osm_mon/collector/vnf_collectors/openstack.py

   1 # Copyright 2018 Whitestack, LLC
   2 # *************************************************************
   3
   4 # This file is part of OSM Monitoring module
   5 # All Rights Reserved to Whitestack, LLC
   6
   7 # Licensed under the Apache License, Version 2.0 (the "License"); you may
   8 # not use this file except in compliance with the License. You may obtain
   9 # a copy of the License at
  10
  11 #         http://www.apache.org/licenses/LICENSE-2.0
  12
  13 # Unless required by applicable law or agreed to in writing, software
  14 # distributed under the License is distributed on an "AS IS" BASIS, WITHOUT
  15 # WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the
  16 # License for the specific language governing permissions and limitations
  17 # under the License.
  18
  19 # For those usages not covered by the Apache License, Version 2.0 please
  20 # contact: bdiaz@whitestack.com or glavado@whitestack.com
  21 ##
  22 from enum import Enum
  23 import logging
  24 import time
  25 from typing import List
  26
  27 from ceilometerclient import client as ceilometer_client
  28 from ceilometerclient.exc import HTTPException
  29 import gnocchiclient.exceptions
  30 from gnocchiclient.v1 import client as gnocchi_client
  31 from keystoneauth1.exceptions.catalog import EndpointNotFound
  32 from keystoneclient.v3 import client as keystone_client
  33 from neutronclient.v2_0 import client as neutron_client
  34 from prometheus_api_client import PrometheusConnect as prometheus_client
  35
  36 from osm_mon.collector.metric import Metric
  37 from osm_mon.collector.utils.openstack import OpenstackUtils
  38 from osm_mon.collector.vnf_collectors.base_vim import BaseVimCollector
  39 from osm_mon.collector.vnf_metric import VnfMetric
  40 from osm_mon.core.common_db import CommonDbClient
  41 from osm_mon.core.config import Config
  42
  43
  44 log = logging.getLogger(__name__)
  45
  46 METRIC_MAPPINGS = {
  47     "average_memory_utilization": "memory.usage",
  48     "disk_read_ops": "disk.read.requests.rate",
  49     "disk_write_ops": "disk.write.requests.rate",
  50     "disk_read_bytes": "disk.read.bytes.rate",
  51     "disk_write_bytes": "disk.write.bytes.rate",
  52     "packets_in_dropped": "network.outgoing.packets.drop",
  53     "packets_out_dropped": "network.incoming.packets.drop",
  54     "packets_received": "network.incoming.packets.rate",
  55     "packets_sent": "network.outgoing.packets.rate",
  56     "cpu_utilization": "cpu",
  57 }
  58
  59 METRIC_MAPPINGS_FOR_PROMETHEUS_TSBD = {
  60     "cpu_utilization": "cpu",
  61     "average_memory_utilization": "memory_usage",
  62     "disk_read_ops": "disk_device_read_requests",
  63     "disk_write_ops": "disk_device_write_requests",
  64     "disk_read_bytes": "disk_device_read_bytes",
  65     "disk_write_bytes": "disk_device_write_bytes",
  66     "packets_in_dropped": "network_incoming_packets_drop",
  67     "packets_out_dropped": "network_outgoing_packets_drop",
  68     "packets_received": "network_incoming_packets",
  69     "packets_sent": "network_outgoing_packets",
  70 }
  71
  72 # Metrics which have new names in Rocky and higher releases
  73 METRIC_MAPPINGS_FOR_ROCKY_AND_NEWER_RELEASES = {
  74     "disk_read_ops": "disk.device.read.requests",
  75     "disk_write_ops": "disk.device.write.requests",
  76     "disk_read_bytes": "disk.device.read.bytes",
  77     "disk_write_bytes": "disk.device.write.bytes",
  78     "packets_received": "network.incoming.packets",
  79     "packets_sent": "network.outgoing.packets",
  80 }
  81
  82 METRIC_MULTIPLIERS = {"cpu": 0.0000001}
  83
  84 METRIC_AGGREGATORS = {"cpu": "rate:mean"}
  85
  86 INTERFACE_METRICS = [
  87     "packets_in_dropped",
  88     "packets_out_dropped",
  89     "packets_received",
  90     "packets_sent",
  91 ]
  92
  93 INSTANCE_DISK = [
  94     "disk_read_ops",
  95     "disk_write_ops",
  96     "disk_read_bytes",
  97     "disk_write_bytes",
  98 ]
  99
 100
 101 class MetricType(Enum):
