Add improvements to default server metrics (#4548)

## Motivation and Context
<!--- Why is this change required? What problem does it solve? -->
<!--- If it fixes an open issue, please link to the issue here -->

## Description
<!--- Describe your changes in detail -->

Add success metric to default server metrics, align naming, add method
for disabling outstanding_requests metric.

## Testing
<!--- Please describe in detail how you tested your changes -->
<!--- Include details of your testing environment, and the tests you ran
to -->
<!--- see how your change affects other areas of the code, etc. -->

## Checklist
<!--- If a checkbox below is not applicable, then please DELETE it
rather than leaving it unchecked -->
- [ ] For changes to the smithy-rs codegen or runtime crates, I have
created a changelog entry Markdown file in the `.changelog` directory,
specifying "client," "server," or both in the `applies_to` key.
- [ ] For changes to the AWS SDK, generated SDK code, or SDK runtime
crates, I have created a changelog entry Markdown file in the
`.changelog` directory, specifying "aws-sdk-rust" in the `applies_to`
key.

----

_By submitting this pull request, I confirm that you can use, modify,
copy, and redistribute this contribution, under the terms of your
choice._

Author: jasgin

CODEOWNERS

@@ -5,6 +5,8 @@
 /codegen-server/                                    @smithy-lang/smithy-rs-server
 /rust-runtime/aws-smithy-http-server/               @smithy-lang/smithy-rs-server
 /rust-runtime/aws-smithy-legacy-http-server/        @smithy-lang/smithy-rs-server
+/rust-runtime/aws-smithy-http-server-metrics        @smithy-lang/smithy-rs-server
+/rust-runtime/aws-smithy-http-server-metrics-macro  @smithy-lang/smithy-rs-server
 
 # Python Server
 /codegen-server-test/python/                        @smithy-lang/smithy-rs-server

examples/pokemon-service/tests/snapshots/metrics_test__metrics_content_via_tcp.snap

@@ -2,6 +2,6 @@
 source: pokemon-service/tests/metrics_test.rs
 expression: metrics_output
 ---
-{"_aws":{"CloudWatchMetrics":[{"Namespace":"Ns","Dimensions":[[]],"Metrics":[{"Name":"found"},{"Name":"outstanding_requests"},{"Name":"http_status_code"},{"Name":"error"},{"Name":"fault"},{"Name":"operation_time","Unit":"Milliseconds"}]}],"Timestamp":"[timestamp]},"found":1,"outstanding_requests":1,"http_status_code":200,"error":0,"fault":0,"operation_time":"[operation_time],"test_request_metric":"test request metric","test_response_metric":"test response metric","requested_pokemon_name":"pikachu","service":"PokemonService","service_version":"2024-03-18","operation":"GetPokemonSpecies","request_id":"[request_id]}
-{"_aws":{"CloudWatchMetrics":[{"Namespace":"Ns","Dimensions":[[]],"Metrics":[{"Name":"authenticated"},{"Name":"outstanding_requests"},{"Name":"http_status_code"},{"Name":"error"},{"Name":"fault"},{"Name":"operation_time","Unit":"Milliseconds"}]}],"Timestamp":"[timestamp]},"authenticated":0,"outstanding_requests":1,"http_status_code":401,"error":1,"fault":0,"operation_time":"[operation_time],"test_request_metric":"test request metric","test_response_metric":"test response metric","user":"ash","service":"PokemonService","service_version":"2024-03-18","operation":"GetStorage","request_id":"[request_id]}
-{"_aws":{"CloudWatchMetrics":[{"Namespace":"Ns","Dimensions":[[]],"Metrics":[{"Name":"supported_region"},{"Name":"outstanding_requests"},{"Name":"http_status_code"},{"Name":"error"},{"Name":"fault"},{"Name":"operation_time","Unit":"Milliseconds"}]}],"Timestamp":"[timestamp]},"supported_region":0,"outstanding_requests":1,"http_status_code":500,"error":0,"fault":1,"operation_time":"[operation_time],"test_request_metric":"test request metric","test_response_metric":"test response metric","requested_region":"trigger500","service":"PokemonService","service_version":"2024-03-18","operation":"CapturePokemon","request_id":"[request_id]}
+{"_aws":{"CloudWatchMetrics":[{"Namespace":"Ns","Dimensions":[[]],"Metrics":[{"Name":"found"},{"Name":"outstanding_requests"},{"Name":"success"},{"Name":"client_error"},{"Name":"server_error"},{"Name":"operation_time","Unit":"Milliseconds"}]}],"Timestamp":"[timestamp]},"found":1,"outstanding_requests":1,"success":1,"client_error":0,"server_error":0,"operation_time":"[operation_time],"test_request_metric":"test request metric","test_response_metric":"test response metric","requested_pokemon_name":"pikachu","service":"PokemonService","service_version":"2024-03-18","operation":"GetPokemonSpecies","request_id":"[request_id],"http_status_code":"200"}
+{"_aws":{"CloudWatchMetrics":[{"Namespace":"Ns","Dimensions":[[]],"Metrics":[{"Name":"authenticated"},{"Name":"outstanding_requests"},{"Name":"success"},{"Name":"client_error"},{"Name":"server_error"},{"Name":"operation_time","Unit":"Milliseconds"}]}],"Timestamp":"[timestamp]},"authenticated":0,"outstanding_requests":1,"success":0,"client_error":1,"server_error":0,"operation_time":"[operation_time],"test_request_metric":"test request metric","test_response_metric":"test response metric","user":"ash","service":"PokemonService","service_version":"2024-03-18","operation":"GetStorage","request_id":"[request_id],"http_status_code":"401"}
+{"_aws":{"CloudWatchMetrics":[{"Namespace":"Ns","Dimensions":[[]],"Metrics":[{"Name":"supported_region"},{"Name":"outstanding_requests"},{"Name":"success"},{"Name":"client_error"},{"Name":"server_error"},{"Name":"operation_time","Unit":"Milliseconds"}]}],"Timestamp":"[timestamp]},"supported_region":0,"outstanding_requests":1,"success":0,"client_error":0,"server_error":1,"operation_time":"[operation_time],"test_request_metric":"test request metric","test_response_metric":"test response metric","requested_region":"trigger500","service":"PokemonService","service_version":"2024-03-18","operation":"CapturePokemon","request_id":"[request_id],"http_status_code":"500"}

