def run(self, command=None, check_list=None, reporter=None):
if check_list or self.jmx_checks is None:
# (Re)set/(re)configure JMXFetch parameters when `check_list` is specified or
# no configuration was found
self.configure(check_list)
try:
command = command or JMX_COLLECT_COMMAND
if len(self.invalid_checks) > 0:
try:
JMXFiles.write_status_file(self.invalid_checks)
except Exception:
log.exception("Error while writing JMX status file")
if len(self.jmx_checks) > 0:
return self._start(self.java_bin_path, self.java_options, self.jmx_checks,
command, reporter, self.tools_jar_path)
else:
# We're exiting purposefully, so exit with zero (supervisor's expected
# code). HACK: Sleep a little bit so supervisor thinks we've started cleanly
# and thus can exit cleanly.
time.sleep(4)
log.info("No valid JMX integration was found. Exiting ...")
except Exception:
log.exception("Error while initiating JMXFetch")
raise
def configure(self, check_list=None):
"""
Instantiate JMXFetch parameters, clean potential previous run leftovers.
"""
JMXFiles.clean_status_file()
self.jmx_checks, self.invalid_checks, self.java_bin_path, self.java_options, self.tools_jar_path = \
self.get_configuration(check_list)
def terminate(self):
"""
Override `terminate` method to properly exit JMXFetch.
"""
JMXFiles.write_exit_file()
self.join()
def run(self):
if self.is_enabled:
JMXFiles.clean_exit_file()
self.jmx_daemon.run()
def _populate_payload_metadata(self, payload, check_statuses, start_event=True):
"""
Periodically populate the payload with metadata related to the system, host, and/or checks.
"""
now = time.time()
# Include system stats on first postback
if start_event and self._is_first_run():
payload['systemStats'] = self.agentConfig.get('system_stats', {})
# Also post an event in the newsfeed
payload['events']['System'] = [{
'api_key': self.agentConfig['api_key'],
'host': payload['internalHostname'],
'timestamp': now,
'event_type':'Agent Startup',
'msg_text': 'Version %s' % get_version()
}]
# Periodically send the host metadata.
if self._should_send_additional_data('host_metadata'):
# gather metadata with gohai
try:
if get_os() != 'windows':
command = "gohai"
else:
command = "gohai\gohai.exe"
gohai_metadata, gohai_log = subprocess.Popen(
[command], stdout=subprocess.PIPE, stderr=subprocess.PIPE
).communicate()
payload['gohai'] = gohai_metadata
if gohai_log:
log.warning("GOHAI LOG | {0}".format(gohai_log))
except OSError as e:
if e.errno == 2: # file not found, expected when install from source
log.info("gohai file not found")
else:
raise e
except Exception as e:
log.warning("gohai command failed with error %s" % str(e))
payload['systemStats'] = get_system_stats()
payload['meta'] = self._get_hostname_metadata()
self.hostname_metadata_cache = payload['meta']
# Add static tags from the configuration file
host_tags = []
if self.agentConfig['tags'] is not None:
host_tags.extend([unicode(tag.strip())
for tag in self.agentConfig['tags'].split(",")])
if self.agentConfig['collect_ec2_tags']:
host_tags.extend(EC2.get_tags(self.agentConfig))
if host_tags:
payload['host-tags']['system'] = host_tags
GCE_tags = GCE.get_tags(self.agentConfig)
if GCE_tags is not None:
payload['host-tags'][GCE.SOURCE_TYPE_NAME] = GCE_tags
# Log the metadata on the first run
if self._is_first_run():
log.info("Hostnames: %s, tags: %s" %
(repr(self.hostname_metadata_cache), payload['host-tags']))
# Periodically send extra hosts metadata (vsphere)
# Metadata of hosts that are not the host where the agent runs, not all the checks use
# that
external_host_tags = []
if self._should_send_additional_data('external_host_tags'):
for check in self.initialized_checks_d:
try:
getter = getattr(check, 'get_external_host_tags')
check_tags = getter()
external_host_tags.extend(check_tags)
except AttributeError:
pass
if external_host_tags:
payload['external_host_tags'] = external_host_tags
# Periodically send agent_checks metadata
if self._should_send_additional_data('agent_checks'):
# Add agent checks statuses and error/warning messages
agent_checks = []
for check in check_statuses:
if check.instance_statuses is not None:
for i, instance_status in enumerate(check.instance_statuses):
agent_checks.append(
(
check.name, check.source_type_name,
instance_status.instance_id,
instance_status.status,
# put error message or list of warning messages in the same field
# it will be handled by the UI
instance_status.error or instance_status.warnings or "",
check.service_metadata[i]
)
)
else:
agent_checks.append(
(
check.name, check.source_type_name,
"initialization",
check.status, repr(check.init_failed_error)
)
)
payload['agent_checks'] = agent_checks
payload['meta'] = self.hostname_metadata_cache # add hostname metadata
# If required by the user, let's create the dd_check:xxx host tags
if self.agentConfig['create_dd_check_tags'] and \
self._should_send_additional_data('dd_check_tags'):
app_tags_list = [DD_CHECK_TAG.format(c.name) for c in self.initialized_checks_d]
app_tags_list.extend([DD_CHECK_TAG.format(cname) for cname
in JMXFiles.get_jmx_appnames()])
if 'system' not in payload['host-tags']:
payload['host-tags']['system'] = []
payload['host-tags']['system'].extend(app_tags_list)
def get_jmx_status():
"""This function tries to read the 2 jmxfetch status file which are yaml file
located in the temp directory.
There are 2 files:
- One generated by the Agent itself, for jmx checks that can't be initialized because
there are missing stuff.
