %s [OPTIONAL: LIST OF CHECKS]: %s" % (command, desc)
print "Example: sudo /etc/init.d/datadog-agent jmx list_matching_attributes tomcat jmx solr"
print "\n"
else:
jmx_command = args[1]
checks_list = args[2:]
confd_directory = get_confd_path(get_os())
jmx_process = JMXFetch(confd_directory, agentConfig)
jmx_process.configure()
should_run = jmx_process.should_run()
if should_run:
jmx_process.run(jmx_command, checks_list, reporter="console")
else:
print "Couldn't find any valid JMX configuration in your conf.d directory: %s" % confd_directory
print "Have you enabled any JMX check ?"
print "If you think it's not normal please get in touch with Datadog Support"
elif 'flare' == command:
Flare.check_user_rights()
case_id = int(args[1]) if len(args) > 1 else None
f = Flare(True, case_id)
f.collect()
try:
f.upload()
except Exception, e:
print 'The upload failed:\n{0}'.format(str(e))
def run(self, config=None):
"""Main loop of the collector"""
# Gracefully exit on sigterm.
signal.signal(signal.SIGTERM, self._handle_sigterm)
# A SIGUSR1 signals an exit with an autorestart
signal.signal(signal.SIGUSR1, self._handle_sigusr1)
# Handle Keyboard Interrupt
signal.signal(signal.SIGINT, self._handle_sigterm)
# A SIGHUP signals a configuration reload
signal.signal(signal.SIGHUP, self._handle_sighup)
# Save the agent start-up stats.
CollectorStatus().persist()
# Intialize the collector.
if not config:
config = get_config(parse_args=True)
self._agentConfig = self._set_agent_config_hostname(config)
hostname = get_hostname(self._agentConfig)
systemStats = get_system_stats()
emitters = self._get_emitters()
# Initialize service discovery
if self._agentConfig.get("service_discovery"):
self.sd_backend = get_sd_backend(self._agentConfig)
# Load the checks.d checks
self._checksd = load_check_directory(self._agentConfig, hostname)
# Initialize the Collector
self.collector = Collector(self._agentConfig, emitters, systemStats, hostname)
# In developer mode, the number of runs to be included in a single collector profile
self.collector_profile_interval = self._agentConfig.get(
"collector_profile_interval", DEFAULT_COLLECTOR_PROFILE_INTERVAL
)
# Configure the watchdog.
self.check_frequency = int(self._agentConfig["check_freq"])
watchdog = self._get_watchdog(self.check_frequency)
# Initialize the auto-restarter
self.restart_interval = int(self._agentConfig.get("restart_interval", RESTART_INTERVAL))
self.agent_start = time.time()
profiled = False
collector_profiled_runs = 0
# Run the main loop.
while self.run_forever:
log.debug("Found {num_checks} checks".format(num_checks=len(self._checksd["initialized_checks"])))
# Setup profiling if necessary
if self.in_developer_mode and not profiled:
try:
profiler = AgentProfiler()
profiler.enable_profiling()
profiled = True
except Exception as e:
log.warn("Cannot enable profiler: %s" % str(e))
# Do the work.
self.collector.run(
checksd=self._checksd, start_event=self.start_event, configs_reloaded=self.configs_reloaded
)
# This flag is used to know if the check configs have been reloaded at the current
# run of the agent yet or not. It's used by the collector to know if it needs to
# look for the AgentMetrics check and pop it out.
# See: https://github.com/DataDog/dd-agent/blob/5.6.x/checks/collector.py#L265-L272
self.configs_reloaded = False
# Look for change in the config template store.
# The self.sd_backend.reload_check_configs flag is set
# to True if a config reload is needed.
if (
self._agentConfig.get("service_discovery")
and self.sd_backend
and not self.sd_backend.reload_check_configs
):
try:
self.sd_backend.reload_check_configs = get_config_store(self._agentConfig).crawl_config_template()
except Exception as e:
log.warn("Something went wrong while looking for config template changes: %s" % str(e))
# Check if we should run service discovery
# The `reload_check_configs` flag can be set through the docker_daemon check or
# using ConfigStore.crawl_config_template
if self._agentConfig.get("service_discovery") and self.sd_backend and self.sd_backend.reload_check_configs:
self.reload_configs()
self.configs_reloaded = True
self.sd_backend.reload_check_configs = False
if profiled:
if collector_profiled_runs >= self.collector_profile_interval:
try:
profiler.disable_profiling()
profiled = False
collector_profiled_runs = 0
except Exception as e:
log.warn("Cannot disable profiler: %s" % str(e))
# Check if we should restart.
if self.autorestart and self._should_restart():
self._do_restart()
# Only plan for next loop if we will continue, otherwise exit quickly.
if self.run_forever:
if watchdog:
watchdog.reset()
if profiled:
collector_profiled_runs += 1
log.debug("Sleeping for {0} seconds".format(self.check_frequency))
time.sleep(self.check_frequency)
# Now clean-up.
try:
CollectorStatus.remove_latest_status()
except Exception:
pass
# Explicitly kill the process, because it might be running as a daemon.
log.info("Exiting. Bye bye.")