rust-runtime/aws-smithy-http-server-metrics/src/default.rs

@@ -74,11 +74,13 @@ pub struct DefaultRequestMetrics {
 #[metrics]
 #[derive(Default, Debug)]
 pub struct DefaultResponseMetrics {
-    pub(crate) http_status_code: Option<u16>,
+    pub(crate) http_status_code: Option<String>,
+    /// Success indicator (1 if 2xx status code, 0 otherwise)
+    pub(crate) success: Option<bool>,
     /// Client error indicator (1 if 4xx status code, 0 otherwise)
-    pub(crate) error: Option<u64>,
-    /// Server fault indicator (1 if 5xx status code, 0 otherwise)
-    pub(crate) fault: Option<u64>,
+    pub(crate) client_error: Option<bool>,
+    /// Server error indicator (1 if 5xx status code, 0 otherwise)
+    pub(crate) server_error: Option<bool>,
     /// Wallclock time from pre-deserialization of the model input to post-serialization of the model output
     pub(crate) operation_time: Option<Duration>,
 }
@@ -103,8 +105,9 @@ pub struct DefaultRequestMetricsConfig {
 pub struct DefaultResponseMetricsConfig {
     pub(crate) disable_all: bool,
     pub(crate) disable_http_status_code: bool,
-    pub(crate) disable_error: bool,
-    pub(crate) disable_fault: bool,
+    pub(crate) disable_success: bool,
+    pub(crate) disable_client_error: bool,
+    pub(crate) disable_server_error: bool,
     pub(crate) disable_operation_time: bool,
 }
 

rust-runtime/aws-smithy-http-server-metrics/src/layer/builder.rs

@@ -44,50 +44,62 @@ macro_rules! impl_disable_methods {
             self
         }
 
-        /// Disable the `request_id` metric
+        /// Disable the default `outstanding_requests` metric
+        pub fn disable_default_outstanding_requests_metric(mut self) -> Self {
+            self.default_req_metrics_config.disable_outstanding_requests = true;
+            self
+        }
+
+        /// Disable the default `request_id` metric
         pub fn disable_default_request_id_metric(mut self) -> Self {
             self.default_req_metrics_config.disable_request_id = true;
             self
         }
 
