Added max_requests_per_connection for downstream connection.

api/envoy/config/cluster/v3/cluster.proto

@@ -737,7 +737,11 @@ message Cluster {
   // is respected by both the HTTP/1.1 and HTTP/2 connection pool
   // implementations. If not specified, there is no limit. Setting this
   // parameter to 1 will effectively disable keep alive.
-  google.protobuf.UInt32Value max_requests_per_connection = 9;
+  //
+  // .. attention::
+  //   This field has been deprecated in favor of the :ref:`max_requests_connection <envoy_api_field_config.core.v3.HttpProtocolOptions.max_requests_connection>` field
+  google.protobuf.UInt32Value max_requests_per_connection = 9
+      [deprecated = true, (envoy.annotations.disallowed_by_default) = true];
 
   // Optional :ref:`circuit breaking <arch_overview_circuit_break>` for the cluster.
   CircuitBreakers circuit_breakers = 10;

api/envoy/config/core/v3/protocol.proto

@@ -40,7 +40,7 @@ message UpstreamHttpProtocolOptions {
   bool auto_san_validation = 2;
 }
 
-// [#next-free-field: 6]
+// [#next-free-field: 7]
 message HttpProtocolOptions {
   option (udpa.annotations.versioning).previous_message_type =
       "envoy.api.v2.core.HttpProtocolOptions";
@@ -104,6 +104,10 @@ message HttpProtocolOptions {
   // If this setting is not specified, the value defaults to ALLOW.
   // Note: upstream responses are not affected by this setting.
   HeadersWithUnderscoresAction headers_with_underscores_action = 5;
+
+  // Optional maximum requests for a downstream connection.
+  // If not specified, there is no limit.
+  google.protobuf.UInt32Value max_requests_connection = 6;
 }
 
 // [#next-free-field: 8]

docs/root/version_history/current.rst

@@ -76,6 +76,7 @@ New Features
 * grpc_json_transcoder: added option :ref:`strict_http_request_validation <envoy_v3_api_field_extensions.filters.http.grpc_json_transcoder.v3.GrpcJsonTranscoder.strict_http_request_validation>` to reject invalid requests early.
 * grpc_json_transcoder: filter can now be configured on per-route/per-vhost level as well. Leaving empty list of services in the filter configuration disables transcoding on the specific route.
 * http: added support for `Envoy::ScopeTrackedObject` for HTTP/1 dispatching. Crashes while inside the dispatching loop should dump debug information.
+* http: added support for :ref:`max_requests_connection <envoy_v3_api_field_config.core.v3.HttpProtocolOptions.max_requests_connection>` for downstream connection.
 * http: added support for :ref:`preconnecting <envoy_v3_api_msg_config.cluster.v3.Cluster.PreconnectPolicy>`. Preconnecting is off by default, but recommended for clusters serving latency-sensitive traffic, especially if using HTTP/1.1.
 * http: change frame flood and abuse checks to the upstream HTTP/2 codec to ON by default. It can be disabled by setting the `envoy.reloadable_features.upstream_http2_flood_checks` runtime key to false.
 * overload: add support for scaling :ref:`transport connection timeouts<envoy_v3_api_enum_value_config.overload.v3.ScaleTimersOverloadActionConfig.TimerType.TRANSPORT_SOCKET_CONNECT>`. This can be used to reduce the TLS handshake timeout in response to overload.
@@ -88,3 +89,4 @@ New Features
 
 Deprecated
 ----------
+* cluster: :ref:`max_requests_per_connection <envoy_v3_api_field_config.cluster.v3.Cluster.max_requests_per_connection>` is deprecated in favor of :ref:`max_requests_connection <envoy_v3_api_field_config.core.v3.HttpProtocolOptions.max_requests_connection>`.

generated_api_shadow/BUILD

@@ -146,6 +146,7 @@ proto_library(
         "//envoy/config/retry/omit_canary_hosts/v2:pkg",
         "//envoy/config/retry/omit_canary_hosts/v3:pkg",
         "//envoy/config/retry/previous_hosts/v2:pkg",
+        "//envoy/config/retry/previous_hosts/v3:pkg",
         "//envoy/config/route/v3:pkg",
         "//envoy/config/tap/v3:pkg",
         "//envoy/config/trace/v3:pkg",

source/common/http/http1/conn_pool.cc

@@ -73,15 +73,15 @@ void ActiveClient::StreamWrapper::onResetStream(StreamResetReason, absl::string_
 