sys.exit(0)
def run(self, config=None):
"""Main loop of the collector"""
# Gracefully exit on sigterm.
signal.signal(signal.SIGTERM, self._handle_sigterm)
# A SIGUSR1 signals an exit with an autorestart
signal.signal(signal.SIGUSR1, self._handle_sigusr1)
# Handle Keyboard Interrupt
signal.signal(signal.SIGINT, self._handle_sigterm)
# Save the agent start-up stats.
CollectorStatus().persist()
# Intialize the collector.
if not config:
config = get_config(parse_args=True)
agentConfig = self._set_agent_config_hostname(config)
systemStats = get_system_stats()
emitters = self._get_emitters(agentConfig)
self.collector = Collector(agentConfig, emitters, systemStats)
# Load the checks.d checks
checksd = load_check_directory(agentConfig)
# Configure the watchdog.
check_frequency = int(agentConfig['check_freq'])
watchdog = self._get_watchdog(check_frequency, agentConfig)
# Initialize the auto-restarter
self.restart_interval = int(agentConfig.get('restart_interval', RESTART_INTERVAL))
self.agent_start = time.time()
# Run the main loop.
while self.run_forever:
# Do the work.
self.collector.run(checksd=checksd, start_event=self.start_event)
# Check if we should restart.
if self.autorestart and self._should_restart():
self._do_restart()
# Only plan for the next loop if we will continue,
# otherwise just exit quickly.
if self.run_forever:
if watchdog:
watchdog.reset()
time.sleep(check_frequency)
# Now clean-up.
try:
CollectorStatus.remove_latest_status()
except:
pass
# Explicitly kill the process, because it might be running
# as a daemon.
log.info("Exiting. Bye bye.")
sys.exit(0)
def main():
options, args = get_parsed_args()
agentConfig = get_config(options=options)
autorestart = agentConfig.get('autorestart', False)
COMMANDS = [
'start',
'stop',
'restart',
'foreground',
'status',
'info',
'check',
]
if len(args) < 1:
sys.stderr.write("Usage: %s %s\n" % (sys.argv[0], "|".join(COMMANDS)))
return 2
command = args[0]
if command not in COMMANDS:
sys.stderr.write("Unknown command: %s\n" % command)
return 3
pid_file = PidFile('dd-agent')
if options.clean:
pid_file.clean()
agent = Agent(pid_file.get_path(), autorestart)
if 'start' == command:
log.info('Start daemon')
agent.start()
elif 'stop' == command:
log.info('Stop daemon')
agent.stop()
elif 'restart' == command:
log.info('Restart daemon')
agent.restart()
elif 'status' == command:
agent.status()
elif 'info' == command:
agent.info(verbose=options.verbose)
elif 'foreground' == command:
logging.info('Running in foreground')
if autorestart:
# Set-up the supervisor callbacks and fork it.
logging.info('Running Agent with auto-restart ON')
def child_func(): agent.run()
def parent_func(): agent.start_event = False
AgentSupervisor.start(parent_func, child_func)
else:
# Run in the standard foreground.
agent.run(config=agentConfig)
elif 'check' == command:
check_name = args[1]
try:
import checks.collector
# Try the old-style check first
print getattr(checks.collector, check_name)(log).check(agentConfig)
except Exception:
# If not an old-style check, try checks.d
checks = load_check_directory(agentConfig)
for check in checks:
if check.name == check_name:
check.run()
print check.get_metrics()
print check.get_events()
if len(args) == 3 and args[2] == 'check_rate':
print "Running 2nd iteration to capture rate metrics"
time.sleep(1)
check.run()
print check.get_metrics()
print check.get_events()
return 0
def testConfigNotFound(self, *args):
copyfile('%s/valid_conf.yaml' % FIXTURE_PATH,
'%s/test_check.yaml' % TEMP_ETC_CONF_DIR)
checks = load_check_directory({"additional_checksd": TEMP_ETC_CHECKS_DIR}, "foo")
self.assertEquals(0, len(checks['init_failed_checks']))
self.assertEquals(0, len(checks['initialized_checks']))
def main():
options, args = get_parsed_args()
agentConfig = get_config(options=options)
autorestart = agentConfig.get('autorestart', False)
hostname = get_hostname(agentConfig)
in_developer_mode = agentConfig.get('developer_mode')
COMMANDS_AGENT = [
'start',
'stop',
'restart',
'status',
'foreground',
]
COMMANDS_NO_AGENT = [
'info',
'check',
'configcheck',
'jmx',
'flare',
]
COMMANDS = COMMANDS_AGENT + COMMANDS_NO_AGENT
if len(args) < 1:
sys.stderr.write("Usage: %s %s\n" % (sys.argv[0], "|".join(COMMANDS)))
return 2
command = args[0]
if command not in COMMANDS:
sys.stderr.write("Unknown command: %s\n" % command)
return 3
# TODO: actually kill the start/stop/restart/status command for 5.11
if command in ['start', 'stop', 'restart', 'status'] and not in_developer_mode:
logging.error('Please use supervisor to manage the agent')
return 1
if command in COMMANDS_AGENT:
agent = Agent(PidFile(PID_NAME, PID_DIR).get_path(), autorestart, in_developer_mode=in_developer_mode)
if 'start' == command:
log.info('Start daemon')
agent.start()
elif 'stop' == command:
log.info('Stop daemon')
agent.stop()
elif 'restart' == command:
log.info('Restart daemon')
agent.restart()
elif 'status' == command:
agent.status()
elif 'info' == command:
return Agent.info(verbose=options.verbose)
elif 'foreground' == command:
log.info('Agent version %s' % get_version())
if autorestart:
