diff --git a/python/sglang/srt/managers/scheduler_metrics_mixin.py b/python/sglang/srt/managers/scheduler_metrics_mixin.py
index 8f5c02670..3d30f72fe 100644
--- a/python/sglang/srt/managers/scheduler_metrics_mixin.py
+++ b/python/sglang/srt/managers/scheduler_metrics_mixin.py
@@ -440,11 +440,12 @@ class SchedulerMetricsMixin:
     def log_decode_stats_every_iteration(
         self: Scheduler, batch: ScheduleBatch, num_accepted_tokens: int
     ):
-        self.metrics_collector.increment_realtime_tokens(
-            # TODO unify this w/ the bumping logic in `Scheduler.num_generated_tokens` accumulator
-            decode_tokens=batch.batch_size() + num_accepted_tokens,
-            dp_cooperation_info=batch.dp_cooperation_info,
-        )
+        if self.enable_metrics:
+            self.metrics_collector.increment_realtime_tokens(
+                # TODO unify this w/ the bumping logic in `Scheduler.num_generated_tokens` accumulator
+                decode_tokens=batch.batch_size() + num_accepted_tokens,
+                dp_cooperation_info=batch.dp_cooperation_info,
+            )
 
     def log_batch_result_stats(
         self: Scheduler,
diff --git a/python/sglang/srt/managers/scheduler_output_processor_mixin.py b/python/sglang/srt/managers/scheduler_output_processor_mixin.py
index e40586c24..ed614fea9 100644
--- a/python/sglang/srt/managers/scheduler_output_processor_mixin.py
+++ b/python/sglang/srt/managers/scheduler_output_processor_mixin.py
@@ -470,12 +470,9 @@ class SchedulerOutputProcessorMixin:
         self.token_to_kv_pool_allocator.free_group_end()
 
         self.forward_ct_decode = (self.forward_ct_decode + 1) % (1 << 30)
-        if (
-            self.current_scheduler_metrics_enabled
-            and self.forward_ct_decode % self.server_args.decode_log_interval == 0
-        ):
-            self.log_decode_stats(can_run_cuda_graph, running_batch=batch)
-        if self.enable_metrics:
+        if self.current_scheduler_metrics_enabled:
+            if self.forward_ct_decode % self.server_args.decode_log_interval == 0:
+                self.log_decode_stats(can_run_cuda_graph, running_batch=batch)
             self.log_decode_stats_every_iteration(
                 batch, num_accepted_tokens=result.num_accepted_tokens
             )