 102     INSTANCE = "instance"
 103     INTERFACE_ALL = "interface_all"
 104     INTERFACE_ONE = "interface_one"
 105     INSTANCEDISK = "instancedisk"
 106
 107
 108 class OpenstackCollector(BaseVimCollector):
 109     def __init__(self, config: Config, vim_account_id: str, vim_session: object):
 110         super().__init__(config, vim_account_id)
 111         self.common_db = CommonDbClient(config)
 112         vim_account = self.common_db.get_vim_account(vim_account_id)
 113         self.backend = self._get_backend(vim_account, vim_session)
 114
 115     def _build_keystone_client(self, vim_account: dict) -> keystone_client.Client:
 116         sess = OpenstackUtils.get_session(vim_account)
 117         return keystone_client.Client(session=sess)
 118
 119     def _get_resource_uuid(
 120         self, nsr_id: str, vnf_member_index: str, vdur_name: str
 121     ) -> str:
 122         vdur = self.common_db.get_vdur(nsr_id, vnf_member_index, vdur_name)
 123         return vdur["vim-id"]
 124
 125     def collect(self, vnfr: dict) -> List[Metric]:
 126         nsr_id = vnfr["nsr-id-ref"]
 127         vnf_member_index = vnfr["member-vnf-index-ref"]
 128         vnfd = self.common_db.get_vnfd(vnfr["vnfd-id"])
 129         # Populate extra tags for metrics
 130         tags = {}
 131         tags["ns_name"] = self.common_db.get_nsr(nsr_id)["name"]
 132         if vnfr["_admin"]["projects_read"]:
 133             tags["project_id"] = vnfr["_admin"]["projects_read"][0]
 134         else:
 135             tags["project_id"] = ""
 136
 137         metrics = []
 138
 139         for vdur in vnfr["vdur"]:
 140             # This avoids errors when vdur records have not been completely filled
 141             if "name" not in vdur:
 142                 continue
 143             vdu = next(filter(lambda vdu: vdu["id"] == vdur["vdu-id-ref"], vnfd["vdu"]))
 144             if "monitoring-parameter" in vdu:
 145                 for param in vdu["monitoring-parameter"]:
 146                     metric_name = param["performance-metric"]
 147                     log.info(f"Using an {type(self.backend)} as backend")
 148                     if type(self.backend) is PrometheusTSBDBackend:
 149                         openstack_metric_name = self.backend.map_metric(metric_name)
 150                     else:
 151                         openstack_metric_name = METRIC_MAPPINGS[metric_name]
 152                     metric_type = self._get_metric_type(metric_name)
 153                     try:
 154                         resource_id = self._get_resource_uuid(
 155                             nsr_id, vnf_member_index, vdur["name"]
 156                         )
 157                     except ValueError:
 158                         log.warning(
 159                             "Could not find resource_uuid for vdur %s, vnf_member_index %s, nsr_id %s. "
 160                             "Was it recently deleted?",
 161                             vdur["name"],
 162                             vnf_member_index,
 163                             nsr_id,
 164                         )
 165                         continue
 166                     try:
 167                         log.info(
 168                             "Collecting metric type: %s and metric_name: %s and resource_id %s and ",
 169                             metric_type,
 170                             metric_name,
 171                             resource_id,
 172                         )
 173                         value = self.backend.collect_metric(
 174                             metric_type, openstack_metric_name, resource_id
 175                         )
 176
 177                         if (
 178                             value is None
 179                             and metric_name
 180                             in METRIC_MAPPINGS_FOR_ROCKY_AND_NEWER_RELEASES
 181                             and type(self.backend) is not PrometheusTSBDBackend
 182                         ):
 183                             # Reattempting metric collection with new metric names.
