def annotate_machine_pu_util(internal_graph, node):
source = InfoGraphNode.get_machine_name_of_pu(node)
machine = InfoGraphNode.get_node(internal_graph, source)
machine_util = InfoGraphNode.get_compute_utilization(machine)
if 'intel/use/compute/utilization' not in machine_util.columns:
sum_util = None
cpu_metric = 'intel/procfs/cpu/utilization_percentage'
pu_util_df = InfoGraphNode.get_compute_utilization(node)
if cpu_metric in pu_util_df.columns:
pu_util = pu_util_df[cpu_metric]
pu_util = pu_util.fillna(0)
machine_util[InfoGraphNode.get_attributes(node)['name']] = pu_util
InfoGraphNode.set_compute_utilization(machine, machine_util)
else:
LOG.info('CPU util not Found use for node {}'.format(InfoGraphNode.get_name(node)))
else:
LOG.debug('Found use for node {}'.format(InfoGraphNode.get_name(node)))
def get_annotated_graph(self,
graph,
ts_from,
ts_to,
utilization=False,
saturation=False):
"""
Collect data from cimmaron tsdb in relation to the specified graph and
time windows and store an annotated subgraph in specified directory
:param graph: (NetworkX Graph) Graph to be annotated with data
:param ts_from: (str) Epoch time representation of start time
:param ts_to: (str) Epoch time representation of stop time
:param utilization: (bool) if True the method calculates also
utilization for each node, if available
:return: NetworkX Graph annotated with telemetry data
"""
TelemetryAnnotation._get_annotated_graph_input_validation(
graph, ts_from, ts_to)
internal_graph = graph.copy()
self.internal_graph = internal_graph
for node in internal_graph.nodes(data=True):
if isinstance(self.telemetry, SnapAnnotation):
queries = list()
try:
queries = self.telemetry.get_queries(
internal_graph, node, ts_from, ts_to)
# queries = self.telemetry.get_queries(graph, node, ts_from, ts_to)
except Exception as e:
LOG.error("Exception: {}".format(e))
LOG.error(e)
import traceback
traceback.print_exc()
if len(queries) != 0:
InfoGraphNode.set_queries(node, queries)
telemetry_data = self.telemetry.get_data(node)
InfoGraphNode.set_telemetry_data(node, telemetry_data)
if utilization and not telemetry_data.empty:
SnapUtils.utilization(internal_graph, node,
self.telemetry)
# if only procfs is available, results needs to be
# propagated at machine level
if InfoGraphNode.get_type(
node) == InfoGraphNodeType.PHYSICAL_PU:
SnapUtils.annotate_machine_pu_util(
internal_graph, node)
if InfoGraphNode.node_is_disk(node):
SnapUtils.annotate_machine_disk_util(
internal_graph, node)
if InfoGraphNode.node_is_nic(node):
SnapUtils.annotate_machine_network_util(
internal_graph, node)
if saturation:
SnapUtils.saturation(internal_graph, node,
self.telemetry)
elif isinstance(self.telemetry, PrometheusAnnotation):
queries = list()
try:
queries = self.telemetry.get_queries(
internal_graph, node, ts_from, ts_to)
# queries = self.telemetry.get_queries(graph, node, ts_from, ts_to)
except Exception as e:
LOG.error("Exception: {}".format(e))
LOG.error(e)
import traceback
traceback.print_exc()
if len(queries) != 0:
InfoGraphNode.set_queries(node, queries)
telemetry_data = self.telemetry.get_data(node)
InfoGraphNode.set_telemetry_data(node, telemetry_data)
# if utilization and not telemetry_data.empty:
#PrometheusUtils.utilization(internal_graph, node, self.telemetry)
# if only procfs is available, results needs to be
# propagated at machine level
#if InfoGraphNode.get_type(node) == InfoGraphNodeType.PHYSICAL_PU:
# PrometheusUtils.annotate_machine_pu_util(internal_graph, node)
#if InfoGraphNode.node_is_disk(node):
# PrometheusUtils.annotate_machine_disk_util(internal_graph, node)
#if InfoGraphNode.node_is_nic(node):
# PrometheusUtils.annotate_machine_network_util(internal_graph, node)
#if saturation:
#PrometheusUtils.saturation(internal_graph, node, self.telemetry)
else:
telemetry_data = self.telemetry.get_data(node)
InfoGraphNode.set_telemetry_data(node, telemetry_data)
if utilization and not telemetry_data.empty:
SnapUtils.utilization(internal_graph, node, self.telemetry)
