xDS: add version gauge

docs/configuration/cluster_manager/cds.rst

@@ -59,6 +59,8 @@ CDS has a statistics tree rooted at *cluster_manager.cds.* with the following st
   :header: Name, Type, Description
   :widths: 1, 1, 2
 
+  config_reload, Counter, Total API fetches that resulted in a config reload due to a different config
   update_attempt, Counter, Total API fetches attempted
   update_success, Counter, Total API fetches completed successfully
   update_failure, Counter, Total API fetches that failed (either network or schema errors)
+  version, Gauge, Hash of the contents from the last successful API fetch

docs/configuration/http_conn_man/rds.rst

@@ -69,7 +69,7 @@ will only perform a full reload if the hash value changes.
 Statistics
 ----------
 
-RDS has a statistics tree rooted at *http.<stat_prefix>.rds.* with the following statistics:
+RDS has a statistics tree rooted at *http.<stat_prefix>.rds.<route_config_name>* with the following statistics:
 
 .. csv-table::
   :header: Name, Type, Description
@@ -79,3 +79,4 @@ RDS has a statistics tree rooted at *http.<stat_prefix>.rds.* with the following
   update_attempt, Counter, Total API fetches attempted
   update_success, Counter, Total API fetches completed successfully
   update_failure, Counter, Total API fetches that failed (either network or schema errors)
+  version, Gauge, Hash of the contents from the last successful API fetch

docs/configuration/listeners/lds.rst

@@ -77,6 +77,8 @@ LDS has a statistics tree rooted at *listener_manager.lds.* with the following s
   :header: Name, Type, Description
   :widths: 1, 1, 2
 
+  config_reload, Counter, Total API fetches that resulted in a config reload due to a different config
   update_attempt, Counter, Total API fetches attempted
   update_success, Counter, Total API fetches completed successfully
   update_failure, Counter, Total API fetches that failed (either network or schema errors)
+  version, Gauge, Hash of the contents from the last successful API fetch

include/envoy/config/subscription.h

@@ -75,18 +75,19 @@ template <class ResourceType> class Subscription {
  * Per subscription stats. @see stats_macros.h
  */
 // clang-format off
-#define ALL_SUBSCRIPTION_STATS(COUNTER) \
+#define ALL_SUBSCRIPTION_STATS(COUNTER, GAUGE) \
   COUNTER(update_attempt)               \
   COUNTER(update_success)               \
   COUNTER(update_failure)               \
-  COUNTER(update_rejected)
+  COUNTER(update_rejected)              \
+  GAUGE(version)
 // clang-format on
 
 /**
  * Struct definition for per subscription stats. @see stats_macros.h
  */
 struct SubscriptionStats {
-  ALL_SUBSCRIPTION_STATS(GENERATE_COUNTER_STRUCT)
+  ALL_SUBSCRIPTION_STATS(GENERATE_COUNTER_STRUCT, GENERATE_GAUGE_STRUCT)
 };
 
 } // namespace Config

source/common/config/filesystem_subscription_impl.h

@@ -63,6 +63,7 @@ class FilesystemSubscriptionImpl : public Config::Subscription<ResourceType>,
       config_update_available = true;
       callbacks_->onConfigUpdate(typed_resources);
       version_info_ = message.version_info();
+      stats_.version_.set(HashUtil::xxHash64(version_info_));
       stats_.update_success_.inc();
       ENVOY_LOG(debug, "Filesystem config update accepted for {}: {}", path_,
                 message.DebugString());

source/common/config/grpc_mux_subscription_impl.h

@@ -55,6 +55,7 @@ class GrpcMuxSubscriptionImpl : public Subscription<ResourceType>,
     stats_.update_success_.inc();
     stats_.update_attempt_.inc();
     version_info_ = version_info;
+    stats_.version_.set(HashUtil::xxHash64(version_info_));
     ENVOY_LOG(debug, "gRPC config for {} accepted with {} resources", type_url_, resources.size());
 
 #ifndef NVLOG

source/common/config/http_subscription_impl.h

@@ -60,7 +60,7 @@ class HttpSubscriptionImpl : public Http::RestApiFetcher,
     request_.mutable_resource_names()->Swap(&resources_vector);
   }
 