 184                             # Some metric names have changed in newer Openstack releases
 185                             log.info(
 186                                 "Reattempting metric collection for type: %s and name: %s and resource_id %s",
 187                                 metric_type,
 188                                 metric_name,
 189                                 resource_id,
 190                             )
 191                             openstack_metric_name = (
 192                                 METRIC_MAPPINGS_FOR_ROCKY_AND_NEWER_RELEASES[
 193                                     metric_name
 194                                 ]
 195                             )
 196                             value = self.backend.collect_metric(
 197                                 metric_type, openstack_metric_name, resource_id
 198                             )
 199                         if value is not None:
 200                             log.info("value: %s", value)
 201                             metric = VnfMetric(
 202                                 nsr_id,
 203                                 vnf_member_index,
 204                                 vdur["name"],
 205                                 metric_name,
 206                                 value,
 207                                 tags,
 208                             )
 209                             metrics.append(metric)
 210                         else:
 211                             log.info("metric value is empty")
 212                     except Exception as e:
 213                         log.exception(
 214                             "Error collecting metric %s for vdu %s"
 215                             % (metric_name, vdur["name"])
 216                         )
 217                         log.info("Error in metric collection: %s" % e)
 218         return metrics
 219
 220     def _get_backend(self, vim_account: dict, vim_session: object):
 221         if vim_account.get("prometheus-config"):
 222             try:
 223                 tsbd = PrometheusTSBDBackend(vim_account)
 224                 log.debug("Using prometheustsbd backend to collect metric")
 225                 return tsbd
 226             except Exception as e:
 227                 log.error(f"Can't create prometheus client, {e}")
 228                 return None
 229         try:
 230             gnocchi = GnocchiBackend(vim_account, vim_session)
 231             gnocchi.client.metric.list(limit=1)
 232             log.debug("Using gnocchi backend to collect metric")
 233             return gnocchi
 234         except (HTTPException, EndpointNotFound):
 235             ceilometer = CeilometerBackend(vim_account, vim_session)
 236             ceilometer.client.capabilities.get()
 237             log.debug("Using ceilometer backend to collect metric")
 238             return ceilometer
 239
 240     def _get_metric_type(self, metric_name: str) -> MetricType:
 241         if metric_name not in INTERFACE_METRICS:
 242             if metric_name not in INSTANCE_DISK:
 243                 return MetricType.INSTANCE
 244             else:
 245                 return MetricType.INSTANCEDISK
 246         else:
 247             return MetricType.INTERFACE_ALL
 248
 249
 250 class OpenstackBackend:
 251     def collect_metric(
 252         self, metric_type: MetricType, metric_name: str, resource_id: str
 253     ):
 254         pass
 255
 256
 257 class PrometheusTSBDBackend(OpenstackBackend):
 258     def __init__(self, vim_account: dict):
 259         self.map = self._build_map(vim_account)
 260         self.cred = vim_account["prometheus-config"].get("prometheus-cred")
 261         self.client = self._build_prometheus_client(
 262             vim_account["prometheus-config"]["prometheus-url"]
 263         )
 264
 265     def _build_prometheus_client(self, url: str) -> prometheus_client:
 266         return prometheus_client(url, disable_ssl=True)
 267
 268     def _build_map(self, vim_account: dict) -> dict:
 269         custom_map = METRIC_MAPPINGS_FOR_PROMETHEUS_TSBD
 270         if "prometheus-map" in vim_account["prometheus-config"]:
 271             custom_map.update(vim_account["prometheus-config"]["prometheus-map"])
 272         return custom_map
 273
 274     def collect_metric(
 275         self, metric_type: MetricType, metric_name: str, resource_id: str
 276     ):
 277         metric = self.query_metric(metric_name, resource_id)
 278         return metric["value"][1] if metric else None
 279
 280     def map_metric(self, metric_name: str):
 281         return self.map[metric_name]
 282
 283     def query_metric(self, metric_name, resource_id=None):
 284         metrics = self.client.get_current_metric_value(metric_name=metric_name)
 285         if resource_id:
 286             metric = next(
 287                 filter(lambda x: resource_id in x["metric"]["resource_id"], metrics)
 288             )
 289             return metric
 290         return metrics
 291
 292
 293 class GnocchiBackend(OpenstackBackend):
 294     def __init__(self, vim_account: dict, vim_session: object):
 295         self.client = self._build_gnocchi_client(vim_account, vim_session)
 296         self.neutron = self._build_neutron_client(vim_account, vim_session)
 297
 298     def _build_gnocchi_client(
 299         self, vim_account: dict, vim_session: object
 300     ) -> gnocchi_client.Client:
 301         return gnocchi_client.Client(session=vim_session)
 302
 303     def _build_neutron_client(
 304         self, vim_account: dict, vim_session: object
 305     ) -> neutron_client.Client:
 306         return neutron_client.Client(session=vim_session)
 307
 308     def collect_metric(
 309         self, metric_type: MetricType, metric_name: str, resource_id: str
 310     ):
 311         if metric_type == MetricType.INTERFACE_ALL:
 312             return self._collect_interface_all_metric(metric_name, resource_id)
 313
 314         elif metric_type == MetricType.INSTANCE:
 315             return self._collect_instance_metric(metric_name, resource_id)
 316
 317         elif metric_type == MetricType.INSTANCEDISK:
 318             return self._collect_instance_disk_metric(metric_name, resource_id)
 319
 320         else:
 321             raise Exception("Unknown metric type %s" % metric_type.value)
