async def get_component_metrics(
tmanager,
component: str,
metric_names: List[str],
instances: List[str],
interval: int,
) -> ComponentMetrics:
"""
Return metrics from the Tmanager over the given interval.
The metrics property is keyed with (metric, instance) to metric values.
Metrics not included in `metric_names` will be truncated.
"""
if not (tmanager and tmanager.host and tmanager.stats_port):
raise Exception("No Tmanager found")
metric_request = tmanager_pb2.MetricRequest()
metric_request.component_name = component
if instances:
metric_request.instance_id.extend(instances)
metric_request.metric.extend(metric_names)
metric_request.interval = interval
url = f"http://{tmanager.host}:{tmanager.stats_port}/stats"
async with httpx.AsyncClient() as client:
response = await client.post(url,
data=metric_request.SerializeToString())
metric_response = tmanager_pb2.MetricResponse()
metric_response.ParseFromString(response.content)
if metric_response.status.status == common_pb2.NOTOK:
if metric_response.status.HasField("message"):
Log.warn("Received response from Tmanager: %s",
metric_response.status.message)
metrics = {}
for metric in metric_response.metric:
instance = metric.instance_id
for instance_metric in metric.metric:
metrics.setdefault(instance_metric.name,
{})[instance] = instance_metric.value
return ComponentMetrics(
interval=metric_response.interval,
component=component,
metrics=metrics,
)
def initializeFromRC(cls, rcfile):
if len(cls.cmdmap) > 0:
return
effective_rc = (rcfile, HERON_RC)[rcfile is None]
Log.debug('Effective RC file is %s', effective_rc)
if os.path.exists(effective_rc):
with open(effective_rc) as f:
cls.cmdmap['*']['*'] = collections.defaultdict(dict)
cls.cmdmap['*']['*']['*'] = ''
for line in f:
m = heron_command_pattern.match(line)
app, value, command, env = '', '', '', ''
if m is not None:
value = cls.remove_comments(
m.group(4).rstrip(os.linesep))
app = (m.group(1), '')[m.group(1) is None
or m.group(1) == '']
command = (m.group(2), '')[m.group(2) is None
or m.group(1) == '']
env = (m.group(3), '')[m.group(3) is None
or m.group(2) == '']
else:
continue
# make sure that all the single args have a boolean value
# associated so that we can load the args to a key value
# structure
args_list = config.insert_bool_values(value.split())
args_list_string = ' '.join(args_list)
if not command or not app or not env:
Log.warn(
"heronrc config entry %s does not have key parameters (command:app:env) ",
line)
continue
if app not in cls.cmdmap:
cls.cmdmap[app] = collections.defaultdict(dict)
if command in cls.cmdmap[app] and env in cls.cmdmap[app][
command]:
cls.cmdmap[app][command][env] = cls.cmdmap[app][
command][env] + ' ' + args_list_string
else:
cls.cmdmap[app][command][env] = args_list_string
Log.debug("RC cmdmap %s", json.dumps(cls.cmdmap))
else:
Log.debug("%s is not an existing file", effective_rc)
def initializeFromRC(cls, rcfile):
if len(cls.cmdmap) > 0:
return
effective_rc = (rcfile, HERON_RC)[rcfile is None]
Log.debug('Effective RC file is %s', effective_rc)
if os.path.exists(effective_rc):
with open(effective_rc) as f:
cls.cmdmap['*']['*'] = collections.defaultdict(dict)
cls.cmdmap['*']['*']['*'] = ''
for line in f:
m = heron_command_pattern.match(line)
app, value, command, env = '', '', '', ''
if m is not None:
value = cls.remove_comments(m.group(4).rstrip(os.linesep))
app = (m.group(1), '')[m.group(1) is None or m.group(1) == '']
command = (m.group(2), '')[m.group(2) is None or m.group(1) == '']
env = (m.group(3), '')[m.group(3) is None or m.group(2) == '']
else:
continue
# make sure that all the single args have a boolean value
# associated so that we can load the args to a key value
# structure
args_list = config.insert_bool_values(value.split())
args_list_string = ' '.join(args_list)
if not command or not app or not env:
Log.warn("heronrc config entry %s does not have key parameters (command:app:env) ",
line)
continue
if app not in cls.cmdmap:
cls.cmdmap[app] = collections.defaultdict(dict)
if command in cls.cmdmap[app] and env in cls.cmdmap[app][command]:
cls.cmdmap[app][command][env] = cls.cmdmap[app][command][env] + ' ' + args_list_string
else:
cls.cmdmap[app][command][env] = args_list_string
Log.debug("RC cmdmap %s", json.dumps(cls.cmdmap))
else:
Log.debug("%s is not an existing file", effective_rc)
def getComponentMetrics(self,
tmaster,
componentName,
metricNames,
instances,
interval,
callback=None):
"""
Get the specified metrics for the given component name of this topology.