# if only procfs is available, results needs to be
# propagated at machine level
if InfoGraphNode.get_type(
node) == InfoGraphNodeType.PHYSICAL_PU:
source = InfoGraphNode.get_machine_name_of_pu(node)
machine = InfoGraphNode.get_node(
internal_graph, source)
machine_util = InfoGraphNode.get_compute_utilization(
machine)
if '/intel/use/compute/utilization' not in machine_util.columns:
sum_util = None
pu_util = InfoGraphNode.get_compute_utilization(
node
)['intel/procfs/cpu/utilization_percentage']
pu_util = pu_util.fillna(0)
if 'intel/procfs/cpu/utilization_percentage' in machine_util.columns:
machine_util = machine_util[
'intel/procfs/cpu/utilization_percentage']
machine_util = machine_util.fillna(0)
sum_util = machine_util.add(pu_util,
fill_value=0)
else:
sum_util = pu_util
if isinstance(sum_util, pandas.Series):
# sum_util.index.name = None
sum_util = pandas.DataFrame(
sum_util,
columns=[
'intel/procfs/cpu/utilization_percentage'
])
InfoGraphNode.set_compute_utilization(
machine, sum_util)
else:
LOG.debug('Found use for node {}'.format(
InfoGraphNode.get_name(node)))
if saturation:
self._saturation(internal_graph, node, self.telemetry)
return internal_graph
def annotate_machine_pu_util(internal_graph, node):
source = InfoGraphNode.get_machine_name_of_pu(node)
machine = InfoGraphNode.get_node(internal_graph, source)
machine_util = InfoGraphNode.get_compute_utilization(machine)
InfoGraphNode.set_compute_utilization(machine, pandas.DataFrame())
def utilization(internal_graph, node, telemetry):
# machine usage
telemetry_data = telemetry.get_data(node)
if 'intel/use/compute/utilization' in telemetry_data:
InfoGraphNode.set_compute_utilization(node,
pandas.DataFrame(telemetry_data['intel/use/compute/utilization'],
columns=['intel/use/compute/utilization']))
# pu usage
if 'intel/procfs/cpu/utilization_percentage' in telemetry_data:
InfoGraphNode.set_compute_utilization(node,
pandas.DataFrame(
telemetry_data['intel/procfs/cpu/utilization_percentage'],
columns=['intel/procfs/cpu/utilization_percentage']))
if 'intel/use/memory/utilization' in telemetry_data:
InfoGraphNode.set_memory_utilization(node, pandas.DataFrame(telemetry_data['intel/use/memory/utilization']))
if 'intel/use/disk/utilization' in telemetry_data:
InfoGraphNode.set_disk_utilization(node, pandas.DataFrame(telemetry_data['intel/use/disk/utilization']))
if 'intel/use/network/utilization' in telemetry_data:
InfoGraphNode.set_network_utilization(node,
pandas.DataFrame(telemetry_data['intel/use/network/utilization']))
# supporting not available /use/ metrics
if 'intel/procfs/meminfo/mem_total' in telemetry_data and 'intel/procfs/meminfo/mem_used' in telemetry_data:
# LOG.info('Found memory utilization procfs')
mem_used = telemetry_data['intel/procfs/meminfo/mem_used'].fillna(0)
mem_total = telemetry_data['intel/procfs/meminfo/mem_total'].fillna(0)
mem_util = mem_used * 100 / mem_total
mem_util.name = 'intel/procfs/memory/utilization_percentage'
InfoGraphNode.set_memory_utilization(node, pandas.DataFrame(mem_util))
if 'intel/procfs/disk/io_time' in telemetry_data:
io_time = telemetry_data['intel/procfs/disk/io_time'].fillna(0)
disk_util = io_time*100/1000
disk_util.name = 'intel/procfs/disk/utilization_percentage'
InfoGraphNode.set_disk_utilization(node, pandas.DataFrame(disk_util))
if 'intel/psutil/net/bytes_recv' in telemetry_data and 'intel/psutil/net/bytes_sent' in telemetry_data:
source=telemetry._source(node)
machine = InfoGraphNode.get_node(internal_graph, source)
nic_speed = InfoGraphNode.get_nic_speed_mbps(machine) * 1000000
net_data = telemetry_data.filter(['timestamp', 'intel/psutil/net/bytes_recv','intel/psutil/net/bytes_sent'], axis=1)
net_data.fillna(0)
net_data['intel/psutil/net/bytes_total'] = net_data['intel/psutil/net/bytes_recv']+net_data['intel/psutil/net/bytes_sent']
net_data_interval = net_data.set_index('timestamp').diff()
net_data_interval['intel/psutil/net/utilization_percentage'] = net_data_interval['intel/psutil/net/bytes_total'] * 100 /nic_speed