Its format is as following:
###
invalid_checks:
jmx: !!python/object/apply:jmxfetch.InvalidJMXConfiguration [You need to have at
least one instance defined in the YAML file for this check]
timestamp: 1391040927.136523
###
- One generated by jmxfetch that return information about the collection of metrics
its format is as following:
###
timestamp: 1391037347435
checks:
failed_checks:
jmx:
- {message: Unable to create instance. Please check your yaml file, status: ERROR}
initialized_checks:
tomcat:
- {message: null, status: OK, metric_count: 7, instance_name: jmx-remihakim.fr-3000}
###
"""
check_statuses = []
java_status_path = JMXFiles.get_status_file_path()
python_status_path = JMXFiles.get_python_status_file_path()
if not os.path.exists(java_status_path) and not os.path.exists(python_status_path):
log.debug("There is no jmx_status file at: %s or at: %s" % (java_status_path, python_status_path))
return []
check_data = defaultdict(lambda: defaultdict(list))
try:
if os.path.exists(java_status_path):
java_jmx_stats = yaml.load(file(java_status_path))
status_age = time.time() - java_jmx_stats.get('timestamp')/1000 # JMX timestamp is saved in milliseconds
jmx_checks = java_jmx_stats.get('checks', {})
if status_age > 60:
check_statuses.append(
CheckStatus("jmx", [
InstanceStatus(
0,
STATUS_ERROR,
error="JMXfetch didn't return any metrics during the last minute"
)
])
)
else:
for check_name, instances in jmx_checks.get('failed_checks', {}).iteritems():
for info in instances:
message = info.get('message', None)
metric_count = info.get('metric_count', 0)
service_check_count = info.get('service_check_count', 0)
status = info.get('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)
check_data[check_name]['service_check_count'].append(service_check_count)
for check_name, instances in jmx_checks.get('initialized_checks', {}).iteritems():
for info in instances:
message = info.get('message', None)
metric_count = info.get('metric_count', 0)
service_check_count = info.get('service_check_count', 0)
status = info.get('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)
check_data[check_name]['service_check_count'].append(service_check_count)
for check_name, data in check_data.iteritems():
check_status = CheckStatus(check_name, data['statuses'],
metric_count=sum(data['metric_count']),
service_check_count=sum(data['service_check_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, [], init_failed_error=excep))
return check_statuses
except Exception:
log.exception("Couldn't load latest jmx status")
return []
def _start(self, path_to_java, java_run_opts, jmx_checks, command, reporter, tools_jar_path):
statsd_port = self.agentConfig.get('dogstatsd_port', "8125")
if reporter is None:
reporter = "statsd:%s" % str(statsd_port)
log.info("Starting jmxfetch:")
try:
path_to_java = path_to_java or "java"
java_run_opts = java_run_opts or ""
path_to_jmxfetch = self._get_path_to_jmxfetch()
path_to_status_file = JMXFiles.get_status_file_path()
if tools_jar_path is None:
classpath = path_to_jmxfetch
else:
classpath = r"%s:%s" % (tools_jar_path, path_to_jmxfetch)
subprocess_args = [
path_to_java, # Path to the java bin
'-classpath',
classpath,
JMXFETCH_MAIN_CLASS,
'--check_period', str(self.check_frequency * 1000), # Period of the main loop of jmxfetch in ms
'--conf_directory', r"%s" % self.confd_path, # Path of the conf.d directory that will be read by jmxfetch,
'--log_level', JAVA_LOGGING_LEVEL.get(self.logging_config.get("log_level"), "INFO"), # Log Level: Mapping from Python log level to log4j log levels
'--log_location', r"%s" % self.logging_config.get('jmxfetch_log_file'), # Path of the log file
'--reporter', reporter, # Reporter to use
'--status_location', r"%s" % path_to_status_file, # Path to the status file to write
command, # Name of the command
]
if Platform.is_windows():
# Signal handlers are not supported on Windows:
# use a file to trigger JMXFetch exit instead
path_to_exit_file = JMXFiles.get_python_exit_file_path()
subprocess_args.insert(len(subprocess_args) - 1, '--exit_file_location')
subprocess_args.insert(len(subprocess_args) - 1, path_to_exit_file)
subprocess_args.insert(4, '--check')
for check in jmx_checks:
subprocess_args.insert(5, check)
# Specify a maximum memory allocation pool for the JVM
if "Xmx" not in java_run_opts and "XX:MaxHeapSize" not in java_run_opts:
java_run_opts += _JVM_DEFAULT_MAX_MEMORY_ALLOCATION
# Specify the initial memory allocation pool for the JVM
if "Xms" not in java_run_opts and "XX:InitialHeapSize" not in java_run_opts:
java_run_opts += _JVM_DEFAULT_INITIAL_MEMORY_ALLOCATION
for opt in java_run_opts.split():
subprocess_args.insert(1, opt)
log.info("Running %s" % " ".join(subprocess_args))
jmx_process = subprocess.Popen(subprocess_args, close_fds=True)
self.jmx_process = jmx_process
# Register SIGINT and SIGTERM signal handlers
self.register_signal_handlers()
# Wait for JMXFetch to return
jmx_process.wait()
return jmx_process.returncode
except OSError:
java_path_msg = "Couldn't launch JMXTerm. Is Java in your PATH ?"
log.exception(java_path_msg)
invalid_checks = {}
for check in jmx_checks:
check_name = check.split('.')[0]
check_name = check_name.encode('ascii', 'ignore')
invalid_checks[check_name] = java_path_msg
JMXFiles.write_status_file(invalid_checks)
raise
except Exception:
log.exception("Couldn't launch JMXFetch")
raise
