Making the github arc runner alarm more generic and scoped to runners…

… only
cds-snc · Jun 11, 2024 · dfaeb5c · dfaeb5c
1 parent ba3bf69
commit dfaeb5c
Show file tree

Hide file tree

Showing 2 changed files with 8 additions and 8 deletions.
diff --git a/aws/eks/cloudwatch_alarms.tf b/aws/eks/cloudwatch_alarms.tf
@@ -939,14 +939,14 @@ resource "aws_cloudwatch_metric_alarm" "aggregating-queues-not-active-5-minutes-
   ok_actions          = [var.sns_alert_critical_arn]
 }
 
-resource "aws_cloudwatch_metric_alarm" "github-arc-runner-write-alarm" {
+resource "aws_cloudwatch_metric_alarm" "github-arc-runner-error-alarm" {
   count               = var.cloudwatch_enabled ? 1 : 0
-  alarm_name          = "github-arc-runner-write-alarm"
-  alarm_description   = "GitHub ARC Runners Are Failing - Check Version Deprecation"
+  alarm_name          = "github-arc-runner-error-alarm"
+  alarm_description   = "GitHub ARC Runners Are Failing"
   comparison_operator = "LessThanThreshold"
   evaluation_periods  = "1"
-  metric_name         = aws_cloudwatch_log_metric_filter.github-arc-write-alarm[0].metric_transformation[0].name
-  namespace           = aws_cloudwatch_log_metric_filter.github-arc-write-alarm[0].metric_transformation[0].namespace
+  metric_name         = aws_cloudwatch_log_metric_filter.github-arc-runner-alarm[0].metric_transformation[0].name
+  namespace           = aws_cloudwatch_log_metric_filter.github-arc-runner-alarm[0].metric_transformation[0].namespace
   period              = "300"
   statistic           = "Sum"
   threshold           = 1

diff --git a/aws/eks/cloudwatch_log.tf b/aws/eks/cloudwatch_log.tf
@@ -167,14 +167,14 @@ resource "aws_cloudwatch_log_metric_filter" "aggregating-queues-are-active" {
   }
 }
 
-resource "aws_cloudwatch_log_metric_filter" "github-arc-write-alarm" {
+resource "aws_cloudwatch_log_metric_filter" "github-arc-runner-alarm" {
   count          = var.cloudwatch_enabled ? 1 : 0
   name           = "GitHub ARC Runners Write Alarm"
-  pattern        = "WRITE ERROR"
+  pattern        = "{ $.kubernetes.pod_name = \"github-arc-ss-${var.env}-*-runner-*\"  && $.log = \"*ERROR*\" }"
   log_group_name = aws_cloudwatch_log_group.notification-canada-ca-eks-application-logs[0].name
 
   metric_transformation {
-    name      = "aggregating-github-arc-write-alarm"
+    name      = "aggregating-github-arc-runner-alarm"
     namespace = "LogMetrics"
     value     = "1"
   }