impl(GCS+gRPC): new options for gRPC metrics (#14094)

Author: coryan

google/cloud/storage/grpc_plugin.h

@@ -60,6 +60,44 @@ struct [[deprecated(
  */
 google::cloud::storage::Client DefaultGrpcClient(Options opts = {});
 
+/**
+ * Enable gRPC telemetry for GCS RPCs.
+ *
+ * Troubleshooting problems with GCS over gRPC is difficult without some
+ * telemetry indicating how the client is configured, and what load balancing
+ * information was available to the gRPC library.
+ *
+ * When this option is enabled (the default), the GCS client will export the
+ * gRPC telemetry discussed in [gRFC/66] and [gRFC/78] to
+ * [Google Cloud Monitoring]. Google Cloud Support can use this information to
+ * more quickly diagnose problems related to GCS and gRPC.
+ *
+ * Sending this data does not incur any billing charges, and requires minimal
+ * CPU (a single RPC every few minutes) or memory (a few KiB to batch the
+ * telemetry).
+ *
+ * [gRFC/66]: https://github.com/grpc/proposal/blob/master/A66-otel-stats.md
+ * [gRFC/78]:
+ * https://github.com/grpc/proposal/blob/master/A78-grpc-metrics-wrr-pf-xds.md
+ * [Google Cloud Monitoring]: https://cloud.google.com/monitoring/docs
+ */
+struct EnableGrpcMetrics {
+  using Type = bool;
+};
+
+/**
+ * gRPC telemetry export period.
+ *
+ * When `EnableGrpcMetrics` is enabled, this option controls the frequency at
+ * which metrics are exported to [Google Cloud Monitoring]. The default is 60
+ * seconds. Values below 60 seconds are ignored.
+ *
+ * [Google Cloud Monitoring]: https://cloud.google.com/monitoring/docs
+ */
+struct GrpcMetricsPeriod {
+  using Type = std::chrono::seconds;
+};
+
 GOOGLE_CLOUD_CPP_INLINE_NAMESPACE_END
 }  // namespace storage_experimental
 }  // namespace cloud

google/cloud/storage/internal/async/default_options.cc

@@ -18,7 +18,6 @@
 #include "google/cloud/storage/async/resume_policy.h"
 #include "google/cloud/storage/async/writer_connection.h"
 #include "google/cloud/storage/internal/grpc/default_options.h"
-#include "google/cloud/storage/internal/grpc/stub.h"
 #include <limits>
 
 namespace google {

google/cloud/storage/internal/grpc/default_options.cc

@@ -31,6 +31,9 @@ namespace storage_internal {
 GOOGLE_CLOUD_CPP_INLINE_NAMESPACE_BEGIN
 namespace {
 
+auto constexpr kMinMetricsPeriod = std::chrono::seconds(5);
+auto constexpr kDefaultMetricsPeriod = std::chrono::seconds(60);
+
 int DefaultGrpcNumChannels(std::string const& endpoint) {
   // When using DirectPath the gRPC library already does load balancing across
   // multiple sockets, it makes little sense to perform additional load
@@ -81,7 +84,15 @@ Options DefaultOptionsGrpc(Options options) {
       "storage.googleapis.com", options);
   options = google::cloud::internal::MergeOptions(
       std::move(options),
-      Options{}.set<EndpointOption>(ep).set<AuthorityOption>(ep));
+      Options{}
+          .set<EndpointOption>(ep)
+          .set<AuthorityOption>(ep)
+          .set<storage_experimental::EnableGrpcMetrics>(true)
+          .set<storage_experimental::GrpcMetricsPeriod>(kDefaultMetricsPeriod));
+  if (options.get<storage_experimental::GrpcMetricsPeriod>() <
+      kMinMetricsPeriod) {
+    options.set<storage_experimental::GrpcMetricsPeriod>(kMinMetricsPeriod);
+  }
   // We can only compute this once the endpoint is known, so take an additional
   // step.
   auto const num_channels =

google/cloud/storage/internal/grpc/default_options_test.cc

@@ -120,6 +120,31 @@ TEST(DefaultOptionsGrpc, DefaultOptionsUploadBuffer) {
   EXPECT_EQ(with_override, 256 * 1024L);
 }
 
+TEST(DefaultOptionsGrpc, MetricsEnabled) {
+  auto const options = DefaultOptionsGrpc(Options{});
+  EXPECT_TRUE(options.get<storage_experimental::EnableGrpcMetrics>());
+}
+
+TEST(DefaultOptionsGrpc, MetricsPeriod) {
+  auto const options = DefaultOptionsGrpc(Options{});
+  EXPECT_GE(options.get<storage_experimental::GrpcMetricsPeriod>(),
+            std::chrono::seconds(60));
+}
+
+TEST(DefaultOptionsGrpc, MinMetricsPeriod) {
+  auto const o0 =
+      DefaultOptionsGrpc(Options{}.set<storage_experimental::GrpcMetricsPeriod>(
+          std::chrono::seconds(0)));
+  EXPECT_GT(o0.get<storage_experimental::GrpcMetricsPeriod>(),
+            std::chrono::seconds(0));
+
+  auto const m5 =
+      DefaultOptionsGrpc(Options{}.set<storage_experimental::GrpcMetricsPeriod>(
+          std::chrono::seconds(-5)));
+  EXPECT_GT(m5.get<storage_experimental::GrpcMetricsPeriod>(),
+            std::chrono::seconds(0));
+}
+
 }  // namespace
 GOOGLE_CLOUD_CPP_INLINE_NAMESPACE_END
 }  // namespace storage_internal