Returns the following dict on success:
{
"metrics": {
<metricname>: {
<instance>: <numeric value>,
<instance>: <numeric value>,
...
}, ...
},
"interval": <numeric value>,
"component": "..."
}
Raises exception on failure.
"""
if not tmaster or not tmaster.host or not tmaster.stats_port:
raise Exception("No Tmaster found")
host = tmaster.host
port = tmaster.stats_port
metricRequest = tmaster_pb2.MetricRequest()
metricRequest.component_name = componentName
if len(instances) > 0:
for instance in instances:
metricRequest.instance_id.append(instance)
for metricName in metricNames:
metricRequest.metric.append(metricName)
metricRequest.interval = interval
# Serialize the metricRequest to send as a payload
# with the HTTP request.
metricRequestString = metricRequest.SerializeToString()
url = "http://{0}:{1}/stats".format(host, port)
request = tornado.httpclient.HTTPRequest(url,
body=metricRequestString,
method='POST',
request_timeout=5)
Log.debug("Making HTTP call to fetch metrics")
Log.debug("url: " + url)
try:
client = tornado.httpclient.AsyncHTTPClient()
result = yield client.fetch(request)
Log.debug("HTTP call complete.")
except tornado.httpclient.HTTPError as e:
raise Exception(str(e))
# Check the response code - error if it is in 400s or 500s
responseCode = result.code
if responseCode >= 400:
message = "Error in getting metrics from Tmaster, code: " + responseCode
Log.error(message)
raise Exception(message)
# Parse the response from tmaster.
metricResponse = tmaster_pb2.MetricResponse()
metricResponse.ParseFromString(result.body)
if metricResponse.status.status == common_pb2.NOTOK:
if metricResponse.status.HasField("message"):
Log.warn("Received response from Tmaster: %s", metricResponse.status.message)
# Form the response.
ret = {}
ret["interval"] = metricResponse.interval
ret["component"] = componentName
ret["metrics"] = {}
for metric in metricResponse.metric:
instance = metric.instance_id
for im in metric.metric:
metricname = im.name
value = im.value
if metricname not in ret["metrics"]:
ret["metrics"][metricname] = {}
ret["metrics"][metricname][instance] = value
raise tornado.gen.Return(ret)
def getMetricsTimeline(tmaster,
component_name,
metric_names,
instances,
start_time,
end_time,
callback=None):
"""
Get the specified metrics for the given component name of this topology.
Returns the following dict on success:
{
"timeline": {
<metricname>: {
<instance>: {
<start_time> : <numeric value>,
<start_time> : <numeric value>,
...
}
...
}, ...
},
"starttime": <numeric value>,
"endtime": <numeric value>,
"component": "..."
}
Returns the following dict on failure:
{
"message": "..."
}
"""
# Tmaster is the proto object and must have host and port for stats.
if not tmaster or not tmaster.host or not tmaster.stats_port:
raise Exception("No Tmaster found")
host = tmaster.host
port = tmaster.stats_port
# Create the proto request object to get metrics.
metricRequest = tmaster_pb2.MetricRequest()
metricRequest.component_name = component_name
# If no instances are give, metrics for all instances
# are fetched by default.
if len(instances) > 0:
for instance in instances:
metricRequest.instance_id.append(instance)
for metricName in metric_names:
metricRequest.metric.append(metricName)
metricRequest.explicit_interval.start = start_time
metricRequest.explicit_interval.end = end_time
metricRequest.minutely = True
# Serialize the metricRequest to send as a payload
# with the HTTP request.
metricRequestString = metricRequest.SerializeToString()
# Form and send the http request.
url = "http://{0}:{1}/stats".format(host, port)
request = tornado.httpclient.HTTPRequest(url,
body=metricRequestString,
method='POST',
request_timeout=5)
Log.debug("Making HTTP call to fetch metrics")
Log.debug("url: " + url)
try:
client = tornado.httpclient.AsyncHTTPClient()
result = yield client.fetch(request)
Log.debug("HTTP call complete.")