-  const std::string versionInfo() const override { return request_.version_info(); }
+  const std::string versionInfo() const override { return version_info_; }
 
   // Http::RestApiFetcher
   void createRequest(Http::Message& request) override {
@@ -83,6 +83,8 @@ class HttpSubscriptionImpl : public Http::RestApiFetcher,
     try {
       callbacks_->onConfigUpdate(typed_resources);
       request_.set_version_info(message.version_info());
+      version_info_ = message.version_info();
+      stats_.version_.set(HashUtil::xxHash64(version_info_));
       stats_.update_success_.inc();
     } catch (const EnvoyException& e) {
       ENVOY_LOG(warn, "REST config update rejected: {}", e.what());
@@ -105,6 +107,7 @@ class HttpSubscriptionImpl : public Http::RestApiFetcher,
   }
 
   std::string path_;
+  std::string version_info_;
   Protobuf::RepeatedPtrField<ProtobufTypes::String> resources_;
   Config::SubscriptionCallbacks<ResourceType>* callbacks_{};
   envoy::api::v2::DiscoveryRequest request_;

source/common/config/utility.h

@@ -5,6 +5,7 @@
 #include "envoy/json/json_object.h"
 #include "envoy/local_info/local_info.h"
 #include "envoy/registry/registry.h"
+#include "envoy/stats/stats.h"
 #include "envoy/upstream/cluster_manager.h"
 
 #include "common/common/assert.h"
@@ -149,7 +150,7 @@ class Utility {
    * @return SubscriptionStats for scope.
    */
   static SubscriptionStats generateStats(Stats::Scope& scope) {
-    return {ALL_SUBSCRIPTION_STATS(POOL_COUNTER(scope))};
+    return {ALL_SUBSCRIPTION_STATS(POOL_COUNTER(scope), POOL_GAUGE(scope))};
   }
 
   /**

source/common/router/rds_impl.cc

@@ -47,7 +47,8 @@ RdsRouteConfigProviderImpl::RdsRouteConfigProviderImpl(
     const std::string& stat_prefix, ThreadLocal::SlotAllocator& tls,
     RouteConfigProviderManagerImpl& route_config_provider_manager)
     : runtime_(runtime), cm_(cm), tls_(tls.allocateSlot()),
-      route_config_name_(rds.route_config_name()), scope_(scope.createScope(stat_prefix + "rds.")),
+      route_config_name_(rds.route_config_name()),
+      scope_(scope.createScope(stat_prefix + "rds." + route_config_name_ + ".")),
       stats_({ALL_RDS_STATS(POOL_COUNTER(*scope_))}),
       route_config_provider_manager_(route_config_provider_manager),
       manager_identifier_(manager_identifier) {

source/common/router/rds_subscription.cc

@@ -50,6 +50,7 @@ void RdsSubscription::parseResponse(const Http::Message& response) {
   resources[0].set_name(route_config_name_);
   callbacks_->onConfigUpdate(resources);
   version_info_ = Envoy::Config::Utility::computeHashedVersion(response_body);
+  stats_.version_.set(HashUtil::xxHash64(version_info_));
   stats_.update_success_.inc();
 }
 

source/common/upstream/cds_subscription.cc

@@ -53,6 +53,7 @@ void CdsSubscription::parseResponse(const Http::Message& response) {
 
   callbacks_->onConfigUpdate(resources);
   version_info_ = Config::Utility::computeHashedVersion(response_body);
+  stats_.version_.set(HashUtil::xxHash64(version_info_));
   stats_.update_success_.inc();
 }
 

source/server/lds_subscription.cc

@@ -53,6 +53,7 @@ void LdsSubscription::parseResponse(const Http::Message& response) {
 
   callbacks_->onConfigUpdate(resources);
   version_info_ = Config::Utility::computeHashedVersion(response_body);
+  stats_.version_.set(HashUtil::xxHash64(version_info_));
   stats_.update_success_.inc();
 }
 

test/common/config/filesystem_subscription_impl_test.cc

@@ -12,19 +12,19 @@ class FilesystemSubscriptionImplTest : public FilesystemSubscriptionTestHarness,
 // Validate that the client can recover from bad JSON responses.
 TEST_F(FilesystemSubscriptionImplTest, BadJsonRecovery) {
   startSubscription({"cluster0", "cluster1"});
-  verifyStats(1, 0, 0, 0);
+  verifyStats(1, 0, 0, 0, 0);
   EXPECT_CALL(callbacks_, onConfigUpdateFailed(_));
   updateFile(";!@#badjso n");
-  verifyStats(2, 0, 0, 1);
+  verifyStats(2, 0, 0, 1, 0);
   deliverConfigUpdate({"cluster0", "cluster1"}, "0", true);
-  verifyStats(3, 1, 0, 1);
+  verifyStats(3, 1, 0, 1, 7148434200721666028);
 }
 
