From bb10e3a1c39ce49553393f6a25f41df768b2b5a6 Mon Sep 17 00:00:00 2001
From: Simo Lin <linsimo.mark@gmail.com>
Date: Sat, 16 Aug 2025 22:36:45 -0700
Subject: [PATCH] [router] fix pd prefill http request complinace issue (#9237)

---
 sgl-router/src/routers/pd_router.rs | 9 +++------
 1 file changed, 3 insertions(+), 6 deletions(-)

diff --git a/sgl-router/src/routers/pd_router.rs b/sgl-router/src/routers/pd_router.rs
index 1d0dd3a60..79589ef5d 100644
--- a/sgl-router/src/routers/pd_router.rs
+++ b/sgl-router/src/routers/pd_router.rs
@@ -886,12 +886,9 @@ impl PDRouter {
 
             tokio::spawn(async move {
                 if let Ok(response) = prefill_future.await {
-                    // Consume at most one small chunk with a very short timeout to advance flow control
-                    let _ = tokio::time::timeout(Duration::from_millis(20), async {
-                        let mut s = response.bytes_stream();
-                        let _ = s.next().await;
-                    })
-                    .await;
+                    // Consume the entire response body to maintain HTTP compliance
+                    // This runs in the background and won't block the decode response
+                    let _ = response.bytes().await;
                 }
             });