 ActiveClient::ActiveClient(HttpConnPoolImplBase& parent)
     : Envoy::Http::ActiveClient(
-          parent, parent.host()->cluster().maxRequestsPerConnection(),
+          parent, parent.host()->cluster().commonHttpProtocolOptions().max_requests_connection(),
           1 // HTTP1 always has a concurrent-request-limit of 1 per connection.
       ) {
   parent.host()->cluster().stats().upstream_cx_http1_total_.inc();
 }
 
 ActiveClient::ActiveClient(HttpConnPoolImplBase& parent, Upstream::Host::CreateConnectionData& data)
     : Envoy::Http::ActiveClient(
-          parent, parent.host()->cluster().maxRequestsPerConnection(),
+          parent, parent.host()->cluster().commonHttpProtocolOptions().max_requests_connection(),
           1, // HTTP1 always has a concurrent-request-limit of 1 per connection.
           data) {
   parent.host()->cluster().stats().upstream_cx_http1_total_.inc();

source/common/http/http2/conn_pool.cc

@@ -57,7 +57,7 @@ uint64_t maxStreamsPerConnection(uint64_t max_streams_config) {
 
 ActiveClient::ActiveClient(HttpConnPoolImplBase& parent)
     : Envoy::Http::ActiveClient(
-          parent, maxStreamsPerConnection(parent.host()->cluster().maxRequestsPerConnection()),
+          parent, maxStreamsPerConnection(parent.host()->cluster().commonHttpProtocolOptions().max_requests_connection(),
           parent.host()->cluster().http2Options().max_concurrent_streams().value()) {
   codec_client_->setCodecClientCallbacks(*this);
   codec_client_->setCodecConnectionCallbacks(*this);
@@ -67,14 +67,16 @@ ActiveClient::ActiveClient(HttpConnPoolImplBase& parent)
 ActiveClient::ActiveClient(Envoy::Http::HttpConnPoolImplBase& parent,
                            Upstream::Host::CreateConnectionData& data)
     : Envoy::Http::ActiveClient(
-          parent, maxStreamsPerConnection(parent.host()->cluster().maxRequestsPerConnection()),
+          parent, maxStreamsPerConnection(parent.host()->cluster().commonHttpProtocolOptions().max_requests_connection()),
           parent.host()->cluster().http2Options().max_concurrent_streams().value(), data) {
   codec_client_->setCodecClientCallbacks(*this);
   codec_client_->setCodecConnectionCallbacks(*this);
   parent.host()->cluster().stats().upstream_cx_http2_total_.inc();
 }
 
-bool ActiveClient::closingWithIncompleteStream() const { return closed_with_active_rq_; }
+bool ActiveClient::closingWithIncompleteStream() const {
+  return closed_with_active_rq_;
+}
 
 RequestEncoder& ActiveClient::newStreamEncoder(ResponseDecoder& response_decoder) {
   return codec_client_->newStream(response_decoder);

source/common/tcp/conn_pool.cc

@@ -16,8 +16,9 @@ namespace Tcp {
 ActiveTcpClient::ActiveTcpClient(Envoy::ConnectionPool::ConnPoolImplBase& parent,
                                  const Upstream::HostConstSharedPtr& host,
                                  uint64_t concurrent_stream_limit)
-    : Envoy::ConnectionPool::ActiveClient(parent, host->cluster().maxRequestsPerConnection(),
-                                          concurrent_stream_limit),
+    : Envoy::ConnectionPool::ActiveClient(
+          parent, host->cluster().commonHttpProtocolOptions().max_requests_connection(),
+          concurrent_stream_limit),
       parent_(parent) {
   Upstream::Host::CreateConnectionData data = host->createConnection(
       parent_.dispatcher(), parent_.socketOptions(), parent_.transportSocketOptions());

source/common/tcp/original_conn_pool.cc

@@ -365,9 +365,9 @@ OriginalConnPoolImpl::PendingRequest::~PendingRequest() {
 
 OriginalConnPoolImpl::ActiveConn::ActiveConn(OriginalConnPoolImpl& parent)
     : parent_(parent),
-      connect_timer_(parent_.dispatcher_.createTimer([this]() -> void { onConnectTimeout(); })),
-      remaining_requests_(parent_.host_->cluster().maxRequestsPerConnection()), timed_out_(false) {
-
+      connect_timer_(parent_.dispatcher_.createTimer([this]() -> void {
+  onConnectTimeout(); })),
+      remaining_requests_(parent_.host_->cluster().commonHttpProtocolOptions().max_requests_connection(), timed_out_(false) {
   parent_.conn_connect_ms_ = std::make_unique<Stats::HistogramCompletableTimespanImpl>(
       parent_.host_->cluster().stats().upstream_cx_connect_ms_, parent_.dispatcher_.timeSource());