def test_credentials_token_noverify():
expected_headers = {'Authorization': 'Bearer mytoken'}
creds = KubeletCredentials({
"verify_tls": "false",
"ca_cert": "ca_cert",
"client_crt": "ignore_me",
"token": "mytoken"
})
assert creds.verify() is False
assert creds.cert_pair() is None
assert creds.headers("https://dummy") == expected_headers
# Make sure we don't leak the token over http
assert creds.headers("http://dummy") is None
instance = {'prometheus_url': 'https://dummy', 'namespace': 'foo'}
scraper = OpenMetricsBaseCheck('prometheus', {}, [instance])
scraper_config = scraper.create_scraper_configuration(instance)
creds.configure_scraper(scraper_config)
assert scraper_config['ssl_ca_cert'] is False
assert scraper_config['ssl_cert'] is None
assert scraper_config['ssl_private_key'] is None
assert scraper_config['extra_headers'] == expected_headers
# Make sure we don't leak the token over http
scraper_config['prometheus_url'] = "http://dummy"
creds.configure_scraper(scraper_config)
assert scraper_config['ssl_ca_cert'] is False
assert scraper_config['ssl_cert'] is None
assert scraper_config['ssl_private_key'] is None
assert scraper_config['extra_headers'] == {}
def test_perform_kubelet_check(monkeypatch):
check = KubeletCheck('kubelet', {}, [{}])
check.kube_health_url = "http://127.0.0.1:10255/healthz"
check.kubelet_credentials = KubeletCredentials({})
monkeypatch.setattr(check, 'service_check', mock.Mock())
instance_tags = ["one:1"]
get = MockResponse()
with mock.patch("requests.get", side_effect=get):
check._perform_kubelet_check(instance_tags)
get.assert_has_calls([
mock.call(
'http://127.0.0.1:10255/healthz',
auth=None,
cert=None,
headers=None,
params={'verbose': True},
proxies=None,
stream=False,
timeout=(10.0, 10.0),
verify=None,
)
])
calls = [mock.call('kubernetes.kubelet.check', 0, tags=instance_tags)]
check.service_check.assert_has_calls(calls)
def check(self, _):
kubelet_conn_info = get_connection_info()
endpoint = kubelet_conn_info.get('url')
if endpoint is None:
raise CheckException(
"Unable to detect the kubelet URL automatically: " +
kubelet_conn_info.get('err', ''))
self.pod_list_url = endpoint.strip("/") + POD_LIST_PATH
self.kubelet_credentials = KubeletCredentials(kubelet_conn_info)
if self.fargate_mode:
pod_list = self.retrieve_pod_list()
for pod in pod_list.get('items', []):
pod_id = pod.get('metadata', {}).get('uid')
tagger_tags = tagger.tag('kubernetes_pod_uid://%s' % pod_id,
tagger.ORCHESTRATOR) or []
tagger_tags.extend(self.tags)
tags = set(tagger_tags)
