Send zendesk ticket when services found with high failure rates

app/celery/scheduled_tasks.py

@@ -45,6 +45,8 @@ from app.models import (
 from app.notifications.process_notifications import send_notification_to_queue
 from app.v2.errors import JobIncompleteError
 
+from app.service.utils import get_services_with_high_failure_rates
+
 
 @notify_celery.task(name="run-scheduled-jobs")
 @statsd(namespace="tasks")
@@ -253,3 +255,28 @@ def check_for_missing_rows_in_completed_jobs():
                     current_app.logger.info(
                         "Processing missing row: {} for job: {}".format(row_to_process.missing_row, job.id))
                     process_row(row, template, job, job.service, sender_id=sender_id)
+
+
+@notify_celery.task(name='check-for-services-with-high-failure-rates-or-sending-to-tv-numbers')
+@statsd(namespace="tasks")
+def check_for_services_with_high_failure_rates_or_sending_to_tv_numbers():
+    services_with_failures = get_services_with_high_failure_rates()
+    # services_sending_to_tv_numbers = dao_find_services_sending_to_tv_numbers(number=100)
+
+    if services_with_failures:
+        message = "{} service(s) have had high permanent-failure rates for sms messages in last 24 hours: ".format(
+            len(services_with_failures)
+        )
+        for service in services_with_failures:
+            message += "service id: {} failure rate: {}, ".format(service["id"], service["permanent_failure_rate"])
+
+        current_app.logger.exception(message)
+
+        if current_app.config['NOTIFY_ENVIRONMENT'] in ['live', 'production', 'test']:
+            zendesk_client.create_ticket(
+                subject="[{}] High failure rates for sms spotted for services".format(
+                    current_app.config['NOTIFY_ENVIRONMENT']
+                ),
+                message=message,
+                ticket_type=zendesk_client.TYPE_INCIDENT
+            )

app/config.py

@@ -276,6 +276,11 @@ class Config(object):
             'schedule': crontab(day_of_week='mon-fri', hour='9,15', minute=0),
             'options': {'queue': QueueNames.PERIODIC}
         },
+        'check-for-services-with-high-failure-rates-or-sending-to-tv-numbers': {
+            'task': 'check-for-services-with-high-failure-rates-or-sending-to-tv-numbers',
+            'schedule': crontab(day_of_week='mon-fri', hour=10, minute=30),
+            'options': {'queue': QueueNames.PERIODIC}
+        },
         'raise-alert-if-letter-notifications-still-sending': {
             'task': 'raise-alert-if-letter-notifications-still-sending',
             'schedule': crontab(hour=16, minute=30),

app/service/statistics.py

@@ -21,6 +21,19 @@ def format_statistics(statistics):
     return counts
 
 
+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)
+
+    if counts['permanent_failure'] + counts['all_other_statuses'] >= threshold:
+        rate = counts['permanent_failure'] / (counts['permanent_failure'] + counts['all_other_statuses'])
+    else:
+        rate = 0
+    return rate
+
+
 def format_admin_stats(statistics):
     counts = create_stats_dict()
 
@@ -94,6 +107,13 @@ def _update_statuses_from_row(update_dict, row):
         update_dict['failed'] += row.count
 
 
+def _count_if_status_is_permanent_failure_from_row(update_dict, row):
+    if row.status == 'permanent-failure':
+        update_dict['permanent_failure'] += row.count
+    else:
+        update_dict['all_other_statuses'] += row.count
+
+
 def create_empty_monthly_notification_status_stats_dict(year):
     utc_month_starts = get_months_for_financial_year(year)
     # nested dicts - data[month][template type][status] = count

app/service/utils.py

@@ -7,6 +7,11 @@ from app.models import (
     MOBILE_TYPE, EMAIL_TYPE,
     KEY_TYPE_TEST, KEY_TYPE_TEAM, KEY_TYPE_NORMAL)
 
+from app.service import statistics
+from datetime import datetime, timedelta
+
+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)]
@@ -52,3 +57,26 @@ def service_allowed_to_send_to(recipient, service, key_type, allow_whitelisted_r
                 whitelist_members
             )
         )
+
+
+def get_services_with_high_failure_rates(rate=0.25, threshold=100):
+    start_date = (datetime.utcnow() - timedelta(days=1)).date()
+    end_date = datetime.utcnow().date()
+
+    stats = fetch_stats_for_all_services_by_date_range(
+        start_date=start_date,
+        end_date=end_date,
+        include_from_test_key=False,
+    )
+    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
+                })
+    return results