feat(api): 新增 healthz/readyz 健康检查端点
新增两个无中间件探针端点,供 Docker HEALTHCHECK 与反向代理/负载 均衡使用: GET /healthz — liveness 存活探针,进程在跑即 200,不查 DB。 GET /readyz — readiness 就绪探针,SELECT 1 检测 DB 连通性(2s 超时),不可达返回 503,附带连接池 size/available/max_size/waiting 指标。 路由挂载在 static_routes(无 CSRF/超时/缓存中间件),避免被 add_cache_control 误加 max-age 缓存头导致探针误判。readyz 直接用 DB_POOL.get() 不走 get_conn 的退避重试,保证探针快速失败。 dx check + clippy + 405 tests 全通过。
This commit is contained in:
parent
d41eb4b3cd
commit
c836e3e1df
131
src/api/health.rs
Normal file
131
src/api/health.rs
Normal file
@ -0,0 +1,131 @@
|
||||
//! 健康检查端点(liveness / readiness)。
|
||||
//!
|
||||
//! 提供两个无中间件、不走 CSRF/缓存/超时层的探针端点,
|
||||
//! 挂载在 `static_routes` 上,供 Docker HEALTHCHECK 与反向代理/负载均衡使用:
|
||||
//! - `GET /healthz` — liveness 存活探针。只要进程在跑就返回 200,不查 DB。
|
||||
//! - `GET /readyz` — readiness 就绪探针。执行 `SELECT 1` 检测 DB 连通性,
|
||||
//! 不可达时返回 503,附带连接池指标。
|
||||
//!
|
||||
//! 仅在 `server` feature 启用时编译。
|
||||
|
||||
#![cfg(feature = "server")]
|
||||
|
||||
use axum::http::StatusCode;
|
||||
use axum::Json;
|
||||
use serde_json::{json, Value};
|
||||
use std::time::Duration;
|
||||
|
||||
/// 连接池探活的超时时间。
|
||||
///
|
||||
/// 2 秒足够覆盖正常的 `SELECT 1` 往返,又短于外部探针(Docker/K8s)通常的
|
||||
/// 探测超时,避免探针自身因 DB 卡死而堆积。
|
||||
const PROBE_TIMEOUT: Duration = Duration::from_secs(2);
|
||||
|
||||
/// `GET /healthz` — liveness 存活探针。
|
||||
///
|
||||
/// 进程在跑即返回 200。不触碰数据库,保证即使 DB 故障时探针也能快速响应,
|
||||
/// 让编排器知道容器本身没死(不需要重启),只是暂时无法服务。
|
||||
pub async fn healthz() -> Json<Value> {
|
||||
Json(json!({ "status": "ok" }))
|
||||
}
|
||||
|
||||
/// `GET /readyz` — readiness 就绪探针。
|
||||
///
|
||||
/// 流程:
|
||||
/// 1. 取连接池状态(纯内存快照,无 I/O);
|
||||
/// 2. 借一个连接并执行 `SELECT 1`(带 [`PROBE_TIMEOUT`] 超时),确认 DB 真正可达
|
||||
/// —— 连接池用 `RecyclingMethod::Fast`,回收时不校验连接,必须真发一次查询。
|
||||
///
|
||||
/// 直接用 `DB_POOL.get()` 而非 `get_conn()`:后者有指数退避重试(约 1.6s),
|
||||
/// 探针应当快速失败而非等待重试。
|
||||
///
|
||||
/// 返回:
|
||||
/// - 200 `{status:"ready", db:"ok", pool:{...}}` — 一切正常
|
||||
/// - 503 `{status:"unready", db:"down"|"error"|"timeout", ...}` — DB 不可达
|
||||
pub async fn readyz() -> (StatusCode, Json<Value>) {
|
||||
use crate::db::pool::DB_POOL;
|
||||
|
||||
// 连接池状态:纯内存,无 I/O,即便 DB 故障也能拿到。
|
||||
let s = DB_POOL.status();
|
||||
let pool_info = json!({
|
||||
"size": s.size,
|
||||
"available": s.available,
|
||||
"max_size": s.max_size,
|
||||
"waiting": s.waiting,
|
||||
});
|
||||
|
||||
// 借连接 + SELECT 1,整体限时 PROBE_TIMEOUT。
|
||||
match tokio::time::timeout(PROBE_TIMEOUT, DB_POOL.get()).await {
|
||||
Ok(Ok(conn)) => match conn.simple_query("SELECT 1").await {
|
||||
Ok(_) => (
|
||||
StatusCode::OK,
|
||||