 // Validate that a file that is initially available results in a successful update.
 TEST_F(FilesystemSubscriptionImplTest, InitialFile) {
   updateFile("{\"versionInfo\": \"0\", \"resources\": []}", false);
   startSubscription({"cluster0", "cluster1"});
-  verifyStats(1, 1, 0, 0);
+  verifyStats(1, 1, 0, 0, 7148434200721666028);
 }
 
 } // namespace

test/common/config/filesystem_subscription_test_harness.h

@@ -83,11 +83,11 @@ class FilesystemSubscriptionTestHarness : public SubscriptionTestHarness {
     EXPECT_EQ(version_, subscription_.versionInfo());
   }
 
-  void verifyStats(uint32_t attempt, uint32_t success, uint32_t rejected,
-                   uint32_t failure) override {
+  void verifyStats(uint32_t attempt, uint32_t success, uint32_t rejected, uint32_t failure,
+                   uint64_t version) override {
     // The first attempt always fail unless there was a file there to begin with.
     SubscriptionTestHarness::verifyStats(attempt, success, rejected,
-                                         failure + (file_at_start_ ? 0 : 1));
+                                         failure + (file_at_start_ ? 0 : 1), version);
   }
 
   const std::string path_;

test/common/config/grpc_subscription_impl_test.cc

@@ -17,54 +17,54 @@ TEST_F(GrpcSubscriptionImplTest, StreamCreationFailure) {
   EXPECT_CALL(callbacks_, onConfigUpdateFailed(_));
   EXPECT_CALL(*timer_, enableTimer(_));
   subscription_->start({"cluster0", "cluster1"}, callbacks_);
-  verifyStats(2, 0, 0, 1);
+  verifyStats(2, 0, 0, 1, 0);
   // Ensure this doesn't cause an issue by sending a request, since we don't
   // have a gRPC stream.
   subscription_->updateResources({"cluster2"});
   // Retry and succeed.
   EXPECT_CALL(*async_client_, start(_, _)).WillOnce(Return(&async_stream_));
   expectSendMessage({"cluster2"}, "");
   timer_cb_();
-  verifyStats(3, 0, 0, 1);
+  verifyStats(3, 0, 0, 1, 0);
 }
 
 // Validate that the client can recover from a remote stream closure via retry.
 TEST_F(GrpcSubscriptionImplTest, RemoteStreamClose) {
   startSubscription({"cluster0", "cluster1"});
-  verifyStats(1, 0, 0, 0);
+  verifyStats(1, 0, 0, 0, 0);
   Http::HeaderMapPtr trailers{new Http::TestHeaderMapImpl{}};
   subscription_->grpcMux().onReceiveTrailingMetadata(std::move(trailers));
   EXPECT_CALL(*timer_, enableTimer(_));
   EXPECT_CALL(callbacks_, onConfigUpdateFailed(_));
   subscription_->grpcMux().onRemoteClose(Grpc::Status::GrpcStatus::Canceled, "");
-  verifyStats(2, 0, 0, 1);
+  verifyStats(2, 0, 0, 1, 0);
   // Retry and succeed.
   EXPECT_CALL(*async_client_, start(_, _)).WillOnce(Return(&async_stream_));
   expectSendMessage({"cluster0", "cluster1"}, "");
   timer_cb_();
-  verifyStats(2, 0, 0, 1);
+  verifyStats(2, 0, 0, 1, 0);
 }
 
