Add striped rate-limiting

Slowness for a single data-set should not result in resource-exhaustion
for the other data sets. This is a resilience measure.

Allow up to 100 requests per minute, with bursting to 5 requests per
second.

We are doing this in the application since we have the knowledge about
URLs in here, rather than in Varnish or nginx.

We need to set rate-limiting high for tests, otherwise the tests fail as
the rate-limiting kicks in.

backdrop/read/api.py

@@ -5,6 +5,7 @@
 
 from flask import Flask, jsonify, request
 from flask_featureflags import FeatureFlag
+from flask_limiter import Limiter
 
 from .query import parse_query_from_request
 from .validation import validate_request_args
@@ -25,6 +26,7 @@
 
 app = Flask("backdrop.read.api")
 
+limiter = Limiter(app)
 feature_flags = FeatureFlag(app)
 
 # Configuration
@@ -153,7 +155,16 @@ def log_error_and_respond(data_set, message, status_code):
     return jsonify(status='error', message=message), status_code
 
 
+def data_group_key():
+    """Key used for rate-limiting requests to data-types"""
+    return '{data_group}:{data_type}'.format(
+        data_group=request.view_args.get('data_group'),
+        data_type=request.view_args.get('data_type'))
+
+
 @app.route('/data/<data_group>/<data_type>', methods=['GET', 'OPTIONS'])
+@limiter.limit(app.config.get('DATA_SET_RATE_LIMIT', '100/minute;5/second'),
+               data_group_key)
 def data(data_group, data_type):
     with statsd.timer('read.route.data.{data_group}.{data_type}'.format(
             data_group=data_group,
@@ -162,7 +173,15 @@ def data(data_group, data_type):
         return fetch(data_set_config)
 
 
+def data_set_key():
+    """Key used for rate-limiting requests to data sets"""
+    return '{data_set_name}'.format(
+        data_set_name=request.view_args.get('data_set_name'))
+
+
 @app.route('/<data_set_name>', methods=['GET', 'OPTIONS'])
+@limiter.limit(app.config.get('DATA_SET_RATE_LIMIT', '100/minute;5/second'),
+               data_set_key)
 @http_validation.etag
 def query(data_set_name):
     with statsd.timer('read.route.{data_set_name}'.format(

backdrop/read/config/test.py

@@ -3,4 +3,6 @@
 MONGO_PORT = 27017
 LOG_LEVEL = "ERROR"
 
+DATA_SET_RATE_LIMIT='10000/second'
+
 from development import STAGECRAFT_URL, STAGECRAFT_DATA_SET_QUERY_TOKEN, SIGNON_API_USER_TOKEN

requirements.txt

@@ -1,6 +1,7 @@
 argh==0.25
 Flask==0.10.1
 Flask-FeatureFlags==0.4
+Flask-Limiter==0.6.4
 gunicorn==19.1.1
 invoke
 isodate==0.5.0