# Submit the heartbeat metric for fargate virtual nodes.
self.gauge(self.NAMESPACE + '.pods.running', 1, tags)
pod_annotations = pod.get('metadata', {}).get('annotations')
if CAPACITY_ANNOTATION_KEY not in pod_annotations:
continue
cpu_val, mem_val = extract_resource_values(
pod_annotations.get(CAPACITY_ANNOTATION_KEY))
if cpu_val == 0 or mem_val == 0:
continue
self.gauge(self.NAMESPACE + '.cpu.capacity', cpu_val, tags)
self.gauge(self.NAMESPACE + '.memory.capacity', mem_val, tags)
def test_silent_tls_warning(caplog, monkeypatch, aggregator):
check = KubeletCheck('kubelet', {}, [{}])
check.kube_health_url = "https://example.com/"
check.kubelet_credentials = KubeletCredentials({'verify_tls': 'false'})
with caplog.at_level(logging.DEBUG):
check._perform_kubelet_check([])
expected_message = 'An unverified HTTPS request is being made to https://example.com/'
for _, _, message in caplog.record_tuples:
assert message != expected_message
def test_report_node_metrics_kubernetes1_18(monkeypatch, aggregator):
check = KubeletCheck('kubelet', {}, [{}])
check.kubelet_credentials = KubeletCredentials({'verify_tls': 'false'})
check.node_spec_url = "http://localhost:10255/spec"
get = mock.MagicMock(status_code=404,
iter_lines=lambda **kwargs: "Error Code")
get.raise_for_status.side_effect = requests.HTTPError('error')
with mock.patch('requests.get', return_value=get):
check._report_node_metrics(['foo:bar'])
aggregator.assert_all_metrics_covered()
def test_credentials_empty():
creds = KubeletCredentials({})
assert creds.verify() is None
assert creds.cert_pair() is None
assert creds.headers("https://dummy") is None
instance = {'prometheus_url': 'https://dummy', 'namespace': 'foo'}
scraper = OpenMetricsBaseCheck('prometheus', {}, [instance])
scraper_config = scraper.create_scraper_configuration(instance)
creds.configure_scraper(scraper_config)
assert scraper_config['ssl_ca_cert'] is None
assert scraper_config['ssl_cert'] is None
assert scraper_config['ssl_private_key'] is None
assert scraper_config['extra_headers'] == {}
def test_credentials_certificates():
creds = KubeletCredentials({
"verify_tls": "true",
"ca_cert": "ca_cert",
"client_crt": "crt",
"client_key": "key",
"token": "ignore_me"
})
assert creds.verify() == "ca_cert"
assert creds.cert_pair() == ("crt", "key")
assert creds.headers("https://dummy") is None
instance = {'prometheus_url': 'https://dummy', 'namespace': 'foo'}
scraper = OpenMetricsBaseCheck('prometheus', {}, [instance])
scraper_config = scraper.create_scraper_configuration(instance)
creds.configure_scraper(scraper_config)
assert scraper_config['ssl_ca_cert'] == "ca_cert"
assert scraper_config['ssl_cert'] == "crt"
assert scraper_config['ssl_private_key'] == "key"
assert scraper_config['extra_headers'] == {}
def check(self, instance):
# Kubelet credential defaults are determined dynamically during every
# check run so we must make sure that configuration is always reset
self.reset_http_config()
kubelet_conn_info = get_connection_info()
endpoint = kubelet_conn_info.get('url')
if endpoint is None:
raise CheckException(
"Unable to detect the kubelet URL automatically: " +
kubelet_conn_info.get('err', ''))
self.kube_health_url = urljoin(endpoint, KUBELET_HEALTH_PATH)
self.node_spec_url = urljoin(endpoint, NODE_SPEC_PATH)
self.pod_list_url = urljoin(endpoint, POD_LIST_PATH)
self.stats_url = urljoin(endpoint, STATS_PATH)
self.instance_tags = instance.get('tags', [])
self.kubelet_credentials = KubeletCredentials(kubelet_conn_info)
# Test the kubelet health ASAP
self._perform_kubelet_check(self.instance_tags)
if 'cadvisor_metrics_endpoint' in instance:
self.cadvisor_scraper_config['prometheus_url'] = instance.get(
'cadvisor_metrics_endpoint',
urljoin(endpoint, CADVISOR_METRICS_PATH))
else:
self.cadvisor_scraper_config['prometheus_url'] = instance.get(
'metrics_endpoint', urljoin(endpoint, CADVISOR_METRICS_PATH))
if 'metrics_endpoint' in instance:
self.log.warning(
'metrics_endpoint is deprecated, please specify cadvisor_metrics_endpoint instead.'
