Add a janitor periodic job to cleanup stale ec2 instances from the node-e2e-tests service account

config/jobs/kubernetes/sig-testing/ec2-e2e.yaml

@@ -347,3 +347,40 @@ periodics:
           requests:
             cpu: 8
             memory: 10Gi
+- interval: 12h
+  cluster: eks-prow-build-cluster
+  name: ci-aws-ec2-janitor
+  annotations:
+    testgrid-dashboards: amazon-ec2
+    testgrid-tab-name: aws-ec2-janitor
+    description: cleanup any EC2 instances older than 2 days
+  labels:
+    preset-e2e-containerd-ec2: "true"
+  decorate: true
+  decoration_config:
+    timeout: 4h
+  spec:
+    serviceAccountName: node-e2e-tests
+    containers:
+      - image: gcr.io/k8s-staging-test-infra/kubekins-e2e:v20230727-ea685f8747-master
+        command:
+          - runner.sh
+        args:
+          - bash
+          - -c
+          - |
+            set -xeu
+            target=$(date -d "2 days ago" +%Y-%m-%d)
+            region="us-east-1"
+            # Is the Launch time *before* two days ago? if so then terminate the instance.
+            for id in `aws ec2 describe-instances --query "Reservations[].Instances[?LaunchTime<='$target'][].{id: InstanceId}" --filters "Name=instance-state-name,Values=running" --output text`
+            do
+              aws ec2 --region us-east-1 terminate-instances --instance-ids $id
+            done
+        resources:
+          limits:
+            cpu: 2
+            memory: 4Gi
+          requests:
+            cpu: 2
+            memory: 4Gi