# Set-up the supervisor callbacks and fork it.
logging.info('Running Agent with auto-restart ON')
def child_func():
agent.start(foreground=True)
def parent_func():
agent.start_event = False
AgentSupervisor.start(parent_func, child_func)
else:
# Run in the standard foreground.
agent.start(foreground=True)
elif 'check' == command:
if len(args) < 2:
sys.stderr.write(
"Usage: %s check <check_name> [check_rate]\n"
"Add check_rate as last argument to compute rates\n"
% sys.argv[0]
)
return 1
check_name = args[1]
try:
import checks.collector
# Try the old-style check first
print getattr(checks.collector, check_name)(log).check(agentConfig)
except Exception:
# If not an old-style check, try checks.d
checks = load_check_directory(agentConfig, hostname)
for check in checks['initialized_checks']:
if check.name == check_name:
if in_developer_mode:
check.run = AgentProfiler.wrap_profiling(check.run)
cs = Collector.run_single_check(check, verbose=True)
print CollectorStatus.render_check_status(cs)
if len(args) == 3 and args[2] == 'check_rate':
print "Running 2nd iteration to capture rate metrics"
time.sleep(1)
cs = Collector.run_single_check(check, verbose=True)
print CollectorStatus.render_check_status(cs)
check.stop()
elif 'configcheck' == command or 'configtest' == command:
configcheck()
sd_configcheck(agentConfig)
elif 'jmx' == command:
jmx_command(args[1:], agentConfig)
elif 'flare' == command:
Flare.check_user_rights()
case_id = int(args[1]) if len(args) > 1 else None
f = Flare(True, case_id)
f.collect()
try:
f.upload()
except Exception as e:
print 'The upload failed:\n{0}'.format(str(e))
return 0
def testDefaultChecks(self):
checks = load_check_directory({"additional_checksd": "/etc/dd-agent/checks.d/"}, "foo")
init_checks_names = [c.name for c in checks['initialized_checks']]
for c in DEFAULT_CHECKS:
self.assertTrue(c in init_checks_names)
def run(self, config=None):
"""Main loop of the collector"""
# Gracefully exit on sigterm.
signal.signal(signal.SIGTERM, self._handle_sigterm)
# A SIGUSR1 signals an exit with an autorestart
signal.signal(signal.SIGUSR1, self._handle_sigusr1)
# Handle Keyboard Interrupt
signal.signal(signal.SIGINT, self._handle_sigterm)
# Save the agent start-up stats.
CollectorStatus().persist()
# Intialize the collector.
if not config:
config = get_config(parse_args=True)
agentConfig = self._set_agent_config_hostname(config)
hostname = get_hostname(agentConfig)
systemStats = get_system_stats()
emitters = self._get_emitters(agentConfig)
# Load the checks.d checks
checksd = load_check_directory(agentConfig, hostname)
self.collector = Collector(agentConfig, emitters, systemStats, hostname)
# Configure the watchdog.
check_frequency = int(agentConfig['check_freq'])
watchdog = self._get_watchdog(check_frequency, agentConfig)
# Initialize the auto-restarter
self.restart_interval = int(agentConfig.get('restart_interval', RESTART_INTERVAL))
self.agent_start = time.time()
# Run the main loop.
while self.run_forever:
# enable profiler if needed
profiled = False
if agentConfig.get('profile', False) and agentConfig.get('profile').lower() == 'yes':
try:
import cProfile
profiler = cProfile.Profile()
profiled = True
profiler.enable()
log.debug("Agent profiling is enabled")
except Exception:
log.warn("Cannot enable profiler")
# Do the work.
self.collector.run(checksd=checksd, start_event=self.start_event)
# disable profiler and printout stats to stdout
if agentConfig.get('profile', False) and agentConfig.get('profile').lower() == 'yes' and profiled:
try:
profiler.disable()
import pstats
from cStringIO import StringIO
s = StringIO()
ps = pstats.Stats(profiler, stream=s).sort_stats("cumulative")
ps.print_stats()
log.debug(s.getvalue())
except Exception:
log.warn("Cannot disable profiler")
# Check if we should restart.
if self.autorestart and self._should_restart():
self._do_restart()
# Only plan for the next loop if we will continue,
# otherwise just exit quickly.
if self.run_forever:
if watchdog:
watchdog.reset()
time.sleep(check_frequency)
# Now clean-up.
try:
CollectorStatus.remove_latest_status()
except Exception:
pass
# Explicitly kill the process, because it might be running
# as a daemon.
log.info("Exiting. Bye bye.")