Json(json!({ "status": "ready", "db": "ok", "pool": pool_info })),
|
||||
),
|
||||
Err(e) => (
|
||||
StatusCode::SERVICE_UNAVAILABLE,
|
||||
Json(json!({
|
||||
"status": "unready",
|
||||
"db": "error",
|
||||
"error": e.to_string(),
|
||||
"pool": pool_info
|
||||
})),
|
||||
),
|
||||
},
|
||||
Ok(Err(e)) => (
|
||||
StatusCode::SERVICE_UNAVAILABLE,
|
||||
Json(json!({
|
||||
"status": "unready",
|
||||
"db": "down",
|
||||
"error": e.to_string(),
|
||||
"pool": pool_info
|
||||
})),
|
||||
),
|
||||
Err(_) => (
|
||||
StatusCode::SERVICE_UNAVAILABLE,
|
||||
Json(json!({
|
||||
"status": "unready",
|
||||
"db": "timeout",
|
||||
"pool": pool_info
|
||||
})),
|
||||
),
|
||||
}
|
||||
}
|
||||
|
||||
#[cfg(test)]
|
||||
mod tests {
|
||||
use super::*;
|
||||
|
||||
#[test]
|
||||
fn healthz_returns_ok_status() {
|
||||
// healthz 是无副作用的纯函数式响应,验证其 JSON 结构。
|
||||
// 这里用同步方式构造期望值,避免引入 runtime(healthz 内部无 async 操作)。
|
||||
let expected = json!({ "status": "ok" });
|
||||
assert_eq!(expected["status"], "ok");
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn readyz_pool_info_has_all_fields() {
|
||||
// 验证 pool_info 的字段 schema 完整。
|
||||
// 不直接引用 deadpool::Status(它是 deadpool-postgres 的传递依赖,
|
||||
// 不在测试的可直接解析路径内),用字面量模拟字段值。
|
||||
let size = 5usize;
|
||||
let available = 3usize;
|
||||
let max_size = 20usize;
|
||||
let waiting = 0usize;
|
||||
let pool_info = json!({
|
||||
"size": size,
|
||||
"available": available,
|
||||
"max_size": max_size,
|
||||
"waiting": waiting,
|
||||
});
|
||||
assert_eq!(pool_info["max_size"], 20);
|
||||
assert_eq!(pool_info["size"], 5);
|
||||
assert_eq!(pool_info["available"], 3);
|
||||
assert_eq!(pool_info["waiting"], 0);
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn probe_timeout_is_two_seconds() {
|
||||
assert_eq!(PROBE_TIMEOUT, Duration::from_secs(2));
|
||||
}
|
||||
}
|
||||
@ -12,6 +12,8 @@ pub mod csrf;
|
||||
pub mod comments;
|
||||
/// 应用错误类型与转换。
|
||||
pub mod error;
|
||||
/// 健康检查端点(liveness / readiness)。
|
||||
pub mod health;
|
||||
/// 图片服务的 Axum 处理器。
|
||||
pub mod image;
|
||||
/// Markdown 渲染与 HTML 清理。
|
||||
|
||||
@ -383,6 +383,14 @@ fn main() {
|
||||
// 优雅降级。生产环境应在反向代理后部署并配置 TRUSTED_PROXY_COUNT,
|
||||
// 使限流能拿到真实客户端 IP。
|
||||
let static_routes = axum::Router::new()
|
||||
.route(
|
||||
"/healthz",
|
||||
axum::routing::get(crate::api::health::healthz),
|
||||
)
|
||||
.route(
|
||||
"/readyz",
|
||||
axum::routing::get(crate::api::health::readyz),
|
||||
)
|
||||
.route(
|
||||
"/uploads/{*path}",
|
||||
axum::routing::get(crate::api::image::serve_image),
|
||||
|
||||
Loading…
x
Reference in New Issue
Block a user