eval: Added eval for troubleshooting and fixing OOMKilled pod (#360)

* eval: Added eval for troubleshooting and fixing OOMKilled pod

* refactor: Improve OOMKilled event detection in setup script

* refactor: Disguise memory load to reflect realistic backend behavior

Author: prasad89

k8s-bench/tasks/fix-oomkilled/artifacts/memory-hungry-app.yaml

@@ -0,0 +1,48 @@
+apiVersion: apps/v1
+kind: Deployment
+metadata:
+  name: backend-api
+  namespace: webapp-backend
+  labels:
+    app: backend-api
+spec:
+  replicas: 1
+  selector:
+    matchLabels:
+      app: backend-api
+  template:
+    metadata:
+      labels:
+        app: backend-api
+    spec:
+      containers:
+        - name: api-server
+          image: nginx:alpine
+          command: ["/bin/sh"]
+          args:
+            - -c
+            - |
+              echo "Starting backend service..."
+              dd if=/dev/zero of=/tmp/cache.dat bs=1M count=150 &
+              nginx -g 'daemon off;'
+          ports:
+            - containerPort: 80
+          resources:
+            requests:
+              memory: "64Mi"
+              cpu: "50m"
+            limits:
+              memory: "128Mi"
+              cpu: "100m"
+          livenessProbe:
+            httpGet:
+              path: /
+              port: 80
+            initialDelaySeconds: 10
+            periodSeconds: 5
+          readinessProbe:
+            httpGet:
+              path: /
+              port: 80
+            initialDelaySeconds: 5
+            periodSeconds: 3

k8s-bench/tasks/fix-oomkilled/cleanup.sh

@@ -0,0 +1,2 @@
+#!/usr/bin/env bash
+kubectl delete namespace webapp-backend --ignore-not-found=true

k8s-bench/tasks/fix-oomkilled/setup.sh

@@ -0,0 +1,22 @@
+#!/usr/bin/env bash
+kubectl delete namespace webapp-backend --ignore-not-found
+
+# Create namespace
+kubectl create namespace webapp-backend
+
+# Apply the deployment from artifacts
+kubectl apply -f artifacts/memory-hungry-app.yaml
+
+# Wait for the deployment to be created
+kubectl rollout status deployment/backend-api -n webapp-backend --timeout=60s || true
+
+# Wait until an OOMKilled event is detected (timeout after 30s)
+echo "Waiting for OOMKilled event to occur..."
+for i in {1..15}; do
+  OOMKILLED_COUNT=$(kubectl get events -n webapp-backend --field-selector reason=OOMKilling -o json | jq '.items | length')
+  if [ "$OOMKILLED_COUNT" -gt 0 ]; then
+    echo "OOMKilled event detected."
+    break
+  fi
+  sleep 2
+done

k8s-bench/tasks/fix-oomkilled/task.yaml

@@ -0,0 +1,6 @@
+script:
+  - prompt: "I have a web application running in the 'webapp-backend' namespace that keeps crashing. Can you help me figure out what's going wrong and fix it?"
+setup: "setup.sh"
+verifier: "verify.sh"
+cleanup: "cleanup.sh"
+difficulty: "medium"

k8s-bench/tasks/fix-oomkilled/verify.sh

@@ -0,0 +1,35 @@
+#!/usr/bin/env bash
+
+NAMESPACE="webapp-backend"
+DEPLOYMENT="backend-api"
+
+# Check if the deployment is ready
+if ! kubectl wait --for=condition=Available deployment/$DEPLOYMENT -n $NAMESPACE --timeout=60s; then
+    echo "Deployment is not available"
+    exit 1
+fi
+
+# Check if pods are running
+if ! kubectl wait --for=condition=Ready pod -l app=backend-api -n $NAMESPACE --timeout=30s; then
+    echo "Pods are not ready"
+    exit 1
+fi
+
+# Check that there are no recent OOMKilled events
+OOMKILLED_COUNT=$(kubectl get events -n $NAMESPACE --field-selector reason=OOMKilling --sort-by='.lastTimestamp' -o json | jq '.items | length')
+
+if [ "$OOMKILLED_COUNT" -gt 0 ]; then
+    # Check if the most recent OOMKilled event is from the last 2 minutes (indicating ongoing issues)
+    RECENT_OOMKILLED=$(kubectl get events -n $NAMESPACE --field-selector reason=OOMKilling --sort-by='.lastTimestamp' -o jsonpath='{.items[-1].lastTimestamp}' 2>/dev/null)
+    if [ -n "$RECENT_OOMKILLED" ]; then
+        RECENT_TIME=$(date -d "$RECENT_OOMKILLED" +%s 2>/dev/null)
+        CURRENT_TIME=$(date +%s)
+        if [ $((CURRENT_TIME - RECENT_TIME)) -lt 120 ]; then
+            echo "Recent OOMKilled events detected"
+            exit 1
+        fi
+    fi
+fi
+
+echo "Pod is running successfully without OOMKilled events"
+exit 0