sys.exit(0)
def main():
options, args = get_parsed_args()
agentConfig = get_config(options=options)
autorestart = agentConfig.get('autorestart', False)
hostname = get_hostname(agentConfig)
in_developer_mode = agentConfig.get('developer_mode')
# do this early on
if agentConfig.get('disable_unsafe_yaml'):
monkey_patch_pyyaml()
COMMANDS_AGENT = [
'start',
'stop',
'restart',
'status',
'foreground',
]
COMMANDS_NO_AGENT = [
'info',
'check',
'configcheck',
'jmx',
'flare',
]
COMMANDS = COMMANDS_AGENT + COMMANDS_NO_AGENT
if len(args) < 1:
sys.stderr.write("Usage: %s %s\n" % (sys.argv[0], "|".join(COMMANDS)))
return 2
command = args[0]
if command not in COMMANDS:
sys.stderr.write("Unknown command: %s\n" % command)
return 3
# TODO: actually kill the start/stop/restart/status command for 5.11
if command in ['start', 'stop', 'restart', 'status'
] and not in_developer_mode:
logging.error('Please use supervisor to manage the agent')
return 1
if command in COMMANDS_AGENT:
agent = Agent(PidFile(PID_NAME, PID_DIR).get_path(),
autorestart,
in_developer_mode=in_developer_mode)
if 'start' == command:
log.info('Start daemon')
agent.start()
elif 'stop' == command:
log.info('Stop daemon')
agent.stop()
elif 'restart' == command:
log.info('Restart daemon')
agent.restart()
elif 'status' == command:
agent.status()
elif 'info' == command:
return Agent.info(verbose=options.verbose)
elif 'foreground' == command:
log.info('Agent version %s' % get_version())
if autorestart:
# Set-up the supervisor callbacks and fork it.
logging.info('Running Agent with auto-restart ON')
def child_func():
agent.start(foreground=True)
def parent_func():
agent.start_event = False
AgentSupervisor.start(parent_func, child_func)
else:
# Run in the standard foreground.
agent.start(foreground=True)
elif 'check' == command:
if len(args) < 2:
sys.stderr.write(
"Usage: %s check <check_name> [check_rate]\n"
"Add check_rate as last argument to compute rates\n" %
sys.argv[0])
return 1
check_name = args[1]
try:
import checks.collector
# Try the old-style check first
print getattr(checks.collector, check_name)(log).check(agentConfig)
except Exception:
# If not an old-style check, try checks.d
checks = load_check_directory(agentConfig, hostname)
for check in checks['initialized_checks']:
if check.name == check_name:
if in_developer_mode:
check.run = AgentProfiler.wrap_profiling(check.run)
cs = Collector.run_single_check(check, verbose=True)
print CollectorStatus.render_check_status(cs)
if len(args) == 3 and args[2] == 'check_rate':
print "Running 2nd iteration to capture rate metrics"
time.sleep(1)
cs = Collector.run_single_check(check, verbose=True)
print CollectorStatus.render_check_status(cs)
else:
print "Check has run only once, if some metrics are missing you can run the command again with the 'check_rate' argument appended at the end to see any other metrics if available."
check.stop()
elif 'configcheck' == command or 'configtest' == command:
sd_configcheck(agentConfig)
return configcheck()
elif 'jmx' == command:
jmx_command(args[1:], agentConfig)
elif 'flare' == command:
Flare.check_user_rights()
case_id = int(args[1]) if len(args) > 1 else None
f = Flare(True, case_id)
f.collect()
try:
f.upload()
except Exception as e:
print 'The upload failed:\n{0}'.format(str(e))
return 0
def main():
options, args = get_parsed_args()
agentConfig = get_config(options=options)
autorestart = agentConfig.get('autorestart', False)
hostname = get_hostname(agentConfig)
in_developer_mode = agentConfig.get('developer_mode')
COMMANDS_AGENT = [
'start',
'stop',
'restart',
'status',
'foreground',
]
COMMANDS_NO_AGENT = [
'info',
'check',
'configcheck',
'jmx',
'flare',
]
COMMANDS = COMMANDS_AGENT + COMMANDS_NO_AGENT
if len(args) < 1:
sys.stderr.write("Usage: %s %s\n" % (sys.argv[0], "|".join(COMMANDS)))
return 2
command = args[0]
if command not in COMMANDS:
sys.stderr.write("Unknown command: %s\n" % command)
return 3
# Deprecation notice
if command not in DD_AGENT_COMMANDS:
# Will become an error message and exit after deprecation period
from utils.deprecations import deprecate_old_command_line_tools
deprecate_old_command_line_tools()
if command in COMMANDS_AGENT:
agent = Agent(PidFile(PID_NAME, PID_DIR).get_path(), autorestart, in_developer_mode=in_developer_mode)
if command in START_COMMANDS:
log.info('Agent version %s' % get_version())
if 'start' == command:
log.info('Start daemon')
agent.start()
elif 'stop' == command:
log.info('Stop daemon')
agent.stop()
elif 'restart' == command:
log.info('Restart daemon')
agent.restart()
elif 'status' == command:
agent.status()
elif 'info' == command:
return Agent.info(verbose=options.verbose)
elif 'foreground' == command:
logging.info('Running in foreground')
if autorestart:
