From c54f4214dfa1055e0315041beb39d4baebc39100 Mon Sep 17 00:00:00 2001
From: shay7sev <shay7sev@gmail.com>
Date: Wed, 22 Jul 2026 14:40:07 +0800
Subject: [PATCH] feat(observability): suppress successful health logs

---
 docs/observability-runbook.md | 15 +++++++
 server/src/main.rs            | 79 +++++++++++++++++++++++++++++++++--
 2 files changed, 91 insertions(+), 3 deletions(-)

diff --git a/docs/observability-runbook.md b/docs/observability-runbook.md
index 3c6c45c..a0127ea 100644
--- a/docs/observability-runbook.md
+++ b/docs/observability-runbook.md
@@ -63,6 +63,15 @@ curl -fsS http://127.0.0.1:39180/health
 
 仪表盘适合浏览和筛选；需要精确关联请求或比较版本时，打开 `Explore`，数据源选择 `Loki`。
 
+仪表盘包含两个日志面板：
+
+- `HTTP Errors`：只显示能够解析出 HTTP `status >= 400` 的结构化日志。
+- `Service Logs (health excluded)`：显示所选服务的普通日志，并默认排除例行 `/health` 请求。
+
+API 不记录成功 `/health` 请求的开始和结束事件，以避免 Docker 日志与 Loki 被固定频率的探针日志占满。
+如果 `/health` 返回非 2xx 或请求在中间件中失败，错误日志仍会保留。需要查看改造前的历史健康检查日志时，
+在 `Explore` 中使用明确包含 `/health` 的查询，并把时间范围设置到对应部署时间之前。
+
 ## 常用 LogQL
 
 查看教学反馈的全部日志：
@@ -71,6 +80,12 @@ curl -fsS http://127.0.0.1:39180/health
 {application="teaching-feedback"}
 ```
 
+排除改造前已采集的健康检查日志：
+
+```logql
+{application="teaching-feedback", service="api"} != "\"path\":\"/health\""
+```
+
 按服务、级别和文本筛选：
 
 ```logql
diff --git a/server/src/main.rs b/server/src/main.rs
index 7b98580..109a022 100644
--- a/server/src/main.rs
+++ b/server/src/main.rs
@@ -12,7 +12,7 @@ use std::{path::PathBuf, sync::Arc, time::Duration};
 use axum::{
     Json, Router,
     extract::{DefaultBodyLimit, MatchedPath},
-    http::{HeaderName, HeaderValue, Request},
+    http::{HeaderName, HeaderValue, Request, StatusCode},
     response::Response,
     routing::get,
 };
@@ -55,6 +55,29 @@ impl From<&Config> for LogContext {
     }
 }
 
+fn should_log_http_request_started(path: &str) -> bool {
+    path != "/health"
+}
+
+fn should_log_http_response(is_health_check: bool, status: StatusCode) -> bool {
+    !is_health_check || !status.is_success()
+}
+
+#[derive(Clone, Copy, Debug)]
+struct HealthCheckResponse;
+
+async fn mark_health_check_response(
+    request: axum::extract::Request,
+    next: axum::middleware::Next,
+) -> Response {
+    let is_health_check = request.uri().path() == "/health";
+    let mut response = next.run(request).await;
+    if is_health_check {
+        response.extensions_mut().insert(HealthCheckResponse);
+    }
+    response
+}
+
 #[tokio::main]
 async fn main() -> Result<(), Box<dyn std::error::Error>> {
     dotenvy::dotenv().ok();
@@ -161,6 +184,9 @@ fn build_app(
                         make_http_request_span(request, &request_span_context)
                     })
                     .on_request(move |request: &Request<_>, span: &Span| {
+                        if !should_log_http_request_started(request.uri().path()) {
+                            return;
+                        }
                         tracing::info!(
                             parent: span,
                             service = "api",
@@ -174,6 +200,11 @@ fn build_app(
                     })
                     .on_response(move |response: &Response, latency: Duration, span: &Span| {
                         let status = response.status();
+                        let is_health_check =
+                            response.extensions().get::<HealthCheckResponse>().is_some();
+                        if !should_log_http_response(is_health_check, status) {
+                            return;
+                        }
                         let latency_ms = u64::try_from(latency.as_millis()).unwrap_or(u64::MAX);
                         let request_id = response
                             .headers()
@@ -226,6 +257,7 @@ fn build_app(
                         },
                     ),
             )
+            .layer(axum::middleware::from_fn(mark_health_check_response))
             .layer(PropagateRequestIdLayer::x_request_id()),
     ))
 }
@@ -304,11 +336,17 @@ async fn shutdown_signal(log_context: LogContext) {
 
 #[cfg(test)]
 mod tests {
-    use axum::{body::Body, http::Request};
+    use axum::{
+        body::Body,
+        http::{Request, StatusCode},
+    };
     use tower::ServiceExt;
     use uuid::Uuid;
 
-    use super::{AppState, LogContext, build_app};
+    use super::{
+        AppState, HealthCheckResponse, LogContext, build_app, should_log_http_request_started,
+        should_log_http_response,
+    };
     use crate::{
         auth::AuthService, config::AuthConfig, speech::SpeechService, summary::SummaryService,
     };
@@ -336,6 +374,41 @@ mod tests {
         .unwrap()
     }
 
+    #[test]
+    fn successful_health_request_is_not_logged() {
+        assert!(!should_log_http_request_started("/health"));
+        assert!(!should_log_http_response(true, StatusCode::OK));
+    }
+
+    #[test]
+    fn failed_health_request_is_logged() {
+        assert!(should_log_http_response(
+            true,
+            StatusCode::INTERNAL_SERVER_ERROR
+        ));
+    }
+
+    #[test]
+    fn ordinary_successful_request_is_logged() {
+        assert!(should_log_http_request_started("/api/v1/profiles"));
+        assert!(should_log_http_response(false, StatusCode::OK));
+    }
+
+    #[tokio::test]
+    async fn health_response_is_marked_for_log_filtering() {
+        let response = app()
+            .oneshot(
+                Request::builder()
+                    .uri("/health")
+                    .body(Body::empty())
+                    .unwrap(),
+            )
+            .await
+            .unwrap();
+
+        assert!(response.extensions().get::<HealthCheckResponse>().is_some());
+    }
+
     #[tokio::test]
     async fn health_generates_a_uuid_request_id() {
         let response = app()