fix(tf): sql alerting policy (#1231)

* fix(tf): wip alerting policy Co-authored-by: Thomas Arrow <[email protected]>
wmde · Nov 9, 2023 · c0c3950 · c0c3950
1 parent db48f1d
commit c0c3950
Show file tree

Hide file tree

Showing 2 changed files with 12 additions and 0 deletions.
diff --git a/tf/modules/CHANGELOG.md b/tf/modules/CHANGELOG.md
@@ -1,3 +1,6 @@
+# tf-module-monitoring-21
+- Attempt to fix the failed logical SQL backup alerting policy
+
 # tf-module-monitoring-20
 - Add alert policy for critical PV usage (Logical SQL Backup)
 - Fix failed logcial SQL backup alert policy by setting duration to 24h instead of 24h 30m

diff --git a/tf/modules/monitoring/sql-backup-failure-alert.tf b/tf/modules/monitoring/sql-backup-failure-alert.tf
@@ -24,6 +24,15 @@ resource "google_monitoring_alert_policy" "alert_policy_sql_logical_backup_failu
     condition_absent {
       filter   = "metric.type=\"logging.googleapis.com/user/${google_logging_metric.sql_logical_backup_success.name}\" AND resource.type=\"k8s_container\""
       duration = "86400s"
+
+      aggregations {
+        cross_series_reducer = "REDUCE_COUNT"
+        per_series_aligner   = "ALIGN_COUNT"
+        alignment_period     = "60s"
+        group_by_fields = [
+          "resource.labels.container_name",
+        ]
+      }
       trigger {
         count = 1
       }