)
self.kubelet_scraper_config['prometheus_url'] = instance.get(
'kubelet_metrics_endpoint', urljoin(endpoint,
KUBELET_METRICS_PATH))
# Kubelet credentials handling
self.kubelet_credentials.configure_scraper(
self.cadvisor_scraper_config)
self.kubelet_credentials.configure_scraper(self.kubelet_scraper_config)
# Legacy cadvisor support
try:
self.cadvisor_legacy_url = self.detect_cadvisor(
endpoint, self.cadvisor_legacy_port)
except Exception as e:
self.log.debug(
'cAdvisor not found, running in prometheus mode: %s', e)
self.pod_list = self.retrieve_pod_list()
self.pod_list_utils = PodListUtils(self.pod_list)
self.pod_tags_by_pvc = self._create_pod_tags_by_pvc(self.pod_list)
self._report_node_metrics(self.instance_tags)
self._report_pods_running(self.pod_list, self.instance_tags)
self._report_container_spec_metrics(self.pod_list, self.instance_tags)
self._report_container_state_metrics(self.pod_list, self.instance_tags)
self.stats = self._retrieve_stats()
self.process_stats_summary(self.pod_list_utils, self.stats,
self.instance_tags,
self.use_stats_summary_as_source)
if self.cadvisor_legacy_url: # Legacy cAdvisor
self.log.debug('processing legacy cadvisor metrics')
self.process_cadvisor(instance, self.cadvisor_legacy_url,
self.pod_list, self.pod_list_utils)
elif self.cadvisor_scraper_config['prometheus_url']: # Prometheus
self.log.debug('processing cadvisor metrics')
self.process(self.cadvisor_scraper_config,
metric_transformers=self.transformers)
if self.kubelet_scraper_config['prometheus_url']: # Prometheus
self.log.debug('processing kubelet metrics')
self.process(self.kubelet_scraper_config,
metric_transformers=self.transformers)
# Free up memory
self.pod_list = None
self.pod_list_utils = None
class KubeletCheck(
CadvisorPrometheusScraperMixin,
OpenMetricsBaseCheck,
CadvisorScraper,
SummaryScraperMixin,
KubeletBase,
):
"""
Collect metrics from Kubelet.
"""
DEFAULT_METRIC_LIMIT = 0
COUNTER_METRICS = {'kubelet_evictions': 'kubelet.evictions'}
VOLUME_METRICS = {
'kubelet_volume_stats_available_bytes':
'kubelet.volume.stats.available_bytes',
'kubelet_volume_stats_capacity_bytes':
'kubelet.volume.stats.capacity_bytes',
'kubelet_volume_stats_used_bytes': 'kubelet.volume.stats.used_bytes',
'kubelet_volume_stats_inodes': 'kubelet.volume.stats.inodes',
'kubelet_volume_stats_inodes_free': 'kubelet.volume.stats.inodes_free',
'kubelet_volume_stats_inodes_used': 'kubelet.volume.stats.inodes_used',
}
VOLUME_TAG_KEYS_TO_EXCLUDE = ['persistentvolumeclaim', 'pod_phase']
def __init__(self, name, init_config, instances):
self.NAMESPACE = 'kubernetes'
if instances is not None and len(instances) > 1:
raise Exception(
'Kubelet check only supports one configured instance.')
inst = instances[0] if instances else None
cadvisor_instance = self._create_cadvisor_prometheus_instance(inst)
kubelet_instance = self._create_kubelet_prometheus_instance(inst)
generic_instances = [cadvisor_instance, kubelet_instance]
super(KubeletCheck, self).__init__(name, init_config,
generic_instances)
self.cadvisor_legacy_port = inst.get('cadvisor_port',
CADVISOR_DEFAULT_PORT)
self.cadvisor_legacy_url = None
self.use_stats_summary_as_source = inst.get(
'use_stats_summary_as_source')
if self.use_stats_summary_as_source is None and sys.platform == 'win32':
self.use_stats_summary_as_source = True
self.cadvisor_scraper_config = self.get_scraper_config(
cadvisor_instance)
# Filter out system slices (empty pod name) to reduce memory footprint
self.cadvisor_scraper_config['_text_filter_blacklist'] = [
'pod_name=""', 'pod=""'
]
self.kubelet_scraper_config = self.get_scraper_config(kubelet_instance)
counter_transformers = {
k: self.send_always_counter
for k in self.COUNTER_METRICS
}
histogram_transformers = {
k: self._histogram_from_seconds_to_microseconds(v)
for k, v in TRANSFORM_VALUE_HISTOGRAMS.items()
}
volume_metric_transformers = {
k: self.append_pod_tags_to_volume_metrics
for k in self.VOLUME_METRICS
}
self.transformers = {}
for d in [
self.CADVISOR_METRIC_TRANSFORMERS,
counter_transformers,
histogram_transformers,
volume_metric_transformers,
]:
self.transformers.update(d)
def _create_kubelet_prometheus_instance(self, instance):
"""
Create a copy of the instance and set default values.
This is so the base class can create a scraper_config with the proper values.
"""
kubelet_instance = deepcopy(instance)
kubelet_instance.update({
'namespace':
self.NAMESPACE,
# We need to specify a prometheus_url so the base class can use it as the key for our config_map,
# we specify a dummy url that will be replaced in the `check()` function. We append it with "kubelet"
# so the key is different than the cadvisor scraper.
'prometheus_url':
instance.get('kubelet_metrics_endpoint', 'dummy_url/kubelet'),
'metrics': [
DEFAULT_GAUGES,
DEPRECATED_GAUGES,
NEW_1_14_GAUGES,
DEFAULT_HISTOGRAMS,
DEPRECATED_HISTOGRAMS,
NEW_1_14_HISTOGRAMS,
DEFAULT_SUMMARIES,
DEPRECATED_SUMMARIES,
NEW_1_14_SUMMARIES,
],
# Defaults that were set when the Kubelet scraper was based on PrometheusScraper
'send_monotonic_counter':
instance.get('send_monotonic_counter', False),
'health_service_check':
instance.get('health_service_check', False),
})
return kubelet_instance
def _create_pod_tags_by_pvc(self, pod_list):
"""
Return a map, e.g.
{
"<kube_namespace>/<persistentvolumeclaim>": [<list_of_pod_tags>],
"<kube_namespace1>/<persistentvolumeclaim1>": [<list_of_pod_tags1>],
}
that can be used to add pod tags to associated volume metrics
"""
pod_tags_by_pvc = defaultdict(set)
pods = pod_list.get('items', [])
for pod in pods:
# get kubernetes namespace of PVC
kube_ns = pod.get('metadata', {}).get('namespace')
if not kube_ns:
continue
# get volumes
volumes = pod.get('spec', {}).get('volumes')
if not volumes:
continue
# get pod id
pod_id = pod.get('metadata', {}).get('uid')
if not pod_id:
self.log.debug('skipping pod with no uid')
continue
# get tags from tagger
tags = tagger.tag('kubernetes_pod_uid://%s' % pod_id,
tagger.ORCHESTRATOR) or None
if not tags:
continue
# remove tags that don't apply to PVCs
for excluded_tag in self.VOLUME_TAG_KEYS_TO_EXCLUDE:
tags = [
t for t in tags if not t.startswith(excluded_tag + ':')
]
# get PVC
for v in volumes:
pvc_name = v.get('persistentVolumeClaim', {}).get('claimName')
if pvc_name:
pod_tags_by_pvc['{}/{}'.format(kube_ns,
pvc_name)].update(tags)
return pod_tags_by_pvc
def check(self, instance):
# Kubelet credential defaults are determined dynamically during every
# check run so we must make sure that configuration is always reset
self.reset_http_config()
kubelet_conn_info = get_connection_info()
endpoint = kubelet_conn_info.get('url')
if endpoint is None:
raise CheckException(
"Unable to detect the kubelet URL automatically: " +
kubelet_conn_info.get('err', ''))
self.kube_health_url = urljoin(endpoint, KUBELET_HEALTH_PATH)
self.node_spec_url = urljoin(endpoint, NODE_SPEC_PATH)
self.pod_list_url = urljoin(endpoint, POD_LIST_PATH)
self.stats_url = urljoin(endpoint, STATS_PATH)
self.instance_tags = instance.get('tags', [])
self.kubelet_credentials = KubeletCredentials(kubelet_conn_info)
# Test the kubelet health ASAP
self._perform_kubelet_check(self.instance_tags)
if 'cadvisor_metrics_endpoint' in instance:
self.cadvisor_scraper_config['prometheus_url'] = instance.get(
'cadvisor_metrics_endpoint',
urljoin(endpoint, CADVISOR_METRICS_PATH))
else:
self.cadvisor_scraper_config['prometheus_url'] = instance.get(
'metrics_endpoint', urljoin(endpoint, CADVISOR_METRICS_PATH))
if 'metrics_endpoint' in instance:
self.log.warning(
'metrics_endpoint is deprecated, please specify cadvisor_metrics_endpoint instead.'
