From 6ce02086a1775170e2c4d5d525f8f478ce478ffd Mon Sep 17 00:00:00 2001
From: Jerry Hu <hushenggang@selectdb.com>
Date: Tue, 26 Nov 2024 12:34:55 +0800
Subject: [PATCH] [opt](brpc) Make the timeout of brpc checking configurable
 (#44481)

---
 be/src/common/config.cpp        | 2 ++
 be/src/common/config.h          | 2 ++
 be/src/runtime/fragment_mgr.cpp | 5 ++++-
 3 files changed, 8 insertions(+), 1 deletion(-)

diff --git a/be/src/common/config.cpp b/be/src/common/config.cpp
index 0cffa30cdca443..63989a76261bb6 100644
--- a/be/src/common/config.cpp
+++ b/be/src/common/config.cpp
@@ -554,6 +554,8 @@ DEFINE_Bool(enable_brpc_builtin_services, "true");
 // Enable brpc connection check
 DEFINE_Bool(enable_brpc_connection_check, "false");
 
+DEFINE_mInt64(brpc_connection_check_timeout_ms, "10000");
+
 // The maximum amount of data that can be processed by a stream load
 DEFINE_mInt64(streaming_load_max_mb, "102400");
 // Some data formats, such as JSON, cannot be streamed.
diff --git a/be/src/common/config.h b/be/src/common/config.h
index caee1f320c103e..29e55e6406390e 100644
--- a/be/src/common/config.h
+++ b/be/src/common/config.h
@@ -1016,6 +1016,8 @@ DECLARE_Bool(enable_brpc_builtin_services);
 
 DECLARE_Bool(enable_brpc_connection_check);
 
+DECLARE_mInt64(brpc_connection_check_timeout_ms);
+
 // Max waiting time to wait the "plan fragment start" rpc.
 // If timeout, the fragment will be cancelled.
 // This parameter is usually only used when the FE loses connection,
diff --git a/be/src/runtime/fragment_mgr.cpp b/be/src/runtime/fragment_mgr.cpp
index 47b007edf5dc9b..1e72fa756d3dd3 100644
--- a/be/src/runtime/fragment_mgr.cpp
+++ b/be/src/runtime/fragment_mgr.cpp
@@ -1062,12 +1062,15 @@ void FragmentMgr::_check_brpc_available(const std::shared_ptr<PBackendService_St
     const std::string message = "hello doris!";
     std::string error_message;
     int32_t failed_count = 0;
+    const int64_t check_timeout_ms =
+            std::max<int64_t>(100, config::brpc_connection_check_timeout_ms);
+
     while (true) {
         PHandShakeRequest request;
         request.set_hello(message);
         PHandShakeResponse response;
         brpc::Controller cntl;
-        cntl.set_timeout_ms(500 * (failed_count + 1));
+        cntl.set_timeout_ms(check_timeout_ms);
         cntl.set_max_retry(10);
         brpc_stub->hand_shake(&cntl, &request, &response, nullptr);