-        /// Disable the `operation_name` metric
+        /// Disable the default `operation_name` metric
         pub fn disable_default_operation_metric(mut self) -> Self {
             self.default_req_metrics_config.disable_operation = true;
             self
         }
 
-        /// Disable the `service_name` metric
+        /// Disable the default `service_name` metric
         pub fn disable_default_service_metric(mut self) -> Self {
             self.default_req_metrics_config.disable_service = true;
             self
         }
 
-        /// Disable the `service_version` metric
+        /// Disable the default `service_version` metric
         pub fn disable_default_service_version_metric(mut self) -> Self {
             self.default_req_metrics_config.disable_service_version = true;
             self
         }
 
-        /// Disable the `http_status_code` metric
+        /// Disable the default `http_status_code` metric
         pub fn disable_default_http_status_code_metric(mut self) -> Self {
             self.default_res_metrics_config.disable_http_status_code = true;
             self
         }
 
-        /// Disable the `error` metric
-        pub fn disable_error_metric(mut self) -> Self {
-            self.default_res_metrics_config.disable_error = true;
+        /// Disable the default `success` metric
+        pub fn disable_default_success_metric(mut self) -> Self {
+            self.default_res_metrics_config.disable_success = true;
+            self
+        }
+
+        /// Disable the default `client_error` metric
+        pub fn disable_default_client_error_metric(mut self) -> Self {
+            self.default_res_metrics_config.disable_client_error = true;
             self
         }
 
-        /// Disable the `fault` metric
-        pub fn disable_fault_metric(mut self) -> Self {
-            self.default_res_metrics_config.disable_fault = true;
+        /// Disable the default `server_error` metric
+        pub fn disable_default_server_error_metric(mut self) -> Self {
+            self.default_res_metrics_config.disable_server_error = true;
             self
         }
 
-        /// Disable the `operation_time` metric
-        pub fn disable_operation_time_metric(mut self) -> Self {
+        /// Disable the default `operation_time` metric
+        pub fn disable_default_operation_time_metric(mut self) -> Self {
             self.default_res_metrics_config.disable_operation_time = true;
             self
         }

rust-runtime/aws-smithy-http-server-metrics/src/lib.rs

@@ -173,8 +173,9 @@
 //! | Metric | Description |
 //! |--------|-------------|
 //! | `http_status_code` | HTTP status code of the response |
-//! | `error` | Client error indicator (4xx status code) |
-//! | `fault` | Server fault indicator (5xx status code) |
+//! | `success` | Success indicator (2xx status code) |
+//! | `client_error` | Client error indicator (4xx status code) |
+//! | `server_error` | Server error indicator (5xx status code) |
 //! | `operation_time` | Timestamp that denotes operation time from pre-deserialization to post-serialization |
 //!
 //! # Platform support

rust-runtime/aws-smithy-http-server-metrics/src/plugin.rs

@@ -307,20 +307,30 @@ fn get_default_response_metrics(
     operation_time: Option<Duration>,
 ) -> DefaultResponseMetrics {
     let status = res.status();
-    let status_code = status.as_u16();
 
-    let error = if (400..500).contains(&status_code) {
-        Some(1)
+    let success = if status.is_success() {
+        Some(true)
     } else {
-        Some(0)
+        Some(false)
     };
 
-    let fault = if status_code >= 500 { Some(1) } else { Some(0) };
+    let client_error = if status.is_client_error() {
+        Some(true)
+    } else {
+        Some(false)
+    };
+
+    let server_error = if status.is_server_error() {
+        Some(true)
+    } else {
+        Some(false)
+    };
 
     DefaultResponseMetrics {
-        http_status_code: Some(status_code),
-        error,
-        fault,
+        http_status_code: Some(status.as_str().into()),
+        success,
+        client_error,
+        server_error,
         operation_time,
     }
 }
@@ -361,8 +371,13 @@ fn configure_default_response_metrics(
         http_status_code: metrics
             .http_status_code
             .filter(|_| !config.disable_http_status_code),
-        error: metrics.error.filter(|_| !config.disable_error),
-        fault: metrics.fault.filter(|_| !config.disable_fault),
+        success: metrics.success.filter(|_| !config.disable_success),
+        client_error: metrics
+            .client_error
+            .filter(|_| !config.disable_client_error),
+        server_error: metrics
+            .server_error
+            .filter(|_| !config.disable_server_error),
         operation_time: metrics
             .operation_time
             .filter(|_| !config.disable_operation_time),