)
self.kubelet_scraper_config['prometheus_url'] = instance.get(
'kubelet_metrics_endpoint', urljoin(endpoint,
KUBELET_METRICS_PATH))
# Kubelet credentials handling
self.kubelet_credentials.configure_scraper(
self.cadvisor_scraper_config)
self.kubelet_credentials.configure_scraper(self.kubelet_scraper_config)
# Legacy cadvisor support
try:
self.cadvisor_legacy_url = self.detect_cadvisor(
endpoint, self.cadvisor_legacy_port)
except Exception as e:
self.log.debug(
'cAdvisor not found, running in prometheus mode: %s', e)
self.pod_list = self.retrieve_pod_list()
self.pod_list_utils = PodListUtils(self.pod_list)
self.pod_tags_by_pvc = self._create_pod_tags_by_pvc(self.pod_list)
self._report_node_metrics(self.instance_tags)
self._report_pods_running(self.pod_list, self.instance_tags)
self._report_container_spec_metrics(self.pod_list, self.instance_tags)
self._report_container_state_metrics(self.pod_list, self.instance_tags)
self.stats = self._retrieve_stats()
self.process_stats_summary(self.pod_list_utils, self.stats,
self.instance_tags,
self.use_stats_summary_as_source)
if self.cadvisor_legacy_url: # Legacy cAdvisor
self.log.debug('processing legacy cadvisor metrics')
self.process_cadvisor(instance, self.cadvisor_legacy_url,
self.pod_list, self.pod_list_utils)
elif self.cadvisor_scraper_config['prometheus_url']: # Prometheus
self.log.debug('processing cadvisor metrics')
self.process(self.cadvisor_scraper_config,
metric_transformers=self.transformers)
if self.kubelet_scraper_config['prometheus_url']: # Prometheus
self.log.debug('processing kubelet metrics')
self.process(self.kubelet_scraper_config,
metric_transformers=self.transformers)
# Free up memory
self.pod_list = None
self.pod_list_utils = None
def _retrieve_node_spec(self):
"""
Retrieve node spec from kubelet.
"""
node_resp = self.perform_kubelet_query(self.node_spec_url)
return node_resp
def _retrieve_stats(self):
"""
Retrieve stats from kubelet.
"""
try:
stats_response = self.perform_kubelet_query(self.stats_url)
stats_response.raise_for_status()
return stats_response.json()
except Exception as e:
self.log.warning('GET on kubelet s `/stats/summary` failed: %s', e)
return {}
def _report_node_metrics(self, instance_tags):
try:
node_resp = self._retrieve_node_spec()
node_resp.raise_for_status()
except requests.HTTPError as e:
if node_resp.status_code == 404:
# ignore HTTPError, for supporting k8s >= 1.18 in a degrated mode
# in 1.18 the /spec can be reactivated from the kubelet config
# in 1.19 the /spec will removed.
return
raise e
node_spec = node_resp.json()
num_cores = node_spec.get('num_cores', 0)
memory_capacity = node_spec.get('memory_capacity', 0)
tags = instance_tags
self.gauge(self.NAMESPACE + '.cpu.capacity', float(num_cores), tags)
self.gauge(self.NAMESPACE + '.memory.capacity', float(memory_capacity),
tags)
def _perform_kubelet_check(self, instance_tags):
"""Runs local service checks"""
service_check_base = self.NAMESPACE + '.kubelet.check'
is_ok = True
url = self.kube_health_url
try:
req = self.perform_kubelet_query(url)
for line in req.iter_lines(decode_unicode=True):
# avoid noise; this check is expected to fail since we override the container hostname
if line.find('hostname') != -1:
continue
matches = re.match(r'\[(.)\]([^\s]+) (.*)?', line)
if not matches or len(matches.groups()) < 2:
continue
service_check_name = service_check_base + '.' + matches.group(
2)
status = matches.group(1)
if status == '+':
self.service_check(service_check_name,
AgentCheck.OK,
tags=instance_tags)
else:
self.service_check(service_check_name,
AgentCheck.CRITICAL,
tags=instance_tags)
is_ok = False
except Exception as e:
self.log.warning('kubelet check %s failed: %s', url, e)
self.service_check(
service_check_base,
AgentCheck.CRITICAL,
message='Kubelet check %s failed: %s' % (url, str(e)),
tags=instance_tags,
)
else:
if is_ok:
self.service_check(service_check_base,
AgentCheck.OK,
tags=instance_tags)
else:
self.service_check(service_check_base,
AgentCheck.CRITICAL,
tags=instance_tags)
def _report_pods_running(self, pods, instance_tags):
"""
Reports the number of running pods on this node and the running
containers in pods, tagged by service and creator.
