make version of ts_mon compatible with appengine

appengine/sheriff_o_matic/monitoring.py

+#!/usr/bin/env python

[agable, 2015/08/12 22:18:08]

This file (appengine module, really) is the meat of getting all this to work. In
order to make it easily re-usable, let's try something:
* Move this file inside appengine_module/gae_ts_mon
* Put monitoring.yaml inside gae_ts_mon
* Put a dispatch.yaml file in s-o-m which points at the monitoring module to
handle /monitoring requests (see
https://cloud.google.com/appengine/docs/python/modules/routing)

That way, for someone else to use this same thing, all they have to do is create
a symlink to gae_ts_mon and add the dispatch.yaml file. That'd be shiny.

+# Copyright (c) 2015 The Chromium Authors. All rights reserved.
+# Use of this source code is governed by a BSD-style license that can be
+# found in the LICENSE file.
+
+"""Send system monitoring data to the timeseries monitoring API."""
+
+import logging
+import webapp2
+
+import gae_ts_mon
+import ts_alerts
+
+from google.appengine.api import app_identity
+from google.appengine.api import modules
+
+
+access_count = gae_ts_mon.CounterMetric('gae/access/count')
+
+
+class InitializeMonitoringHandler(webapp2.RequestHandler):
+
+  def get(self):
+    service = app_identity.get_application_id()
+    version = modules.get_current_version_name()
+    instance_id = int(modules.get_current_instance_id())
+    endpoint = 'pubsub://chrome-infra-mon-pubsub/monacq'
+    gae_ts_mon.initialize(job_name=version, instance=instance_id,
+                          service_name=service, endpoint=endpoint)
+    self.response.set_status(200, 'Initialized instance of ts_mon.')
+
+
+class MonitoringHandler(webapp2.RequestHandler):
+
+  ''' Called by cron jobs every 5 minutes to update metrics. '''
+  def get(self, key=None):
+    if key:
+      url = app_identity.get_default_version_hostname()
+      access_count.increment(fields={'url_path': url + '/' + key,
+                                     'request_path': key,
+                                     'request_type': 'GET'})

[agable, 2015/08/12 22:18:08]

The request type should be supplied by whoever calls the /monitoring endpoint
(because someone else is reporting a hit to their own endpoint, which may have
been a GET or a POST or something else), not hardcoded to GET.

+      return
+    gae_ts_mon.flush()
+    self.response.write('Sheriff-o-matic metrics updated.')
+    return
+
+  def post(self):

[agable, 2015/08/12 22:18:08]

This appears to be unreachable / dead code. Sucks that we can't get POST to
work, but shouldn't leave this hanging around to confuse folks.

+    key = self.request.body['endpoint']
+    if key not in self.ACCESS_COUNT.keys():
+      self.response.write('Unknown key %s', key)
+      self.abort(400)
+      return
+    self.ACCESS_COUNT[key] += 1
+    self.response.write('Post success: ' + key)
+
+
+app = webapp2.WSGIApplication([
+    ('/_ah/start', InitializeMonitoringHandler),
+    ('/monitoring', MonitoringHandler),
+    ('/monitoring/(.*)', MonitoringHandler)
+], debug=True)
+

[agable, 2015/08/12 22:18:08]

nit: no empty line