# Step 1: Start continuous load
while true; do
  curl -s -o /dev/null -w "%{http_code}\n" "$GATEWAY_URL/health" >> /tmp/responses.log
  sleep 0.1
done &
LOAD_PID=$!

# Step 2: Get pod name
POD=$(kubectl get pods -l app=mcpgateway -o jsonpath='{.items[0].metadata.name}')
echo "Deleting pod: $POD"

# Step 3: Delete pod
kubectl delete pod $POD

# Step 4: Wait for new pod
kubectl wait --for=condition=ready pod -l app=mcpgateway --timeout=60s

# Step 5: Stop load generator
kill $LOAD_PID

# Step 6: Analyze responses
echo "Total requests: $(wc -l < /tmp/responses.log)"
echo "Failed requests: $(grep -v 200 /tmp/responses.log | wc -l)"
grep -v 200 /tmp/responses.log | sort | uniq -c

# Step 1: Check current restart count
kubectl get pods -l app=mcpgateway -o jsonpath='{.items[0].status.containerStatuses[0].restartCount}'

# Step 2: Simulate liveness failure
# Option A: Exec into pod and kill the process
POD=$(kubectl get pods -l app=mcpgateway -o jsonpath='{.items[0].metadata.name}')
kubectl exec $POD -- kill 1

# Option B: If app supports it, trigger unhealthy state via API

# Step 3: Watch pod status
kubectl get pods -l app=mcpgateway -w

# Step 4: Wait for restart
sleep 30

# Step 5: Verify restart count increased
kubectl get pods -l app=mcpgateway -o jsonpath='{.items[0].status.containerStatuses[0].restartCount}'

# Step 6: Verify pod is healthy again
kubectl get pods -l app=mcpgateway
curl -s "$GATEWAY_URL/health" | jq .

# Step 1: Start continuous load in background
while true; do
  START=$(date +%s%N)
  HTTP_CODE=$(curl -s -o /dev/null -w "%{http_code}" "$GATEWAY_URL/api/tools" \
    -H "Authorization: Bearer $TOKEN")
  END=$(date +%s%N)
  LATENCY=$(( ($END - $START) / 1000000 ))
  echo "$(date +%H:%M:%S) $HTTP_CODE ${LATENCY}ms"
  sleep 0.2
done > /tmp/rolling-update.log &
LOAD_PID=$!

# Step 2: Trigger rolling update (change image or env var)
kubectl set env deployment/mcpgateway ROLLING_UPDATE_TEST=$(date +%s)

# Step 3: Watch rollout
kubectl rollout status deployment/mcpgateway

# Step 4: Stop load generator
kill $LOAD_PID

# Step 5: Analyze results
echo "Total requests: $(wc -l < /tmp/rolling-update.log)"
echo "Failed requests: $(grep -v " 200 " /tmp/rolling-update.log | wc -l)"
echo "Max latency: $(awk '{print $3}' /tmp/rolling-update.log | sort -n | tail -1)"

# Step 6: Show any errors
grep -v " 200 " /tmp/rolling-update.log | head -10

# Step 1: Check rollout history
kubectl rollout history deployment/mcpgateway

# Step 2: Start load generator
while true; do
  curl -s -o /dev/null -w "%{http_code}\n" "$GATEWAY_URL/health"
  sleep 0.1
done > /tmp/rollback.log &
LOAD_PID=$!

# Step 3: Trigger rollback
kubectl rollout undo deployment/mcpgateway

# Step 4: Wait for rollback
kubectl rollout status deployment/mcpgateway

# Step 5: Stop load
kill $LOAD_PID

# Step 6: Verify no errors
grep -v 200 /tmp/rollback.log | wc -l

# Step 1: Identify node with gateway pods
NODE=$(kubectl get pods -l app=mcpgateway -o jsonpath='{.items[0].spec.nodeName}')
echo "Draining node: $NODE"

# Step 2: Start load generator
while true; do
  curl -s -o /dev/null -w "%{http_code}\n" "$GATEWAY_URL/health"
  sleep 0.1
done > /tmp/drain.log &
LOAD_PID=$!

# Step 3: Drain the node
kubectl drain $NODE --ignore-daemonsets --delete-emptydir-data

# Step 4: Watch pods reschedule
kubectl get pods -l app=mcpgateway -o wide -w &

# Step 5: Wait for reschedule
sleep 30

# Step 6: Stop load
kill $LOAD_PID

# Step 7: Analyze
grep -v 200 /tmp/drain.log | wc -l

# Step 8: Uncordon node
kubectl uncordon $NODE

# Step 1: Check current replicas
kubectl get deployment mcpgateway -o jsonpath='{.spec.replicas}'

# Step 2: Scale up
kubectl scale deployment mcpgateway --replicas=5

# Step 3: Watch pods come up
kubectl get pods -l app=mcpgateway -w

# Step 4: Wait for all ready
kubectl wait --for=condition=ready pod -l app=mcpgateway --timeout=120s

# Step 5: Verify traffic distribution
for i in {1..20}; do
  curl -s "$GATEWAY_URL/health" | jq -r '.hostname // empty'
done | sort | uniq -c

# Step 6: Verify all pods serving traffic