:param pods: pod list object
:param instance_tags: list of tags
"""
pods_tag_counter = defaultdict(int)
containers_tag_counter = defaultdict(int)
for pod in pods.get('items', []):
# Containers reporting
containers = pod.get('status', {}).get('containerStatuses', [])
has_container_running = False
for container in containers:
container_id = container.get('containerID')
if not container_id:
self.log.debug('skipping container with no id')
continue
if "running" not in container.get('state', {}):
continue
has_container_running = True
tags = tagger.tag(replace_container_rt_prefix(container_id),
tagger.LOW) or None
if not tags:
continue
tags += instance_tags
hash_tags = tuple(sorted(tags))
containers_tag_counter[hash_tags] += 1
# Pod reporting
if not has_container_running:
continue
pod_id = pod.get('metadata', {}).get('uid')
if not pod_id:
self.log.debug('skipping pod with no uid')
continue
tags = tagger.tag('kubernetes_pod_uid://%s' % pod_id,
tagger.LOW) or None
if not tags:
continue
tags += instance_tags
hash_tags = tuple(sorted(tags))
pods_tag_counter[hash_tags] += 1
for tags, count in iteritems(pods_tag_counter):
self.gauge(self.NAMESPACE + '.pods.running', count, list(tags))
for tags, count in iteritems(containers_tag_counter):
self.gauge(self.NAMESPACE + '.containers.running', count,
list(tags))
def _report_container_spec_metrics(self, pod_list, instance_tags):
"""Reports pod requests & limits by looking at pod specs."""
for pod in pod_list.get('items', []):
pod_name = pod.get('metadata', {}).get('name')
pod_phase = pod.get('status', {}).get('phase')
if self._should_ignore_pod(pod_name, pod_phase):
continue
for ctr in pod['spec']['containers']:
if not ctr.get('resources'):
continue
c_name = ctr.get('name', '')
cid = None
for ctr_status in pod['status'].get('containerStatuses', []):
if ctr_status.get('name') == c_name:
# it is already prefixed with 'runtime://'
cid = ctr_status.get('containerID')
break
if not cid:
continue
pod_uid = pod.get('metadata', {}).get('uid')
if self.pod_list_utils.is_excluded(cid, pod_uid):
continue
tags = tagger.tag(replace_container_rt_prefix(cid),
tagger.HIGH)
if not tags:
continue
tags += instance_tags
try:
for resource, value_str in iteritems(
ctr.get('resources', {}).get('requests', {})):
value = self.parse_quantity(value_str)
self.gauge(
'{}.{}.requests'.format(self.NAMESPACE, resource),
value, tags)
except (KeyError, AttributeError) as e:
self.log.debug(
"Unable to retrieve container requests for %s: %s",
c_name, e)
try:
for resource, value_str in iteritems(
ctr.get('resources', {}).get('limits', {})):
value = self.parse_quantity(value_str)
self.gauge(
'{}.{}.limits'.format(self.NAMESPACE, resource),
value, tags)
except (KeyError, AttributeError) as e:
self.log.debug(
"Unable to retrieve container limits for %s: %s",
c_name, e)
def _report_container_state_metrics(self, pod_list, instance_tags):
"""Reports container state & reasons by looking at container statuses"""
if pod_list.get('expired_count'):
self.gauge(self.NAMESPACE + '.pods.expired',
pod_list.get('expired_count'),
tags=instance_tags)
for pod in pod_list.get('items', []):
pod_name = pod.get('metadata', {}).get('name')
pod_uid = pod.get('metadata', {}).get('uid')
if not pod_name or not pod_uid:
continue
for ctr_status in pod['status'].get('containerStatuses', []):
c_name = ctr_status.get('name')
cid = ctr_status.get('containerID')