# Set-up the supervisor callbacks and fork it.
logging.info('Running Agent with auto-restart ON')
def child_func():
agent.start(foreground=True)
def parent_func():
agent.start_event = False
AgentSupervisor.start(parent_func, child_func)
else:
# Run in the standard foreground.
agent.start(foreground=True)
elif 'check' == command:
if len(args) < 2:
sys.stderr.write(
"Usage: %s check <check_name> [check_rate]\n"
"Add check_rate as last argument to compute rates\n"
% sys.argv[0]
)
return 1
check_name = args[1]
try:
import checks.collector
# Try the old-style check first
print getattr(checks.collector, check_name)(log).check(agentConfig)
except Exception:
# If not an old-style check, try checks.d
checks = load_check_directory(agentConfig, hostname)
for check in checks['initialized_checks']:
if check.name == check_name:
if in_developer_mode:
check.run = AgentProfiler.wrap_profiling(check.run)
cs = Collector.run_single_check(check, verbose=True)
print CollectorStatus.render_check_status(cs)
if len(args) == 3 and args[2] == 'check_rate':
print "Running 2nd iteration to capture rate metrics"
time.sleep(1)
cs = Collector.run_single_check(check, verbose=True)
print CollectorStatus.render_check_status(cs)
check.stop()
elif 'configcheck' == command or 'configtest' == command:
configcheck()
sd_configcheck(agentConfig)
elif 'jmx' == command:
jmx_command(args[1:], agentConfig)
elif 'flare' == command:
Flare.check_user_rights()
case_id = int(args[1]) if len(args) > 1 else None
f = Flare(True, case_id)
f.collect()
try:
f.upload()
except Exception as e:
print 'The upload failed:\n{0}'.format(str(e))
return 0
def run(self, config=None):
"""Main loop of the collector"""
# Gracefully exit on sigterm.
signal.signal(signal.SIGTERM, self._handle_sigterm)
# A SIGUSR1 signals an exit with an autorestart
signal.signal(signal.SIGUSR1, self._handle_sigusr1)
# Handle Keyboard Interrupt
signal.signal(signal.SIGINT, self._handle_sigterm)
# A SIGHUP signals a configuration reload
signal.signal(signal.SIGHUP, self._handle_sighup)
# Save the agent start-up stats.
CollectorStatus().persist()
# Intialize the collector.
if not config:
config = get_config(parse_args=True)
self._agentConfig = self._set_agent_config_hostname(config)
hostname = get_hostname(self._agentConfig)
systemStats = get_system_stats(
proc_path=self._agentConfig.get('procfs_path', '/proc').rstrip('/')
)
emitters = self._get_emitters()
# Initialize service discovery
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)
# In developer mode, the number of runs to be included in a single collector profile
try:
self.collector_profile_interval = int(
self._agentConfig.get('collector_profile_interval', DEFAULT_COLLECTOR_PROFILE_INTERVAL))
except ValueError:
log.warn('collector_profile_interval is invalid. '
'Using default value instead (%s).' % DEFAULT_COLLECTOR_PROFILE_INTERVAL)
self.collector_profile_interval = DEFAULT_COLLECTOR_PROFILE_INTERVAL
# Configure the watchdog.
self.check_frequency = int(self._agentConfig['check_freq'])
watchdog = self._get_watchdog(self.check_frequency)
# Initialize the auto-restarter
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.allow_profiling and self.in_developer_mode and not profiled:
try:
profiler = AgentProfiler()
profiler.enable_profiling()
profiled = True
except Exception as e:
log.warn("Cannot enable profiler: %s" % str(e))
if self.reload_configs_flag:
if isinstance(self.reload_configs_flag, set):
self.reload_configs(checks_to_reload=self.reload_configs_flag)
else:
self.reload_configs()
# Do the work. Pass `configs_reloaded` to let the collector know if it needs to
# look for the AgentMetrics check and pop it out.
self.collector.run(checksd=self._checksd,
start_event=self.start_event,
configs_reloaded=True if self.reload_configs_flag else False)
