1 files changed, 89 insertions, 0 deletions
diff --git a/scripts/cleanup-benchmark-data.sh b/scripts/cleanup-benchmark-data.sh
new file mode 100644
index 0000000..48ba187
--- /dev/null
+++ b/scripts/cleanup-benchmark-data.sh
@@ -0,0 +1,89 @@
+#!/bin/bash
+# Cleanup script: Delete benchmark data from Prometheus
+# This uses the Prometheus Admin API to selectively remove benchmark metrics
+# Run from repo root: ./scripts/cleanup-benchmark-data.sh [prometheus_url]
+
+set -e
+
+PROMETHEUS_URL="${1:-http://localhost:9090}"
+
+echo "=== Prometheus Benchmark Data Cleanup ==="
+echo ""
+echo "Prometheus URL: $PROMETHEUS_URL"
+echo ""
+
+# Check if port-forward is needed
+if [[ "$PROMETHEUS_URL" == *"localhost"* ]]; then
+  echo "Note: Make sure you have port-forward running:"
+  echo "  kubectl port-forward -n monitoring svc/prometheus-kube-prometheus-prometheus 9090:9090"
+  echo ""
+fi
+
+# Metrics to delete
+METRICS=(
+  "epimetheus_benchmark_cpu_usage"
+  "epimetheus_benchmark_memory_bytes"
+  "epimetheus_benchmark_disk_io_bytes"
+  "epimetheus_benchmark_network_rx_bytes"
+  "epimetheus_benchmark_network_tx_bytes"
+  "epimetheus_benchmark_requests_total"
+  "epimetheus_benchmark_errors_total"
+  "epimetheus_benchmark_response_time_ms"
+  "epimetheus_benchmark_active_connections"
+  "epimetheus_benchmark_queue_depth"
+)
+
+echo "Step 1: Deleting benchmark metrics..."
+echo ""
+
+SUCCESS_COUNT=0
+ERROR_COUNT=0
+
+for METRIC in "${METRICS[@]}"; do
+  echo "  Deleting: $METRIC"
+
+  # Delete series endpoint returns HTTP 204 No Content on success
+  HTTP_CODE=$(curl -s -o /dev/null -w "%{http_code}" -X POST -g "${PROMETHEUS_URL}/api/v1/admin/tsdb/delete_series?match[]=${METRIC}")
+
+  if [ "$HTTP_CODE" == "204" ] || [ "$HTTP_CODE" == "200" ]; then
+    echo "    ✅ Success (HTTP $HTTP_CODE)"
+    SUCCESS_COUNT=$((SUCCESS_COUNT + 1))
+  else
+    echo "    ❌ Error: HTTP $HTTP_CODE"
+    ERROR_COUNT=$((ERROR_COUNT + 1))
+  fi
+done
+
+echo ""
+echo "Deletion summary: $SUCCESS_COUNT succeeded, $ERROR_COUNT failed"
+echo ""
+
+if [ $ERROR_COUNT -eq 0 ]; then
+  echo "Step 2: Cleaning up tombstones..."
+  echo ""
+
+  # Clean tombstones endpoint returns HTTP 204 No Content on success
+  CLEANUP_HTTP_CODE=$(curl -s -o /dev/null -w "%{http_code}" -X POST "${PROMETHEUS_URL}/api/v1/admin/tsdb/clean_tombstones")
+
+  if [ "$CLEANUP_HTTP_CODE" == "204" ] || [ "$CLEANUP_HTTP_CODE" == "200" ]; then
+    echo "  ✅ Tombstones cleaned successfully (HTTP $CLEANUP_HTTP_CODE)"
+    echo ""
+    echo "🎉 Cleanup complete!"
+    echo ""
+    echo "Note: Prometheus may take a few moments to compact the database"
+    echo "and free up disk space."
+  else
+    echo "  ❌ Error cleaning tombstones: HTTP $CLEANUP_HTTP_CODE"
+    exit 1
+  fi
+else
+  echo "⚠️  Some deletions failed. Skipping tombstone cleanup."
+  echo "Check Prometheus admin API is enabled with:"
+  echo "  kubectl get prometheus -n monitoring prometheus-kube-prometheus-prometheus -o yaml | grep -A5 additionalArgs"
+  exit 1
+fi
+
+echo ""
+echo "To verify deletion, run:"
+echo "  curl -s '${PROMETHEUS_URL}/api/v1/label/__name__/values' | jq '.data | map(select(startswith(\"epimetheus_benchmark\")))'"
+echo ""