 // Validate that When the management server gets multiple requests for the same version, it can
 // ignore later ones. This allows the nonce to be used.
 TEST_F(GrpcSubscriptionImplTest, RepeatedNonce) {
   InSequence s;
   startSubscription({"cluster0", "cluster1"});
-  verifyStats(1, 0, 0, 0);
+  verifyStats(1, 0, 0, 0, 0);
   // First with the initial, empty version update to "0".
   updateResources({"cluster2"});
-  verifyStats(2, 0, 0, 0);
+  verifyStats(2, 0, 0, 0, 0);
   deliverConfigUpdate({"cluster0", "cluster2"}, "0", false);
-  verifyStats(3, 0, 1, 0);
+  verifyStats(3, 0, 1, 0, 0);
   deliverConfigUpdate({"cluster0", "cluster2"}, "0", true);
-  verifyStats(4, 1, 1, 0);
+  verifyStats(4, 1, 1, 0, 7148434200721666028);
   // Now with version "0" update to "1".
   updateResources({"cluster3"});
-  verifyStats(5, 1, 1, 0);
+  verifyStats(5, 1, 1, 0, 7148434200721666028);
   deliverConfigUpdate({"cluster3"}, "1", false);
-  verifyStats(6, 1, 2, 0);
+  verifyStats(6, 1, 2, 0, 7148434200721666028);
   deliverConfigUpdate({"cluster3"}, "1", true);
-  verifyStats(7, 2, 2, 0);
+  verifyStats(7, 2, 2, 0, 13237225503670494420U);
 }
 
 } // namespace

test/common/config/http_subscription_impl_test.cc

@@ -15,11 +15,11 @@ TEST_F(HttpSubscriptionImplTest, OnRequestReset) {
   EXPECT_CALL(*timer_, enableTimer(_));
   EXPECT_CALL(callbacks_, onConfigUpdateFailed(_));
   http_callbacks_->onFailure(Http::AsyncClient::FailureReason::Reset);
-  verifyStats(1, 0, 0, 1);
+  verifyStats(1, 0, 0, 1, 0);
   timerTick();
-  verifyStats(2, 0, 0, 1);
+  verifyStats(2, 0, 0, 1, 0);
   deliverConfigUpdate({"cluster0", "cluster1"}, "0", true);
-  verifyStats(3, 1, 0, 1);
+  verifyStats(3, 1, 0, 1, 7148434200721666028);
 }
 
 // Validate that the client can recover from bad JSON responses.
@@ -32,12 +32,12 @@ TEST_F(HttpSubscriptionImplTest, BadJsonRecovery) {
   EXPECT_CALL(*timer_, enableTimer(_));
   EXPECT_CALL(callbacks_, onConfigUpdateFailed(_));
   http_callbacks_->onSuccess(std::move(message));
-  verifyStats(1, 0, 0, 1);
+  verifyStats(1, 0, 0, 1, 0);
   request_in_progress_ = false;
   timerTick();
-  verifyStats(2, 0, 0, 1);
+  verifyStats(2, 0, 0, 1, 0);
   deliverConfigUpdate({"cluster0", "cluster1"}, "0", true);
-  verifyStats(3, 1, 0, 1);
+  verifyStats(3, 1, 0, 1, 7148434200721666028);
 }
 
 } // namespace

test/common/config/subscription_impl_test.cc

@@ -42,8 +42,9 @@ class SubscriptionImplTest : public testing::TestWithParam<SubscriptionType> {
     test_harness_->expectSendMessage(cluster_names, version);
   }
 
-  void verifyStats(uint32_t attempt, uint32_t success, uint32_t rejected, uint32_t failure) {
-    test_harness_->verifyStats(attempt, success, rejected, failure);
+  void verifyStats(uint32_t attempt, uint32_t success, uint32_t rejected, uint32_t failure,
+                   uint64_t version) {
+    test_harness_->verifyStats(attempt, success, rejected, failure, version);
   }
 
   void deliverConfigUpdate(const std::vector<std::string> cluster_names, const std::string& version,
@@ -61,57 +62,57 @@ INSTANTIATE_TEST_CASE_P(SubscriptionImplTest, SubscriptionImplTest,
 // Validate basic request-response succeeds.
 TEST_P(SubscriptionImplTest, InitialRequestResponse) {
   startSubscription({"cluster0", "cluster1"});
-  verifyStats(1, 0, 0, 0);
+  verifyStats(1, 0, 0, 0, 0);
   deliverConfigUpdate({"cluster0", "cluster1"}, "0", true);
-  verifyStats(2, 1, 0, 0);
+  verifyStats(2, 1, 0, 0, 7148434200721666028);
 }
 