self.reload_configs_flag = False
# Look for change in the config template store.
# The self.sd_backend.reload_check_configs flag is set
# to True if a config reload is needed.
if self._agentConfig.get('service_discovery') and self.sd_backend and \
not self.sd_backend.reload_check_configs:
try:
self.sd_backend.reload_check_configs = get_config_store(
self._agentConfig).crawl_config_template()
except Exception as e:
log.warn('Something went wrong while looking for config template changes: %s' % str(e))
# Check if we should run service discovery
# The `reload_check_configs` flag can be set through the docker_daemon check or
# using ConfigStore.crawl_config_template
if self._agentConfig.get('service_discovery') and self.sd_backend and \
self.sd_backend.reload_check_configs:
self.reload_configs_flag = self.sd_backend.reload_check_configs
self.sd_backend.reload_check_configs = False
if profiled:
if collector_profiled_runs >= self.collector_profile_interval:
try:
profiler.disable_profiling()
profiled = False
collector_profiled_runs = 0
except Exception as e:
log.warn("Cannot disable profiler: %s" % str(e))
# Check if we should restart.
if self.autorestart and self._should_restart():
self._do_restart()
# Only plan for next loop if we will continue, otherwise exit quickly.
if self.run_forever:
if watchdog:
watchdog.reset()
if profiled:
collector_profiled_runs += 1
log.debug("Sleeping for {0} seconds".format(self.check_frequency))
time.sleep(self.check_frequency)
# Now clean-up.
try:
CollectorStatus.remove_latest_status()
except Exception:
pass
# Explicitly kill the process, because it might be running as a daemon.
log.info("Exiting. Bye bye.")
sys.exit(0)
def run(self, config=None):
"""Main loop of the collector"""
# Gracefully exit on sigterm.
signal.signal(signal.SIGTERM, self._handle_sigterm)
# A SIGUSR1 signals an exit with an autorestart
signal.signal(signal.SIGUSR1, self._handle_sigusr1)
# Handle Keyboard Interrupt
signal.signal(signal.SIGINT, self._handle_sigterm)
# A SIGHUP signals a configuration reload
signal.signal(signal.SIGHUP, self._handle_sighup)
# Save the agent start-up stats.
CollectorStatus().persist()
# Intialize the collector.
if not config:
config = get_config(parse_args=True)
self._agentConfig = self._set_agent_config_hostname(config)
hostname = get_hostname(self._agentConfig)
systemStats = get_system_stats()
emitters = self._get_emitters()
# Load the checks.d checks
self._checksd = load_check_directory(self._agentConfig, hostname)
# Initialize the Collector
self.collector = Collector(self._agentConfig, emitters, systemStats, hostname)
# In developer mode, the number of runs to be included in a single collector profile
self.collector_profile_interval = self._agentConfig.get('collector_profile_interval',
DEFAULT_COLLECTOR_PROFILE_INTERVAL)
# Configure the watchdog.
self.check_frequency = int(self._agentConfig['check_freq'])
watchdog = self._get_watchdog(self.check_frequency)
# Initialize the auto-restarter
self.restart_interval = int(self._agentConfig.get('restart_interval', RESTART_INTERVAL))
self.agent_start = time.time()
profiled = False
collector_profiled_runs = 0
# Run the main loop.
while self.run_forever:
log.debug("Found {num_checks} checks".format(num_checks=len(self._checksd['initialized_checks'])))
# Setup profiling if necessary
if self.in_developer_mode and not profiled:
try:
profiler = AgentProfiler()
profiler.enable_profiling()
profiled = True
except Exception as e:
log.warn("Cannot enable profiler: %s" % str(e))
# Do the work.
self.collector.run(checksd=self._checksd,
start_event=self.start_event,
configs_reloaded=self.configs_reloaded)
if self.configs_reloaded:
self.configs_reloaded = False
if profiled:
if collector_profiled_runs >= self.collector_profile_interval:
try:
profiler.disable_profiling()
profiled = False
collector_profiled_runs = 0
except Exception as e:
log.warn("Cannot disable profiler: %s" % str(e))
# Check if we should restart.
if self.autorestart and self._should_restart():
self._do_restart()
# Only plan for next loop if we will continue, otherwise exit quickly.
if self.run_forever:
if watchdog:
watchdog.reset()
if profiled:
collector_profiled_runs += 1
log.info("Sleeping for {0} seconds".format(self.check_frequency))
time.sleep(self.check_frequency)
# Now clean-up.
try:
CollectorStatus.remove_latest_status()
except Exception:
pass
# Explicitly kill the process, because it might be running as a daemon.
log.info("Exiting. Bye bye.")
sys.exit(0)
def main():
options, args = get_parsed_args()
agentConfig = get_config(options=options)
autorestart = agentConfig.get('autorestart', False)
hostname = get_hostname(agentConfig)
COMMANDS = [
'start',
'stop',
'restart',
'foreground',
'status',
'info',
'check',
'configcheck',
'jmx',
]
if len(args) < 1:
sys.stderr.write("Usage: %s %s\n" % (sys.argv[0], "|".join(COMMANDS)))
return 2
command = args[0]
if command not in COMMANDS:
sys.stderr.write("Unknown command: %s\n" % command)
return 3
pid_file = PidFile('dd-agent')
if options.clean:
pid_file.clean()
agent = Agent(pid_file.get_path(), autorestart)
if command in START_COMMANDS:
log.info('Agent version %s' % get_version())
if 'start' == command:
log.info('Start daemon')
agent.start()
elif 'stop' == command:
log.info('Stop daemon')
agent.stop()
elif 'restart' == command:
log.info('Restart daemon')
agent.restart()
elif 'status' == command:
agent.status()
elif 'info' == command:
return agent.info(verbose=options.verbose)
elif 'foreground' == command:
logging.info('Running in foreground')
if autorestart:
