diff --git a/python/sglang/srt/model_executor/model_runner.py b/python/sglang/srt/model_executor/model_runner.py
index ac7c44569..677d71d8c 100644
--- a/python/sglang/srt/model_executor/model_runner.py
+++ b/python/sglang/srt/model_executor/model_runner.py
@@ -2164,6 +2164,12 @@ class ModelRunner(ModelRunnerKVCacheMixin):
     def forward_idle(
         self, forward_batch: ForwardBatch, pp_proxy_tensors=None
     ) -> Union[LogitsProcessorOutput, PPProxyTensors]:
+        # In DP Attention, IDLE batches are padded (batch_size > 0) for MLP sync.
+        # in this case, we need to reinit the forward metadata, otherwise the stale
+        # metadata causes batch_size mismatch in attention kernel(e.g. NSA Indexer).
+        if forward_batch.batch_size > 0:
+            self.attn_backend.init_forward_metadata(forward_batch)
+
         kwargs = {}
         if self.support_pp:
             kwargs["pp_proxy_tensors"] = pp_proxy_tensors