flex_ts_mon/setup.py

# Copyright 2018 The Chromium OS Authors. All rights reserved.
# Use of this source code is governed by a BSD-style license that can be
# found in the LICENSE file.

"""Global singleton setup for ts-mon.

Ts-mon setup must happen post-fork, or else the gunicorn workers will all use
the same hostname and task_num arguments, causing conflicts when they flush
metrics.

This can be accomplished with flask's "app.before_first_request" decorator
function. See flask's documentation:
    http://flask.pocoo.org/docs/0.12/api/#flask.Flask.before_first_request
"""

from __future__ import absolute_import
from __future__ import division
from __future__ import print_function

import os
import socket

from chromite.lib import cros_logging as logging

from infra_libs.ts_mon.common import interface
from infra_libs.ts_mon.common import monitors
from infra_libs.ts_mon.common import standard_metrics
from infra_libs.ts_mon.common import targets
import googleapiclient.discovery

from flex_ts_mon import gae_app_identity_shim as app_identity
from flex_ts_mon import gae_modules_shim as modules


_WasSetup = False

_PRODX_MON_ENDPOINT = 'https://prodxmon-pa.googleapis.com/v1:insert'
_PRODXMON_SERVICE_ACCOUNT_EMAIL = (
    'app-engine-metric-publishers@'
    'prodx-mon-chrome-infra.google.com.iam.gserviceaccount.com')
_REGION = 'appengine'
_FLUSH_INTERVAL_SECS = 60


def SetupTsMonGlobalState(debug_file=None):
  """Sets up the ts-mon flushing thread.

  Args:
    debug_file: If non-none, send metrics to this path instead of to PubSub.
  """
  global _WasSetup  # pylint: disable=global-statement
  if not _WasSetup:
    _SetupTsMon(debug_file)
    _WasSetup = True


def _SetupTsMon(debug_file):
  """Sets up ts-mon global state given parsed argparse options.

  Args:
    debug_file: If non-none, send metrics to this path instead of to PubSub.
  """
  googleapiclient.discovery.logger.setLevel(logging.WARNING)
  try:
    if debug_file:
      endpoint = 'file://' + debug_file
    else:
      endpoint = _PRODX_MON_ENDPOINT

    _SetupAppEngineTarget()
    _SetupMonitorForEndpoint(endpoint)

    # TODO(phobbs): set up the same global app engine metrics as
    # gae_ts_mon.shared
    standard_metrics.init()

    interface.state.flush_mode = 'auto'
    # pylint: disable=protected-access
    interface.state.flush_thread = interface._FlushThread(_FLUSH_INTERVAL_SECS)
    interface.state.flush_thread.start()

    logging.notice('ts_mon was set up.')
  except Exception as e:
    logging.warning('Failed to configure ts_mon, monitoring is disabled: %s', e,
                    exc_info=True)


def _SetupAppEngineTarget():
  """Creates the ts-mon target for an App Engine app."""
  fqdn = socket.getfqdn().lower()
  host = fqdn.split('.')[0]

  # Note - any graphs using metrics from these apps need a taskless
  # precomputation to remove task_num.
  task_hostname = 'autogen:' + host
  task_num = os.getpid()

  service_name = app_identity.get_application_id()
  job_name = modules.get_current_module_name()

  interface.state.target = targets.TaskTarget(
      service_name, job_name, _REGION, task_hostname, task_num=task_num)

  logging.info('Initializing ts_mon with service_name=%s, job_name=%s, '
               'hostname=%s', service_name, job_name, task_hostname)


def _SetupMonitorForEndpoint(endpoint):
  """Sets up a ts-mon monitor.

  Args:
    endpoint: The endpoint to send metrics to. Starts with file:// or https://,
        or is "none".
  """
  if endpoint.startswith('file://'):
    interface.state.global_monitor = monitors.DebugMonitor(
        endpoint[len('file://'):])
  elif endpoint.startswith('https://'):
    logging.info('Using https monitor %s with %s', endpoint,
                 _PRODXMON_SERVICE_ACCOUNT_EMAIL)
    credentials = monitors.DelegateServiceAccountCredentials(
        _PRODXMON_SERVICE_ACCOUNT_EMAIL,
        monitors.GCECredentials())
    interface.state.global_monitor = monitors.HttpsMonitor(
        endpoint, credentials)
    interface.state.use_new_proto = True
  elif endpoint.lower() == 'none':
    logging.info('ts_mon monitoring has been explicitly disabled')
  else:
    logging.error('ts_mon monitoring is disabled because the endpoint provided'
                  ' is invalid or not supported: %s', endpoint)