Optimized request prefill error messages

Signed-off-by: learner0810 <zhongjun.li@daocloud.io>

Author: learner0810

Dockerfile.sidecar

@@ -11,7 +11,7 @@ COPY go.mod go.mod
 COPY go.sum go.sum
 # cache deps before building and copying source so that we don't need to re-download as much
 # and so that source changes don't invalidate our downloaded layer
-RUN go mod download
+RUN go env -w GOPROXY=https://goproxy.cn && go mod download
 
 # Copy the go source
 COPY cmd/pd-sidecar/main.go cmd/cmd.go

cmd/pd-sidecar/main.go

@@ -43,7 +43,7 @@ var (
 
 func main() {
 	port := flag.String("port", "8000", "the port the sidecar is listening on")
-	vLLMPort := flag.String("vllm-port", "8001", "the port vLLM is listening on")
+	vLLMPort := flag.String("vllm-port", "30611", "the port vLLM is listening on")
 	vLLMDataParallelSize := flag.Int("data-parallel-size", 1, "the vLLM DATA-PARALLEL-SIZE value")
 	connector := flag.String("connector", proxy.ConnectorNIXLV2, "the P/D connector being used. Supported: "+strings.Join(supportedConnectors, ", "))
 	prefillerUseTLS := flag.Bool("prefiller-use-tls", false, "whether to use TLS when sending requests to prefillers")
@@ -120,7 +120,7 @@ func main() {
 	if *decoderUseTLS {
 		scheme = "https"
 	}
-	targetURL, err := url.Parse(scheme + "://localhost:" + *vLLMPort)
+	targetURL, err := url.Parse(scheme + "://10.20.100.240:" + *vLLMPort)
 	if err != nil {
 		logger.Error(err, "failed to create targetURL")
 		return

pkg/sidecar/proxy/connector_nixlv2.go

@@ -130,10 +130,26 @@ func (s *Server) runNIXLProtocolV2(w http.ResponseWriter, r *http.Request, prefi
 	)
 
 	if isHTTPError(pw.statusCode) {
-		s.logger.Error(err, "request failed", "code", pw.statusCode)
+		s.logger.Error(err, "request failed", "code", pw.statusCode, "body", pw.buffer.String())
 		prefillSpan.SetStatus(codes.Error, "prefill request failed")
 		prefillSpan.End()
-		w.WriteHeader(pw.statusCode)
+
+		if shouldFallbackToDecode(pw) {
+			s.logger.Info("fallback to decode", "request_id", uuidStr)
+			r.Body = io.NopCloser(strings.NewReader(string(original)))
+			s.decoderProxy.ServeHTTP(w, r)
+		} else {
+			for key, values := range pw.Header() {
+				for _, v := range values {
+					w.Header().Add(key, v)
+				}
+			}
+			w.WriteHeader(pw.statusCode)
+			_, err := w.Write([]byte(pw.buffer.String()))
+			if err != nil {
+				s.logger.Error(err, "failed to send error response to client")
+			}
+		}
 		return
 	}
 	prefillSpan.End()

pkg/sidecar/proxy/proxy_helpers.go

@@ -142,3 +142,11 @@ func (s *Server) createDecoderProxyHandler(decoderURL *url.URL, decoderInsecureS
 func isHTTPError(statusCode int) bool {
 	return statusCode < http.StatusOK || statusCode >= http.StatusMultipleChoices
 }
+
+// shouldFallbackToDecode returns false for client error 4xx status codes (400–451). For all other status codes, it returns true.
+func shouldFallbackToDecode(pw *bufferedResponseWriter) bool {
+	if pw.statusCode >= http.StatusBadRequest && pw.statusCode <= http.StatusUnavailableForLegalReasons {
+		return false
+	}
+	return true
+}