diff --git a/docs/dev.md b/docs/dev.md
index a9f8d4c3..efd2023a 100644
--- a/docs/dev.md
+++ b/docs/dev.md
@@ -27,6 +27,7 @@ const(
     DEFAULT=2
     VERBOSE=3
     DEBUG=4
+    TRACE=5
 )
 ```
 
@@ -54,12 +55,12 @@ The guidelines are written in the context of a k8s controller. Our [ext-proc](..
    * `V(DEBUG)` and above: Use your best judgement. 
 
 4. Metric scraping loops. These loops run at a very high frequency, and logs can be very spammy if not handled properly.
-    * `klog.V(DEBUG).InfoS`
+    * `klog.V(TRACE).InfoS`
       * Transient errors/warnings, such as failure to get response from a pod.
       * Important state changes, such as updating a metric.
 
 5. Misc 
    1. Periodic (every 5s) debug loop which prints the current pods and metrics.
       * `klog.WarningS` If the metrics are not fresh enough, which indicates an error occurred during the metric scraping loop.
-      * `klog.V(VERBOSE).InfoS`
+      * `klog.V(DEBUG).InfoS`
          *  This is very important to debug the request scheduling algorithm, and yet not spammy compared to the metric scraping loop logs.
\ No newline at end of file
diff --git a/pkg/ext-proc/backend/provider.go b/pkg/ext-proc/backend/provider.go
index a97ca747..8bf67257 100644
--- a/pkg/ext-proc/backend/provider.go
+++ b/pkg/ext-proc/backend/provider.go
@@ -80,7 +80,7 @@ func (p *Provider) Init(refreshPodsInterval, refreshMetricsInterval time.Duratio
 		for {
 			time.Sleep(refreshMetricsInterval)
 			if err := p.refreshMetricsOnce(); err != nil {
-				klog.V(logutil.DEBUG).Infof("Failed to refresh metrics: %v", err)
+				klog.V(logutil.TRACE).Infof("Failed to refresh metrics: %v", err)
 			}
 		}
 	}()
@@ -135,12 +135,12 @@ func (p *Provider) refreshMetricsOnce() error {
 	defer func() {
 		d := time.Since(start)
 		// TODO: add a metric instead of logging
-		klog.V(logutil.DEBUG).Infof("Refreshed metrics in %v", d)
+		klog.V(logutil.TRACE).Infof("Refreshed metrics in %v", d)
 	}()
 	var wg sync.WaitGroup
 	errCh := make(chan error)
 	processOnePod := func(key, value any) bool {
-		klog.V(logutil.DEBUG).Infof("Processing pod %v and metric %v", key, value)
+		klog.V(logutil.TRACE).Infof("Processing pod %v and metric %v", key, value)
 		pod := key.(Pod)
 		existing := value.(*PodMetrics)
 		wg.Add(1)
@@ -152,7 +152,7 @@ func (p *Provider) refreshMetricsOnce() error {
 				return
 			}
 			p.UpdatePodMetrics(pod, updated)
-			klog.V(logutil.DEBUG).Infof("Updated metrics for pod %s: %v", pod, updated.Metrics)
+			klog.V(logutil.TRACE).Infof("Updated metrics for pod %s: %v", pod, updated.Metrics)
 		}()
 		return true
 	}
diff --git a/pkg/ext-proc/backend/vllm/metrics.go b/pkg/ext-proc/backend/vllm/metrics.go
index 68196d11..8800868a 100644
--- a/pkg/ext-proc/backend/vllm/metrics.go
+++ b/pkg/ext-proc/backend/vllm/metrics.go
@@ -171,6 +171,6 @@ func getLatestMetric(metricFamilies map[string]*dto.MetricFamily, metricName str
 			latest = m
 		}
 	}
-	klog.V(logutil.DEBUG).Infof("Got metric value %+v for metric %v", latest, metricName)
+	klog.V(logutil.TRACE).Infof("Got metric value %+v for metric %v", latest, metricName)
 	return latest, nil
 }
diff --git a/pkg/ext-proc/util/logging/logging_const.go b/pkg/ext-proc/util/logging/logging_const.go
index 57c44cf6..a6131d18 100644
--- a/pkg/ext-proc/util/logging/logging_const.go
+++ b/pkg/ext-proc/util/logging/logging_const.go
@@ -4,4 +4,5 @@ const (
 	DEFAULT = 2
 	VERBOSE = 3
 	DEBUG   = 4
+	TRACE   = 5
 )