 322
 323     def _collect_interface_all_metric(self, openstack_metric_name, resource_id):
 324         total_measure = None
 325         interfaces = self.client.resource.search(
 326             resource_type="instance_network_interface",
 327             query={"=": {"instance_id": resource_id}},
 328         )
 329         for interface in interfaces:
 330             try:
 331                 measures = self.client.metric.get_measures(
 332                     openstack_metric_name, resource_id=interface["id"], limit=1
 333                 )
 334                 if measures:
 335                     if not total_measure:
 336                         total_measure = 0.0
 337                     total_measure += measures[-1][2]
 338             except (gnocchiclient.exceptions.NotFound, TypeError) as e:
 339                 # Gnocchi in some Openstack versions raise TypeError instead of NotFound
 340                 log.debug(
 341                     "No metric %s found for interface %s: %s",
 342                     openstack_metric_name,
 343                     interface["id"],
 344                     e,
 345                 )
 346         return total_measure
 347
 348     def _collect_instance_disk_metric(self, openstack_metric_name, resource_id):
 349         value = None
 350         instances = self.client.resource.search(
 351             resource_type="instance_disk",
 352             query={"=": {"instance_id": resource_id}},
 353         )
 354         for instance in instances:
 355             try:
 356                 measures = self.client.metric.get_measures(
 357                     openstack_metric_name, resource_id=instance["id"], limit=1
 358                 )
 359                 if measures:
 360                     value = measures[-1][2]
 361
 362             except gnocchiclient.exceptions.NotFound as e:
 363                 log.debug(
 364                     "No metric %s found for instance disk %s: %s",
 365                     openstack_metric_name,
 366                     instance["id"],
 367                     e,
 368                 )
 369         return value
 370
 371     def _collect_instance_metric(self, openstack_metric_name, resource_id):
 372         value = None
 373         try:
 374             aggregation = METRIC_AGGREGATORS.get(openstack_metric_name)
 375
 376             try:
 377                 measures = self.client.metric.get_measures(
 378                     openstack_metric_name,
 379                     aggregation=aggregation,
 380                     start=time.time() - 1200,
 381                     resource_id=resource_id,
 382                 )
 383                 if measures:
 384                     value = measures[-1][2]
 385             except (
 386                 gnocchiclient.exceptions.NotFound,
 387                 gnocchiclient.exceptions.BadRequest,
 388                 TypeError,
 389             ) as e:
 390                 # CPU metric in previous Openstack versions do not support rate:mean aggregation method
 391                 # Gnocchi in some Openstack versions raise TypeError instead of NotFound or BadRequest
 392                 if openstack_metric_name == "cpu":
 393                     log.debug(
 394                         "No metric %s found for instance %s: %s",
 395                         openstack_metric_name,
 396                         resource_id,
 397                         e,
 398                     )
 399                     log.info(
 400                         "Retrying to get metric %s for instance %s without aggregation",
 401                         openstack_metric_name,
 402                         resource_id,
 403                     )
 404                     measures = self.client.metric.get_measures(
 405                         openstack_metric_name, resource_id=resource_id, limit=1
 406                     )
 407                 else:
 408                     raise e
 409                 # measures[-1] is the last measure
 410                 # measures[-2] is the previous measure
 411                 # measures[x][2] is the value of the metric
 412                 if measures and len(measures) >= 2:
 413                     value = measures[-1][2] - measures[-2][2]
 414             if value:
 415                 # measures[-1][0] is the time of the reporting interval
 416                 # measures[-1][1] is the duration of the reporting interval
 417                 if aggregation:
 418                     # If this is an aggregate, we need to divide the total over the reported time period.
 419                     # Even if the aggregation method is not supported by Openstack, the code will execute it
 420                     # because aggregation is specified in METRIC_AGGREGATORS
 421                     value = value / measures[-1][1]
 422                 if openstack_metric_name in METRIC_MULTIPLIERS:
 423                     value = value * METRIC_MULTIPLIERS[openstack_metric_name]
 424         except gnocchiclient.exceptions.NotFound as e:
 425             log.debug(
 426                 "No metric %s found for instance %s: %s",
 427                 openstack_metric_name,
 428                 resource_id,
 429                 e,
 430             )
 431         return value
 432
 433
 434 class CeilometerBackend(OpenstackBackend):
 435     def __init__(self, vim_account: dict, vim_session: object):
 436         self.client = self._build_ceilometer_client(vim_account, vim_session)
 437
 438     def _build_ceilometer_client(
 439         self, vim_account: dict, vim_session: object
 440     ) -> ceilometer_client.Client:
 441         return ceilometer_client.Client("2", session=vim_session)
 442
 443     def collect_metric(
 444         self, metric_type: MetricType, metric_name: str, resource_id: str
 445     ):
 446         if metric_type != MetricType.INSTANCE:
 447             raise NotImplementedError(
 448                 "Ceilometer backend only support instance metrics"
 449             )
 450         measures = self.client.samples.list(
 451             meter_name=metric_name,
 452             limit=1,
 453             q=[{"field": "resource_id", "op": "eq", "value": resource_id}],
 454         )
 455         return measures[0].counter_volume if measures else None