except tornado.httpclient.HTTPError as e:
raise Exception(str(e))
# Check the response code - error if it is in 400s or 500s
responseCode = result.code
if responseCode >= 400:
message = "Error in getting metrics from Tmaster, code: " + responseCode
Log.error(message)
raise Exception(message)
# Parse the response from tmaster.
metricResponse = tmaster_pb2.MetricResponse()
metricResponse.ParseFromString(result.body)
if metricResponse.status.status == common_pb2.NOTOK:
if metricResponse.status.HasField("message"):
Log.warn("Received response from Tmaster: %s", metricResponse.status.message)
# Form the response.
ret = {}
ret["starttime"] = start_time
ret["endtime"] = end_time
ret["component"] = component_name
ret["timeline"] = {}
# Loop through all the metrics
# One instance corresponds to one metric, which can have
# multiple IndividualMetrics for each metricname requested.
for metric in metricResponse.metric:
instance = metric.instance_id
# Loop through all individual metrics.
for im in metric.metric:
metricname = im.name
if metricname not in ret["timeline"]:
ret["timeline"][metricname] = {}
if instance not in ret["timeline"][metricname]:
ret["timeline"][metricname][instance] = {}
# We get minutely metrics.
# Interval-values correspond to the minutely mark for which
# this metric value corresponds to.
for interval_value in im.interval_values:
ret["timeline"][metricname][instance][interval_value.interval.start] = interval_value.value
raise tornado.gen.Return(ret)
def getComponentMetrics(self,
tmaster,
componentName,
metricNames,
instances,
interval,
callback=None):
"""
Get the specified metrics for the given component name of this topology.
Returns the following dict on success:
{
"metrics": {
<metricname>: {
<instance>: <numeric value>,
<instance>: <numeric value>,
...
}, ...
},
"interval": <numeric value>,
"component": "..."
}
Raises exception on failure.
"""
if not tmaster or not tmaster.host or not tmaster.stats_port:
raise Exception("No Tmaster found")
host = tmaster.host
port = tmaster.stats_port
metricRequest = tmaster_pb2.MetricRequest()
metricRequest.component_name = componentName
if len(instances) > 0:
for instance in instances:
metricRequest.instance_id.append(instance)
for metricName in metricNames:
metricRequest.metric.append(metricName)
metricRequest.interval = interval
# Serialize the metricRequest to send as a payload
# with the HTTP request.
metricRequestString = metricRequest.SerializeToString()
url = "http://{0}:{1}/stats".format(host, port)
request = tornado.httpclient.HTTPRequest(url,
body=metricRequestString,
method='POST',
request_timeout=5)
Log.debug("Making HTTP call to fetch metrics")
Log.debug("url: " + url)
try:
client = tornado.httpclient.AsyncHTTPClient()
result = yield client.fetch(request)
Log.debug("HTTP call complete.")
except tornado.httpclient.HTTPError as e:
raise Exception(str(e))
# Check the response code - error if it is in 400s or 500s
responseCode = result.code
if responseCode >= 400:
message = "Error in getting metrics from Tmaster, code: " + responseCode
Log.error(message)
raise Exception(message)
# Parse the response from tmaster.
metricResponse = tmaster_pb2.MetricResponse()
metricResponse.ParseFromString(result.body)
if metricResponse.status.status == common_pb2.NOTOK:
if metricResponse.status.HasField("message"):
Log.warn("Received response from Tmaster: %s",
metricResponse.status.message)
# Form the response.
ret = {}
ret["interval"] = metricResponse.interval
ret["component"] = componentName
ret["metrics"] = {}
for metric in metricResponse.metric:
instance = metric.instance_id
for im in metric.metric:
metricname = im.name
value = im.value
if metricname not in ret["metrics"]:
ret["metrics"][metricname] = {}
ret["metrics"][metricname][instance] = value
raise tornado.gen.Return(ret)
async def get_metrics_timeline(
tmanager: tmanager_pb2.TManagerLocation,
component_name: str,
metric_names: List[str],
instances: List[str],
start_time: int,
end_time: int,
callback=None,
) -> MetricsTimeline:
"""
Get the specified metrics for the given component name of this topology.
"""
# Tmanager is the proto object and must have host and port for stats.
if not tmanager or not tmanager.host or not tmanager.stats_port:
raise Exception("No Tmanager found")
# Create the proto request object to get metrics.
request_parameters = tmanager_pb2.MetricRequest()
request_parameters.component_name = component_name
# If no instances are given, metrics for all instances
# are fetched by default.
request_parameters.instance_id.extend(instances)
request_parameters.metric.extend(metric_names)