net_data_pct = pandas.DataFrame(net_data_interval['intel/psutil/net/utilization_percentage'])
InfoGraphNode.set_network_utilization(node, net_data_pct)
elif 'intel/procfs/iface/bytes_recv' in telemetry_data and 'intel/procfs/iface/bytes_recv' in telemetry_data:
source=telemetry._source(node)
machine = InfoGraphNode.get_node(internal_graph, source)
nic_speed = InfoGraphNode.get_nic_speed_mbps(machine) * 1000000
net_data = telemetry_data.filter(['timestamp', 'intel/procfs/iface/bytes_recv','intel/procfs/iface/bytes_sent'], axis=1)
net_data.fillna(0)
net_data['intel/psutil/net/bytes_total'] = net_data['intel/procfs/iface/bytes_recv']+net_data['intel/procfs/iface/bytes_sent']
net_data_interval = net_data.set_index('timestamp').diff()
net_data_interval['intel/psutil/net/utilization_percentage'] = net_data_interval['intel/psutil/net/bytes_total'] * 100 /nic_speed
net_data_pct = pandas.DataFrame(net_data_interval['intel/psutil/net/utilization_percentage'])
InfoGraphNode.set_network_utilization(node, net_data_pct)
if 'intel/docker/stats/cgroups/cpu_stats/cpu_usage/total' in telemetry_data:
# Container node
#cpu util
cpu_data = telemetry_data.filter(['timestamp', 'intel/docker/stats/cgroups/cpu_stats/cpu_usage/total'], axis=1)
cpu_data_interval = cpu_data.set_index('timestamp').diff()
#util data in nanoseconds
cpu_data_interval['intel/docker/stats/cgroups/cpu_stats/cpu_usage/percentage'] = cpu_data_interval['intel/docker/stats/cgroups/cpu_stats/cpu_usage/total'] / 10000000
cpu_data_pct = pandas.DataFrame(cpu_data_interval['intel/docker/stats/cgroups/cpu_stats/cpu_usage/percentage'])
InfoGraphNode.set_compute_utilization(node, cpu_data_pct)
if "intel/docker/stats/cgroups/memory_stats/usage/usage" in telemetry_data:
#container mem util
source=telemetry._source(node)
machine = InfoGraphNode.get_node(internal_graph, source)
local_mem = int(InfoGraphNode.get_attributes(machine).get("local_memory"))
mem_data = telemetry_data.filter(['timestamp', "intel/docker/stats/cgroups/memory_stats/usage/usage"], axis=1)
mem_data["intel/docker/stats/cgroups/memory_stats/usage/percentage"] = mem_data["intel/docker/stats/cgroups/memory_stats/usage/usage"]/local_mem * 100
mem_data_pct = pandas.DataFrame(mem_data["intel/docker/stats/cgroups/memory_stats/usage/percentage"])
InfoGraphNode.set_memory_utilization(node, mem_data_pct)
if "intel/docker/stats/network/tx_bytes" in telemetry_data:
#container network util
source=telemetry._source(node)
machine = InfoGraphNode.get_node(internal_graph, source)
nic_speed = InfoGraphNode.get_nic_speed_mbps(machine) * 1000000
net_data = telemetry_data.filter(['timestamp', "intel/docker/stats/network/tx_bytes","intel/docker/stats/network/rx_bytes"], axis=1)
net_data.fillna(0)
net_data['intel/docker/stats/network/bytes_total'] = net_data["intel/docker/stats/network/tx_bytes"]+net_data["intel/docker/stats/network/rx_bytes"]
net_data_interval = net_data.set_index('timestamp').diff()
net_data_interval['intel/docker/stats/network/utilization_percentage'] = net_data_interval['intel/docker/stats/network/bytes_total'] * 100 /nic_speed
net_data_pct = pandas.DataFrame(net_data_interval['intel/docker/stats/network/utilization_percentage'])
InfoGraphNode.set_network_utilization(node, net_data_pct)
if "intel/docker/stats/cgroups/blkio_stats/io_time_recursive/value" in telemetry_data:
#container disk util
disk_data = telemetry_data.filter(['timestamp', "intel/docker/stats/cgroups/blkio_stats/io_time_recursive/value"], axis=1)
disk_data_interval = disk_data.set_index('timestamp').diff()
#util data in milliseconds
disk_data_interval["intel/docker/stats/cgroups/blkio_stats/io_time_recursive/percentage"] = \
disk_data_interval["intel/docker/stats/cgroups/blkio_stats/io_time_recursive/value"] / 1000000
disk_data_pct = pandas.DataFrame(disk_data_interval["intel/docker/stats/cgroups/blkio_stats/io_time_recursive/percentage"])
InfoGraphNode.set_disk_utilization(node, disk_data_pct)