if not c_name or not cid:
continue
if self.pod_list_utils.is_excluded(cid, pod_uid):
continue
tags = tagger.tag(replace_container_rt_prefix(cid),
tagger.ORCHESTRATOR)
if not tags:
continue
tags += instance_tags
restart_count = ctr_status.get('restartCount', 0)
self.gauge(self.NAMESPACE + '.containers.restarts',
restart_count, tags)
for (metric_name, field_name) in [('state', 'state'),
('last_state', 'lastState')]:
c_state = ctr_status.get(field_name, {})
for state_name in ['terminated', 'waiting']:
state_reasons = WHITELISTED_CONTAINER_STATE_REASONS.get(
state_name, [])
self._submit_container_state_metric(
metric_name, state_name, c_state, state_reasons,
tags)
def _submit_container_state_metric(self, metric_name, state_name, c_state,
state_reasons, tags):
reason_tags = []
state_value = c_state.get(state_name)
if state_value:
reason = state_value.get('reason', '')
if reason.lower() in state_reasons:
reason_tags.append('reason:%s' % (reason))
else:
return
gauge_name = '{}.containers.{}.{}'.format(self.NAMESPACE,
metric_name, state_name)
self.gauge(gauge_name, 1, tags + reason_tags)
@staticmethod
def parse_quantity(string):
"""
Parse quantity allows to convert the value in the resources spec like:
resources:
requests:
cpu: "100m"
memory": "200Mi"
limits:
memory: "300Mi"
:param string: str
:return: float
"""
number, unit = '', ''
for char in string:
if char.isdigit() or char == '.':
number += char
else:
unit += char
return float(number) * FACTORS.get(unit, 1)
@staticmethod
def _should_ignore_pod(name, phase):
"""
Pods that are neither pending or running should not be counted
in resource requests and limits.
"""
if not name or phase not in ["Running", "Pending"]:
return True
return False
def send_always_counter(self, metric, scraper_config, hostname=None):
metric_name_with_namespace = '{}.{}'.format(
scraper_config['namespace'], self.COUNTER_METRICS[metric.name])
for sample in metric.samples:
val = sample[self.SAMPLE_VALUE]
if not self._is_value_valid(val):
self.log.debug("Metric value is not supported for metric %s",
sample[self.SAMPLE_NAME])
continue
custom_hostname = self._get_hostname(hostname, sample,
scraper_config)
# Determine the tags to send
tags = self._metric_tags(metric.name,
val,
sample,
scraper_config,
hostname=custom_hostname)
self.monotonic_count(metric_name_with_namespace,
val,
tags=tags,
hostname=custom_hostname)
def append_pod_tags_to_volume_metrics(self,
metric,
scraper_config,
hostname=None):
metric_name_with_namespace = '{}.{}'.format(
scraper_config['namespace'], self.VOLUME_METRICS[metric.name])
for sample in metric.samples:
val = sample[self.SAMPLE_VALUE]
if not self._is_value_valid(val):
self.log.debug("Metric value is not supported for metric %s",
sample[self.SAMPLE_NAME])
continue
custom_hostname = self._get_hostname(hostname, sample,
scraper_config)
# Determine the tags to send
tags = self._metric_tags(metric.name,
val,
sample,
scraper_config,
hostname=custom_hostname)
pvc_name, kube_ns = None, None
for label_name, label_value in iteritems(
sample[self.SAMPLE_LABELS]):
if label_name == "persistentvolumeclaim":
pvc_name = label_value
elif label_name == "namespace":
kube_ns = label_value
if pvc_name and kube_ns:
break
pod_tags = self.pod_tags_by_pvc.get(
'{}/{}'.format(kube_ns, pvc_name), {})
tags.extend(pod_tags)
self.gauge(metric_name_with_namespace,
val,
tags=list(set(tags)),
hostname=custom_hostname)