fix(alerts): alert monitoring + trendlines only for absolute

frontend/src/lib/components/Alerts/insightAlertsLogic.ts

@@ -3,7 +3,7 @@ import { loaders } from 'kea-loaders'
 import api from 'lib/api'
 import { insightVizDataLogic } from 'scenes/insights/insightVizDataLogic'
 
-import { GoalLine, InsightThresholdType } from '~/queries/schema'
+import { AlertConditionType, GoalLine, InsightThresholdType } from '~/queries/schema'
 import { getBreakdown, isInsightVizNode, isTrendsQuery } from '~/queries/utils'
 import { InsightLogicProps } from '~/types'
 
@@ -67,6 +67,7 @@ export const insightAlertsLogic = kea<insightAlertsLogicType>([
                 alerts.flatMap((alert) => {
                     if (
                         alert.threshold.configuration.type !== InsightThresholdType.ABSOLUTE ||
+                        alert.condition.type !== AlertConditionType.ABSOLUTE_VALUE ||
                         !alert.threshold.configuration.bounds
                     ) {
                         return []

posthog/tasks/alerts/checks.py

@@ -34,12 +34,22 @@
     alert_calculation_interval_to_relativedelta,
 )
 from posthog.tasks.alerts.trends import check_trends_alert
+import time
+import math
 
 
 logger = structlog.get_logger(__name__)
 
 
-class AlertCheckException(Exception): ...
+class AlertCheckException(Exception):
+    """
+    Required for custom exceptions to pass stack trace to sentry.
+    Subclassing through other ways doesn't transfer the traceback.
+    https://stackoverflow.com/a/69963663/5540417
+    """
+
+    def __init__(self, err):
+        self.__traceback__ = err.__traceback__
 
 
 HOURLY_ALERTS_BACKLOG_GAUGE = Gauge(
@@ -102,6 +112,9 @@ def alerts_backlog_task() -> None:
 
     DAILY_ALERTS_BACKLOG_GAUGE.set(daily_alerts_breaching_sla)
 
+    # sleeping 30s for prometheus to pick up the metrics sent during task
+    time.sleep(30)
+
 
 @shared_task(
     ignore_result=True,
@@ -158,6 +171,8 @@ def check_alert_task(alert_id: str) -> None:
 
 
 def check_alert(alert_id: str) -> None:
+    task_start_time = time.time()
+
     try:
         alert = AlertConfiguration.objects.get(id=alert_id, enabled=True)
     except AlertConfiguration.DoesNotExist:
@@ -199,8 +214,15 @@ def check_alert(alert_id: str) -> None:
         check_alert_and_notify_atomically(alert)
     except Exception as err:
         ALERT_CHECK_ERROR_COUNTER.inc()
+
         logger.exception(AlertCheckException(err))
-        capture_exception(AlertCheckException(err))
+        capture_exception(
+            AlertCheckException(err),
+            tags={
+                "alert_configuration_id": alert_id,
+            },
+        )
+
         # raise again so alert check is retried depending on error type
         raise
     finally:
@@ -209,6 +231,13 @@ def check_alert(alert_id: str) -> None:
         alert.is_calculating = False
         alert.save()
 
+        task_duration = time.time() - task_start_time
+
+        # Ensure task runs at least 40s
+        # for prometheus to pick up the metrics sent during task
+        time_left_to_run = 40 - math.floor(task_duration)
+        time.sleep(time_left_to_run)
+
 
 @transaction.atomic
 def check_alert_and_notify_atomically(alert: AlertConfiguration) -> None: