Simplify failure rate by building separate query

app/celery/scheduled_tasks.py

@@ -266,11 +266,7 @@ def check_for_services_with_high_failure_rates_or_sending_to_tv_numbers():
     message = ""
 
     services_with_failures = get_services_with_high_failure_rates(start_date=start_date, end_date=end_date)
-    services_sending_to_tv_numbers = dao_find_services_sending_to_tv_numbers(
-        threshold=100,
-        start_date=start_date,
-        end_date=end_date
-    )
+    services_sending_to_tv_numbers = dao_find_services_sending_to_tv_numbers(start_date=start_date, end_date=end_date)
 
     if services_with_failures:
         message += "{} service(s) have had high permanent-failure rates for sms messages in last 24 hours:\n".format(
@@ -291,6 +287,7 @@ def check_for_services_with_high_failure_rates_or_sending_to_tv_numbers():
         current_app.logger.exception(message)
 
         if current_app.config['NOTIFY_ENVIRONMENT'] in ['live', 'production', 'test']:
+            message += "\nThings to do: contact service? revoke their key?"
             zendesk_client.create_ticket(
                 subject="[{}] High failure rates for sms spotted for services".format(
                     current_app.config['NOTIFY_ENVIRONMENT']

app/dao/services_dao.py

@@ -524,9 +524,7 @@ def dao_fetch_active_users_for_service(service_id):
 
 
 def dao_find_services_sending_to_tv_numbers(start_date, end_date, threshold=100):
-
     return db.session.query(
-        Service.name.label('service_name'),
         Notification.service_id.label('service_id'),
         func.count(Notification.id).label('notification_count')
     ).filter(
@@ -541,7 +539,27 @@ def dao_find_services_sending_to_tv_numbers(start_date, end_date, threshold=100)
         Service.active == True,
     ).group_by(
         Notification.service_id,
-        Service.name
     ).having(
         func.count(Notification.id) > threshold
     ).all()
+
+
+def dao_find_real_sms_notification_count_by_status_for_live_services(start_date, end_date):
+    #  only works within services' retention period
+    return db.session.query(
+        Notification.service_id.label('service_id'),
+        Notification.status.label('status'),
+        func.count(Notification.id).label('count')
+    ).filter(
+        Notification.service_id == Service.id,
+        Notification.created_at >= start_date,
+        Notification.created_at <= end_date,
+        Notification.key_type != KEY_TYPE_TEST,
+        Notification.notification_type == SMS_TYPE,
+        Service.restricted == False,  # noqa
+        Service.research_mode == False,
+        Service.active == True,
+    ).group_by(
+        Notification.service_id,
+        Notification.status
+    ).all()

app/service/statistics.py

@@ -24,8 +24,7 @@ def format_statistics(statistics):
 def get_rate_of_permanent_failures_for_service(statistics, threshold=100):
     counts = {"permanent_failure": 0, "all_other_statuses": 0}
     for row in statistics:
-        if row.notification_type == 'sms':
-            _count_if_status_is_permanent_failure_from_row(counts, row)
+        _count_if_status_is_permanent_failure_from_row(counts, row)
 
     if counts['permanent_failure'] + counts['all_other_statuses'] >= threshold:
         rate = counts['permanent_failure'] / (counts['permanent_failure'] + counts['all_other_statuses'])

app/service/utils.py

@@ -2,6 +2,8 @@ import itertools
 
 from notifications_utils.recipients import allowed_to_send_to
 
+from app.dao.services_dao import dao_find_real_sms_notification_count_by_status_for_live_services
+
 from app.models import (
     ServiceWhitelist,
     MOBILE_TYPE, EMAIL_TYPE,
@@ -9,8 +11,6 @@ from app.models import (
 
 from app.service import statistics
 
-from app.dao.fact_notification_status_dao import fetch_stats_for_all_services_by_date_range
-
 
 def get_recipients_from_request(request_json, key, type):
     return [(type, recipient) for recipient in request_json.get(key)]
@@ -59,20 +59,18 @@ def service_allowed_to_send_to(recipient, service, key_type, allow_whitelisted_r
 
 
 def get_services_with_high_failure_rates(start_date, end_date, rate=0.25, threshold=100):
-    stats = fetch_stats_for_all_services_by_date_range(
-        start_date=start_date.date(),
-        end_date=end_date.date(),
-        include_from_test_key=False,
+    stats = dao_find_real_sms_notification_count_by_status_for_live_services(
+        start_date=start_date,
+        end_date=end_date,
     )
     results = []
     for service_id, rows in itertools.groupby(stats, lambda x: x.service_id):
         rows = list(rows)
-        if not rows[0].restricted and not rows[0].research_mode and rows[0].active:
-            permanent_failure_rate = statistics.get_rate_of_permanent_failures_for_service(rows, threshold=threshold)
-            if permanent_failure_rate >= rate:
-                results.append({
-                    'id': str(rows[0].service_id),
-                    'name': rows[0].name,
-                    'permanent_failure_rate': permanent_failure_rate
-                })
+
+        permanent_failure_rate = statistics.get_rate_of_permanent_failures_for_service(rows, threshold=threshold)
+        if permanent_failure_rate >= rate:
+            results.append({
+                'id': str(rows[0].service_id),
+                'permanent_failure_rate': permanent_failure_rate
+            })
     return results