# Set-up the supervisor callbacks and fork it.
logging.info('Running Agent with auto-restart ON')
def child_func(): agent.run()
def parent_func(): agent.start_event = False
AgentSupervisor.start(parent_func, child_func)
else:
# Run in the standard foreground.
agent.run(config=agentConfig)
elif 'check' == command:
check_name = args[1]
try:
import checks.collector
# Try the old-style check first
print getattr(checks.collector, check_name)(log).check(agentConfig)
except Exception:
# If not an old-style check, try checks.d
checks = load_check_directory(agentConfig, hostname)
for check in checks['initialized_checks']:
if check.name == check_name:
check.run()
print check.get_metrics()
print check.get_events()
if len(args) == 3 and args[2] == 'check_rate':
print "Running 2nd iteration to capture rate metrics"
time.sleep(1)
check.run()
print check.get_metrics()
print check.get_events()
elif 'configcheck' == command or 'configtest' == command:
osname = get_os()
all_valid = True
for conf_path in glob.glob(os.path.join(get_confd_path(osname), "*.yaml")):
basename = os.path.basename(conf_path)
try:
check_yaml(conf_path)
except Exception, e:
all_valid = False
print "%s contains errors:\n %s" % (basename, e)
else:
print "%s is valid" % basename
if all_valid:
print "All yaml files passed. You can now run the Datadog agent."
return 0
else:
print("Fix the invalid yaml files above in order to start the Datadog agent. "
"A useful external tool for yaml parsing can be found at "
"http://yaml-online-parser.appspot.com/")
return 1
def testConfigETCOnly(self, *args):
copyfile("%s/valid_conf.yaml" % FIXTURE_PATH, "%s/test_check.yaml" % TEMP_ETC_CONF_DIR)
copyfile("%s/valid_check_1.py" % FIXTURE_PATH, "%s/test_check.py" % TEMP_ETC_CHECKS_DIR)
checks = load_check_directory({"additional_checksd": TEMP_ETC_CHECKS_DIR}, "foo")
self.assertEquals(1, len(checks["initialized_checks"]))
def main():
options, args = get_parsed_args()
agentConfig = get_config(options=options)
autorestart = agentConfig.get("autorestart", False)
hostname = get_hostname(agentConfig)
in_developer_mode = agentConfig.get("developer_mode")
COMMANDS_AGENT = ["start", "stop", "restart", "status", "foreground"]
COMMANDS_NO_AGENT = ["info", "check", "configcheck", "jmx", "flare"]
COMMANDS = COMMANDS_AGENT + COMMANDS_NO_AGENT
if len(args) < 1:
sys.stderr.write("Usage: %s %s\n" % (sys.argv[0], "|".join(COMMANDS)))
return 2
command = args[0]
if command not in COMMANDS:
sys.stderr.write("Unknown command: %s\n" % command)
return 3
# Deprecation notice
if command not in DD_AGENT_COMMANDS:
# Will become an error message and exit after deprecation period
from utils.deprecations import deprecate_old_command_line_tools
deprecate_old_command_line_tools()
if command in COMMANDS_AGENT:
agent = Agent(PidFile(PID_NAME, PID_DIR).get_path(), autorestart, in_developer_mode=in_developer_mode)
if command in START_COMMANDS:
log.info("Agent version %s" % get_version())
if "start" == command:
log.info("Start daemon")
agent.start()
elif "stop" == command:
log.info("Stop daemon")
agent.stop()
elif "restart" == command:
log.info("Restart daemon")
agent.restart()
elif "status" == command:
agent.status()
elif "info" == command:
return Agent.info(verbose=options.verbose)
elif "foreground" == command:
logging.info("Running in foreground")
if autorestart:
# Set-up the supervisor callbacks and fork it.
logging.info("Running Agent with auto-restart ON")
def child_func():
agent.start(foreground=True)
def parent_func():
agent.start_event = False
AgentSupervisor.start(parent_func, child_func)
else:
# Run in the standard foreground.
agent.start(foreground=True)
elif "check" == command:
if len(args) < 2:
sys.stderr.write(
"Usage: %s check <check_name> [check_rate]\n"
"Add check_rate as last argument to compute rates\n" % sys.argv[0]
)
return 1
check_name = args[1]
try:
import checks.collector
# Try the old-style check first
print getattr(checks.collector, check_name)(log).check(agentConfig)
except Exception:
# If not an old-style check, try checks.d
checks = load_check_directory(agentConfig, hostname)
for check in checks["initialized_checks"]:
if check.name == check_name:
if in_developer_mode:
check.run = AgentProfiler.wrap_profiling(check.run)
cs = Collector.run_single_check(check, verbose=True)
print CollectorStatus.render_check_status(cs)
if len(args) == 3 and args[2] == "check_rate":
print "Running 2nd iteration to capture rate metrics"
time.sleep(1)
cs = Collector.run_single_check(check, verbose=True)
print CollectorStatus.render_check_status(cs)
check.stop()
elif "configcheck" == command or "configtest" == command:
configcheck()
if agentConfig.get("service_discovery", False):