request_parameters.explicit_interval.start = start_time
request_parameters.explicit_interval.end = end_time
request_parameters.minutely = True
# Form and send the http request.
url = f"http://{tmanager.host}:{tmanager.stats_port}/stats"
Log.debug(f"Making HTTP call to fetch metrics: {url}")
async with httpx.AsyncClient() as client:
result = await client.post(url,
data=request_parameters.SerializeToString())
# Check the response code - error if it is in 400s or 500s
if result.status_code >= 400:
message = f"Error in getting metrics from Tmanager, code: {result.code}"
raise Exception(message)
# Parse the response from tmanager.
response_data = tmanager_pb2.MetricResponse()
response_data.ParseFromString(result.content)
if response_data.status.status == common_pb2.NOTOK:
if response_data.status.HasField("message"):
Log.warn("Received response from Tmanager: %s",
response_data.status.message)
timeline = {}
# Loop through all the metrics
# One instance corresponds to one metric, which can have
# multiple IndividualMetrics for each metricname requested.
for metric in response_data.metric:
instance = metric.instance_id
# Loop through all individual metrics.
for im in metric.metric:
metricname = im.name
if metricname not in timeline:
timeline[metricname] = {}
if instance not in timeline[metricname]:
timeline[metricname][instance] = {}
# We get minutely metrics.
# Interval-values correspond to the minutely mark for which
# this metric value corresponds to.
for interval_value in im.interval_values:
timeline[metricname][instance][
interval_value.interval.start] = interval_value.value
return MetricsTimeline(
starttime=start_time,
endtime=end_time,
component=component_name,
timeline=timeline,
)
def getMetricsTimeline(tmaster,
component_name,
metric_names,
instances,
start_time,
end_time,
callback=None):
"""
Get the specified metrics for the given component name of this topology.
Returns the following dict on success:
{
"timeline": {
<metricname>: {
<instance>: {
<start_time> : <numeric value>,
<start_time> : <numeric value>,
...
}
...
}, ...
},
"starttime": <numeric value>,
"endtime": <numeric value>,
"component": "..."
}
Returns the following dict on failure:
{
"message": "..."
}
"""
# Tmaster is the proto object and must have host and port for stats.
if not tmaster or not tmaster.host or not tmaster.stats_port:
raise Exception("No Tmaster found")
host = tmaster.host
port = tmaster.stats_port
# Create the proto request object to get metrics.
metricRequest = tmaster_pb2.MetricRequest()
metricRequest.component_name = component_name
# If no instances are give, metrics for all instances
# are fetched by default.
if len(instances) > 0:
for instance in instances:
metricRequest.instance_id.append(instance)
for metricName in metric_names:
metricRequest.metric.append(metricName)
metricRequest.explicit_interval.start = start_time
metricRequest.explicit_interval.end = end_time
metricRequest.minutely = True
# Serialize the metricRequest to send as a payload
# with the HTTP request.
metricRequestString = metricRequest.SerializeToString()
# Form and send the http request.
url = "http://{0}:{1}/stats".format(host, port)
request = tornado.httpclient.HTTPRequest(url,
body=metricRequestString,
method='POST',
request_timeout=5)
Log.debug("Making HTTP call to fetch metrics")
Log.debug("url: " + url)
try:
client = tornado.httpclient.AsyncHTTPClient()
result = yield client.fetch(request)
Log.debug("HTTP call complete.")
except tornado.httpclient.HTTPError as e:
raise Exception(str(e))
# Check the response code - error if it is in 400s or 500s
responseCode = result.code
if responseCode >= 400:
message = "Error in getting metrics from Tmaster, code: " + responseCode
Log.error(message)
raise Exception(message)
# Parse the response from tmaster.
metricResponse = tmaster_pb2.MetricResponse()
metricResponse.ParseFromString(result.body)
if metricResponse.status.status == common_pb2.NOTOK:
if metricResponse.status.HasField("message"):
Log.warn("Received response from Tmaster: %s",
metricResponse.status.message)
# Form the response.
ret = {}
ret["starttime"] = start_time
ret["endtime"] = end_time
ret["component"] = component_name
ret["timeline"] = {}
# Loop through all the metrics
# One instance corresponds to one metric, which can have
# multiple IndividualMetrics for each metricname requested.
for metric in metricResponse.metric:
instance = metric.instance_id
# Loop through all individual metrics.
for im in metric.metric:
metricname = im.name
if metricname not in ret["timeline"]:
ret["timeline"][metricname] = {}
if instance not in ret["timeline"][metricname]:
ret["timeline"][metricname][instance] = {}
# We get minutely metrics.
# Interval-values correspond to the minutely mark for which
# this metric value corresponds to.
for interval_value in im.interval_values:
ret["timeline"][metricname][instance][
interval_value.interval.start] = interval_value.value
raise tornado.gen.Return(ret)
