Keep important tests (#34966)

In https://github.com/getsentry/sentry/pull/34906 I pinned snuba to unblock CI, but we shouldn't pin snuba for much longer. This unpins snuba and fixes the failing tests (please see https://github.com/getsentry/sentry/pull/34906, https://github.com/getsentry/sentry/runs/6538111731?check_suite_focus=true) that fail with snuba@ 3cacc16753696de0953e80af5dcdabefc65d370e.

I don't know this part of Sentry very much, so going to be triaging this to people who can determine the best thing to do (fix these tests, adapt to new snuba stuff?). This is purely just to unblock CI and keep snuba on nightly.

src/sentry/conf/server.py

@@ -1941,7 +1941,7 @@ SENTRY_DEVSERVICES = {
     ),
     "snuba": lambda settings, options: (
         {
-            "image": "getsentry/snuba:f063336085e7be0ccbdb52791aaf97882ba7a26f" if not APPLE_ARM64
+            "image": "getsentry/snuba:nightly" if not APPLE_ARM64
             # We cross-build arm64 images on GH's Apple Intel runners
             else "ghcr.io/getsentry/snuba-arm64-dev:latest",
             "pull": True,

tests/sentry/incidents/test_subscription_processor.py

@@ -9,7 +9,6 @@ from django.utils import timezone
 from exam import fixture, patcher
 from freezegun import freeze_time
 
-from sentry.constants import CRASH_RATE_ALERT_AGGREGATE_ALIAS, CRASH_RATE_ALERT_SESSION_COUNT_ALIAS
 from sentry.incidents.logic import (
     CRITICAL_TRIGGER_LABEL,
     WARNING_TRIGGER_LABEL,
@@ -1596,9 +1595,17 @@ class ProcessUpdateTest(ProcessUpdateBaseClass):
         )
 
 
-class CrashRateAlertProcessUpdateTest(ProcessUpdateBaseClass):
+class MetricsCrashRateAlertProcessUpdateTest(ProcessUpdateBaseClass, SessionMetricsTestCase):
+    entity_subscription_metrics = patcher("sentry.snuba.entity_subscription.metrics")
+
     def setUp(self):
         super().setUp()
+        for status in ["exited", "crashed"]:
+            self.store_session(
+                self.build_session(
+                    status=status,
+                )
+            )
 
     @fixture
     def sub(self):
@@ -1610,7 +1617,7 @@ class CrashRateAlertProcessUpdateTest(ProcessUpdateBaseClass):
     def crash_rate_alert_rule(self):
         rule = self.create_alert_rule(
             projects=[self.project],
-            dataset=QueryDatasets.SESSIONS,
+            dataset=QueryDatasets.METRICS,
             name="JustAValidRule",
             query="",
             aggregate="percentage(sessions_crashed, sessions) AS _crash_rate_alert_aggregate",
@@ -1649,6 +1656,7 @@ class CrashRateAlertProcessUpdateTest(ProcessUpdateBaseClass):
         )
 
     def send_crash_rate_alert_update(self, rule, value, subscription, time_delta=None, count=EMPTY):
+        org_id = self.organization.id
         self.email_action_handler.reset_mock()
         if time_delta is None:
             time_delta = timedelta()
@@ -1663,6 +1671,17 @@ class CrashRateAlertProcessUpdateTest(ProcessUpdateBaseClass):
         with self.feature(
             ["organizations:incidents", "organizations:performance-view"]
         ), self.capture_on_commit_callbacks(execute=True):
+            if value is None:
+                numerator, denominator = 0, 0
+            else:
+                if count is EMPTY:
+                    numerator, denominator = value.as_integer_ratio()
+                else:
+                    denominator = count
+                    numerator = int(value * denominator)
+            session_status = resolve_tag_key(org_id, "session.status")
+            tag_value_init = resolve_weak(org_id, "init")
+            tag_value_crashed = resolve_weak(org_id, "crashed")
             processor.process_update(
                 {
                     "subscription_id": subscription.subscription_id
@@ -1670,12 +1689,12 @@ class CrashRateAlertProcessUpdateTest(ProcessUpdateBaseClass):
                     else uuid4().hex,
                     "values": {
                         "data": [
+                            {"project_id": 8, session_status: tag_value_init, "value": denominator},
                             {
-                                CRASH_RATE_ALERT_AGGREGATE_ALIAS: value,
-                                CRASH_RATE_ALERT_SESSION_COUNT_ALIAS: randint(1, 100)
-                                if count is EMPTY
-                                else count,
-                            }
+                                "project_id": 8,
+                                session_status: tag_value_crashed,
+                                "value": numerator,
+                            },
                         ]
                     },
                     "timestamp": timestamp,
@@ -2102,75 +2121,6 @@ class CrashRateAlertProcessUpdateTest(ProcessUpdateBaseClass):
         self.assert_trigger_exists_with_status(incident, trigger, TriggerStatus.ACTIVE)
         self.assert_action_handler_called_with_actions(incident, [])
 
-
-class MetricsCrashRateAlertProcessUpdateTest(
-    CrashRateAlertProcessUpdateTest, SessionMetricsTestCase
-):
-    entity_subscription_metrics = patcher("sentry.snuba.entity_subscription.metrics")
-
-    def setUp(self):
-        super().setUp()
-        for status in ["exited", "crashed"]:
-            self.store_session(
-                self.build_session(
-                    status=status,
-                )
-            )
-        rule = self.crash_rate_alert_rule
-        snuba_query = rule.snuba_query
-        snuba_query.dataset = QueryDatasets.METRICS.value
-        snuba_query.save()
-
-    def send_crash_rate_alert_update(self, rule, value, subscription, time_delta=None, count=EMPTY):
-        org_id = self.organization.id
-        self.email_action_handler.reset_mock()
-        if time_delta is None:
-            time_delta = timedelta()
-        processor = SubscriptionProcessor(subscription)
-
-        if time_delta is not None:
-            timestamp = timezone.now() + time_delta
-        else:
-            timestamp = timezone.now()
-        timestamp = timestamp.replace(tzinfo=pytz.utc, microsecond=0)
-
-        with self.feature(
-            ["organizations:incidents", "organizations:performance-view"]
-        ), self.capture_on_commit_callbacks(execute=True):
-            if value is None:
-                numerator, denominator = 0, 0
-            else:
-                if count is EMPTY:
-                    numerator, denominator = value.as_integer_ratio()
-                else:
-                    denominator = count
-                    numerator = int(value * denominator)
-            session_status = resolve_tag_key(org_id, "session.status")
-            tag_value_init = resolve_weak(org_id, "init")
-            tag_value_crashed = resolve_weak(org_id, "crashed")
-            processor.process_update(
-                {
-                    "subscription_id": subscription.subscription_id
-                    if subscription
-                    else uuid4().hex,
-                    "values": {
-                        "data": [
-                            {"project_id": 8, session_status: tag_value_init, "value": denominator},
-                            {
-                                "project_id": 8,
-                                session_status: tag_value_crashed,
-                                "value": numerator,
-                            },
-                        ]
-                    },
-                    "timestamp": timestamp,
-                    "interval": 1,
-                    "partition": 1,
-                    "offset": 1,
-                }
-            )
-        return processor
-
     def test_ensure_case_when_no_metrics_index_not_found_is_handled_gracefully(self):
         MetricsKeyIndexer.objects.all().delete()
         rule = self.crash_rate_alert_rule