# set the TRACE_CONFIG flag to True to make load_check_directory return
# the source of config objects.
# Then call load_check_directory here and pass the result to sd_configcheck
# to avoid circular imports
agentConfig[TRACE_CONFIG] = True
configs = {
# check_name: (config_source, config)
}
print ("\nLoading check configurations...\n\n")
configs = load_check_directory(agentConfig, hostname)
sd_configcheck(agentConfig, configs)
elif "jmx" == command:
jmx_command(args[1:], agentConfig)
elif "flare" == command:
Flare.check_user_rights()
case_id = int(args[1]) if len(args) > 1 else None
f = Flare(True, case_id)
f.collect()
try:
f.upload()
except Exception, e:
print "The upload failed:\n{0}".format(str(e))
def run(self, config=None):
"""Main loop of the collector"""
# Gracefully exit on sigterm.
signal.signal(signal.SIGTERM, self._handle_sigterm)
# A SIGUSR1 signals an exit with an autorestart
signal.signal(signal.SIGUSR1, self._handle_sigusr1)
# Handle Keyboard Interrupt
signal.signal(signal.SIGINT, self._handle_sigterm)
# A SIGHUP signals a configuration reload
signal.signal(signal.SIGHUP, self._handle_sighup)
# Save the agent start-up stats.
CollectorStatus().persist()
# Intialize the collector.
if not config:
config = get_config(parse_args=True)
self._agentConfig = self._set_agent_config_hostname(config)
hostname = get_hostname(self._agentConfig)
systemStats = get_system_stats(
proc_path=self._agentConfig.get('procfs_path', '/proc').rstrip('/')
)
emitters = self._get_emitters()
# Initialize service discovery
if self._agentConfig.get('service_discovery'):
self.sd_backend = get_sd_backend(self._agentConfig)
# Load the checks.d checks
self._checksd = load_check_directory(self._agentConfig, hostname)
# Initialize the Collector
self.collector = Collector(self._agentConfig, emitters, systemStats, hostname)
# In developer mode, the number of runs to be included in a single collector profile
self.collector_profile_interval = self._agentConfig.get('collector_profile_interval',
DEFAULT_COLLECTOR_PROFILE_INTERVAL)
# Configure the watchdog.
self.check_frequency = int(self._agentConfig['check_freq'])
watchdog = self._get_watchdog(self.check_frequency)
# Initialize the auto-restarter
self.restart_interval = int(self._agentConfig.get('restart_interval', RESTART_INTERVAL))
self.agent_start = time.time()
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:
try:
profiler = AgentProfiler()
profiler.enable_profiling()
profiled = True
except Exception as e:
log.warn("Cannot enable profiler: %s" % str(e))
# Do the work.
self.collector.run(checksd=self._checksd,
start_event=self.start_event,
configs_reloaded=self.configs_reloaded)
# This flag is used to know if the check configs have been reloaded at the current
# run of the agent yet or not. It's used by the collector to know if it needs to
# look for the AgentMetrics check and pop it out.
# See: https://github.com/DataDog/dd-agent/blob/5.6.x/checks/collector.py#L265-L272
self.configs_reloaded = False
# Look for change in the config template store.
# The self.sd_backend.reload_check_configs flag is set
# to True if a config reload is needed.
if self._agentConfig.get('service_discovery') and self.sd_backend and \
not self.sd_backend.reload_check_configs:
try:
self.sd_backend.reload_check_configs = get_config_store(
self._agentConfig).crawl_config_template()
except Exception as e:
log.warn('Something went wrong while looking for config template changes: %s' % str(e))
# Check if we should run service discovery
# The `reload_check_configs` flag can be set through the docker_daemon check or
# using ConfigStore.crawl_config_template
if self._agentConfig.get('service_discovery') and self.sd_backend and \
self.sd_backend.reload_check_configs:
self.reload_configs()
self.configs_reloaded = True
self.sd_backend.reload_check_configs = False
if profiled:
if collector_profiled_runs >= self.collector_profile_interval:
try:
profiler.disable_profiling()
profiled = False
collector_profiled_runs = 0
except Exception as e:
log.warn("Cannot disable profiler: %s" % str(e))
# Check if we should restart.
if self.autorestart and self._should_restart():
self._do_restart()
# Only plan for next loop if we will continue, otherwise exit quickly.
if self.run_forever:
if watchdog:
watchdog.reset()
if profiled:
collector_profiled_runs += 1
log.debug("Sleeping for {0} seconds".format(self.check_frequency))
time.sleep(self.check_frequency)
# Now clean-up.
try:
CollectorStatus.remove_latest_status()
except Exception:
pass
# Explicitly kill the process, because it might be running as a daemon.
log.info("Exiting. Bye bye.")
sys.exit(0)