From 88bc96a26e04a61c9909f6119d643c5ae6e7ceb2 Mon Sep 17 00:00:00 2001
From: zghh <1069308575@qq.com>
Date: Tue, 25 Jun 2024 21:46:23 +0800
Subject: [PATCH] fix the problem of configuring multiple gRPC endpoints where
 the first endpoint times out and cannot switch to the second endpoint

---
 WORKSPACE                                     | 10 +++++
 .../org_golang_google_grpc_clientconn.patch   | 41 +++++++++++++++++++
 2 files changed, 51 insertions(+)
 create mode 100644 third_party/org_golang_google_grpc_clientconn.patch

diff --git a/WORKSPACE b/WORKSPACE
index 46e8e7c3c017..30f2b3d09c5a 100644
--- a/WORKSPACE
+++ b/WORKSPACE
@@ -159,6 +159,16 @@ git_repository(
     # gazelle args: -go_prefix github.com/gogo/protobuf -proto legacy
 )
 
+git_repository(
+    name = "org_golang_google_grpc",
+    commit = "1055b481ed2204a29d233286b9b50c42b63f8825",
+    patch_args = ["-p1"],
+    patches = [
+        "//third_party:org_golang_google_grpc_clientconn.patch",
+    ],
+    remote = "https://github.com/grpc/grpc-go",
+)
+
 load("@rules_oci//oci:pull.bzl", "oci_pull")
 
 # A multi-arch base image
diff --git a/third_party/org_golang_google_grpc_clientconn.patch b/third_party/org_golang_google_grpc_clientconn.patch
new file mode 100644
index 000000000000..f1505c998c73
--- /dev/null
+++ b/third_party/org_golang_google_grpc_clientconn.patch
@@ -0,0 +1,41 @@
+diff --git a/clientconn.go b/clientconn.go
+index 95a7459b..a7acdabc 100644
+--- a/clientconn.go
++++ b/clientconn.go
+@@ -1323,18 +1323,11 @@ func (ac *addrConn) resetTransport() {
+ 		// Give dial more time as we keep failing to connect.
+ 		dialDuration = backoffFor
+ 	}
+-	// We can potentially spend all the time trying the first address, and
+-	// if the server accepts the connection and then hangs, the following
+-	// addresses will never be tried.
+-	//
+-	// The spec doesn't mention what should be done for multiple addresses.
+-	// https://github.com/grpc/grpc/blob/master/doc/connection-backoff.md#proposed-backoff-algorithm
+-	connectDeadline := time.Now().Add(dialDuration)
+ 
+ 	ac.updateConnectivityState(connectivity.Connecting, nil)
+ 	ac.mu.Unlock()
+ 
+-	if err := ac.tryAllAddrs(acCtx, addrs, connectDeadline); err != nil {
++	if err := ac.tryAllAddrs(acCtx, addrs, dialDuration); err != nil {
+ 		ac.cc.resolveNow(resolver.ResolveNowOptions{})
+ 		// After exhausting all addresses, the addrConn enters
+ 		// TRANSIENT_FAILURE.
+@@ -1377,7 +1370,7 @@ func (ac *addrConn) resetTransport() {
+ // tryAllAddrs tries to creates a connection to the addresses, and stop when at
+ // the first successful one. It returns an error if no address was successfully
+ // connected, or updates ac appropriately with the new transport.
+-func (ac *addrConn) tryAllAddrs(ctx context.Context, addrs []resolver.Address, connectDeadline time.Time) error {
++func (ac *addrConn) tryAllAddrs(ctx context.Context, addrs []resolver.Address, dialDuration time.Duration) error {
+ 	var firstConnErr error
+ 	for _, addr := range addrs {
+ 		if ctx.Err() != nil {
+@@ -1397,6 +1390,7 @@ func (ac *addrConn) tryAllAddrs(ctx context.Context, addrs []resolver.Address, c
+ 
+ 		channelz.Infof(logger, ac.channelzID, "Subchannel picks a new address %q to connect", addr.Addr)
+ 
++		connectDeadline := time.Now().Add(dialDuration)
+ 		err := ac.createTransport(ctx, addr, copts, connectDeadline)
+ 		if err == nil {
+ 			return nil