 // Validate that multiple streamed updates succeed.
 TEST_P(SubscriptionImplTest, ResponseStream) {
   startSubscription({"cluster0", "cluster1"});
-  verifyStats(1, 0, 0, 0);
+  verifyStats(1, 0, 0, 0, 0);
   deliverConfigUpdate({"cluster0", "cluster1"}, "0", true);
-  verifyStats(2, 1, 0, 0);
+  verifyStats(2, 1, 0, 0, 7148434200721666028);
   deliverConfigUpdate({"cluster0", "cluster1"}, "1", true);
-  verifyStats(3, 2, 0, 0);
+  verifyStats(3, 2, 0, 0, 13237225503670494420U);
 }
 
 // Validate that the client can reject a config.
 TEST_P(SubscriptionImplTest, RejectConfig) {
   startSubscription({"cluster0", "cluster1"});
-  verifyStats(1, 0, 0, 0);
+  verifyStats(1, 0, 0, 0, 0);
   deliverConfigUpdate({"cluster0", "cluster1"}, "0", false);
-  verifyStats(2, 0, 1, 0);
+  verifyStats(2, 0, 1, 0, 0);
 }
 
 // Validate that the client can reject a config and accept the same config later.
 TEST_P(SubscriptionImplTest, RejectAcceptConfig) {
   startSubscription({"cluster0", "cluster1"});
-  verifyStats(1, 0, 0, 0);
+  verifyStats(1, 0, 0, 0, 0);
   deliverConfigUpdate({"cluster0", "cluster1"}, "0", false);
-  verifyStats(2, 0, 1, 0);
+  verifyStats(2, 0, 1, 0, 0);
   deliverConfigUpdate({"cluster0", "cluster1"}, "0", true);
-  verifyStats(3, 1, 1, 0);
+  verifyStats(3, 1, 1, 0, 7148434200721666028);
 }
 
 // Validate that the client can reject a config and accept another config later.
 TEST_P(SubscriptionImplTest, RejectAcceptNextConfig) {
   startSubscription({"cluster0", "cluster1"});
-  verifyStats(1, 0, 0, 0);
+  verifyStats(1, 0, 0, 0, 0);
   deliverConfigUpdate({"cluster0", "cluster1"}, "0", false);
-  verifyStats(2, 0, 1, 0);
+  verifyStats(2, 0, 1, 0, 0);
   deliverConfigUpdate({"cluster0", "cluster1"}, "1", true);
-  verifyStats(3, 1, 1, 0);
+  verifyStats(3, 1, 1, 0, 13237225503670494420U);
 }
 
 // Validate that stream updates send a message with the updated resources.
 TEST_P(SubscriptionImplTest, UpdateResources) {
   startSubscription({"cluster0", "cluster1"});
-  verifyStats(1, 0, 0, 0);
+  verifyStats(1, 0, 0, 0, 0);
   deliverConfigUpdate({"cluster0", "cluster1"}, "0", true);
-  verifyStats(2, 1, 0, 0);
+  verifyStats(2, 1, 0, 0, 7148434200721666028);
   updateResources({"cluster2"});
-  verifyStats(3, 1, 0, 0);
+  verifyStats(3, 1, 0, 0, 7148434200721666028);
 }
 
 } // namespace

test/common/config/subscription_test_harness.h

@@ -49,12 +49,13 @@ class SubscriptionTestHarness {
   virtual void deliverConfigUpdate(const std::vector<std::string> cluster_names,
                                    const std::string& version, bool accept) PURE;
 
-  virtual void verifyStats(uint32_t attempt, uint32_t success, uint32_t rejected,
-                           uint32_t failure) {
+  virtual void verifyStats(uint32_t attempt, uint32_t success, uint32_t rejected, uint32_t failure,
+                           uint64_t version) {
     EXPECT_EQ(attempt, stats_.update_attempt_.value());
     EXPECT_EQ(success, stats_.update_success_.value());
     EXPECT_EQ(rejected, stats_.update_rejected_.value());
     EXPECT_EQ(failure, stats_.update_failure_.value());
+    EXPECT_EQ(version, stats_.version_.value());
   }
 
   Stats::MockIsolatedStatsStore stats_store_;