agents: implement gRPC retry policies

Implement client-side retry policies for gRPC connections to improve
resilience against transient network failures. The retry policy
automatically retries failed requests on UNAVAILABLE status with
exponential backoff.

Retry policy configuration:
- Max attempts: 5 retries
- Initial backoff: 500ms
- Max backoff: 5 seconds
- Backoff multiplier: 2.0x
- Retryable status: UNAVAILABLE

The retry policy is configured via gRPC service config JSON and applies
to all RPC methods. This ensures robust connectivity to the N|Solid
Console even during temporary network issues or server restarts.

Author: santigimeno

agents/grpc/src/grpc_agent.cc

@@ -80,6 +80,12 @@ const int CONSOLE_ID_SIZE = 36;
 
 const seconds DEFAULT_GRPC_TIMEOUT = seconds{ 60 };
 
+// Retry policy configuration
+constexpr size_t retry_max_attempts = 5;
+constexpr auto retry_initial_backoff = std::chrono::milliseconds(500);
+constexpr auto retry_max_backoff = std::chrono::seconds(5);
+constexpr float retry_backoff_multiplier = 2.0f;
+
 JSThreadMetrics::JSThreadMetrics(SharedEnvInst envinst):
     metrics_(ThreadMetrics::Create(envinst)) {
 }
@@ -1068,6 +1074,11 @@ int GrpcAgent::config(const json& config) {
         }
       }
 
+      opts.retry_policy_max_attempts = retry_max_attempts;
+      opts.retry_policy_initial_backoff = retry_initial_backoff;
+      opts.retry_policy_max_backoff = retry_max_backoff;
+      opts.retry_policy_backoff_multiplier = retry_backoff_multiplier;
+
       nsolid_service_stub_ =
           GrpcClient::MakeNSolidServiceStub(opts, tls_keylog_file_);
 

agents/grpc/src/grpc_client.cc

@@ -73,6 +73,40 @@ std::shared_ptr<Channel>
   grpc_arguments.SetInt(GRPC_ARG_KEEPALIVE_PERMIT_WITHOUT_CALLS, 1);
   grpc_arguments.SetInt(GRPC_ARG_HTTP2_MAX_PINGS_WITHOUT_DATA, 0);
 
+  static const auto kServiceConfigJson = std::string_view{R"(
+  {
+    "methodConfig": [
+      {
+        "name": [{}],
+        "retryPolicy": {
+          "maxAttempts": %0000000000u,
+          "initialBackoff": "%0000000000.1fs",
+          "maxBackoff": "%0000000000.1fs",
+          "backoffMultiplier": %0000000000.1f,
+          "retryableStatusCodes": [
+            "UNAVAILABLE"
+          ]
+        }
+      }
+    ]
+  })"};
+
+  // Allocate string with buffer large enough to hold the formatted json config
+  auto service_config = std::string(kServiceConfigJson.size(), '\0');
+  float initial_backoff = options.retry_policy_initial_backoff.count();
+  float max_backoff = options.retry_policy_max_backoff.count();
+  float backoff_multiplier = options.retry_policy_backoff_multiplier;
+  std::snprintf(
+    service_config.data(),
+    service_config.size(),
+    kServiceConfigJson.data(),
+    options.retry_policy_max_attempts,
+    std::min(std::max(initial_backoff, 0.f), 999999999.f),
+    std::min(std::max(max_backoff, 0.f), 999999999.f),
+    std::min(std::max(backoff_multiplier, 0.f), 999999999.f));
+
+  grpc_arguments.SetServiceConfigJSON(service_config);
+
   return CreateCustomChannel(options.endpoint,
                              MakeCredentials(options, tls_keylog_file),
                              grpc_arguments);

deps/opentelemetry-cpp/otlp-http-exporter.gyp

@@ -64,19 +64,21 @@
         'ENABLE_ASYNC_EXPORT',
         'ENABLE_OTLP_GRPC_CREDENTIAL_PREVIEW',
         'OPENTELEMETRY_STL_VERSION=2020',
+        'ENABLE_OTLP_RETRY_PREVIEW',
       ],
       'dependencies': [
         '../protobuf/protobuf.gyp:protobuf',
         '../curl/curl.gyp:curl',
         '../grpc/grpc.gyp:grpc++',
-	'../protobuf/abseil.gyp:abseil_proto',
+        '../protobuf/abseil.gyp:abseil_proto',
         '../zlib/zlib.gyp:zlib',
       ],
       'direct_dependent_settings': {
         'defines': [
           'ENABLE_ASYNC_EXPORT',
           'ENABLE_OTLP_GRPC_CREDENTIAL_PREVIEW',
           'OPENTELEMETRY_STL_VERSION=2020',
+          'ENABLE_OTLP_RETRY_PREVIEW',
         ],
         'include_dirs': [
           'api/include',