Add health endpoint (#808)

We currently use `/ctrl/status` as a health check endpoint. The
introduction of auth delegation (#804) applies, when enabled, to all
`/ctrl` endpoints, including the health check. That means that the
health checker needs to have credentials.

Instead, this adds a separate health check endpoint `/pub/health` which
is part of the `/pub` path, which is not credentialed. This does not
include the version information returned by the `/ctrl/status` endpoint,
but does a health check.

plane/plane-tests/tests/docker_command.rs

@@ -70,7 +70,7 @@ async fn test_resource_limits() {
     let docker = bollard::Docker::connect_with_local_defaults().unwrap();
     let backend_name = BackendName::new_random();
     let _container_id = ContainerId::from(format!("plane-test-{}", backend_name));
-    let mut executor_config = DockerExecutorConfig::from_image_with_defaults("debian:bookworm");
+    let mut executor_config = DockerExecutorConfig::from_image_with_defaults("alpine:latest");
 
     let resource_limits: ResourceLimits = serde_json::from_value(serde_json::json!( {
         "cpu_period": 1000000,

plane/plane-tests/tests/health.rs

@@ -0,0 +1,13 @@
+use common::test_env::TestEnvironment;
+use plane_test_macro::plane_test;
+
+mod common;
+
+#[plane_test]
+async fn controller_health_check(env: TestEnvironment) {
+    let controller = env.controller().await;
+    let client = controller.client();
+    let result = client.health_check().await;
+
+    assert!(result.is_ok());
+}

plane/src/client/mod.rs

@@ -194,6 +194,12 @@ impl PlaneClient {
         let cluster_state: ClusterState = authed_get(&self.client, &url).await?;
         Ok(cluster_state)
     }
+
+    pub async fn health_check(&self) -> Result<(), PlaneClientError> {
+        let url = self.controller_address.join("/pub/health");
+        self.client.get(url.url).send().await?;
+        Ok(())
+    }
 }
 
 async fn get_response<T: DeserializeOwned>(response: Response) -> Result<T, PlaneClientError> {

plane/src/controller/mod.rs

@@ -31,6 +31,7 @@ use axum::{
 use forward_auth::forward_layer;
 use futures_util::never::Never;
 use serde::{Deserialize, Serialize};
+use serde_json::{json, Value};
 use std::net::{SocketAddr, TcpListener};
 use tokio::{
     sync::oneshot::{self},
@@ -80,6 +81,18 @@ pub async fn status(
     }))
 }
 
+pub async fn health(State(controller): State<Controller>) -> Result<Json<Value>, Response> {
+    controller
+        .db
+        .health_check()
+        .await
+        .or_internal_error("Database health check failed")?;
+
+    Ok(Json(json!({
+        "status": "ok"
+    })))
+}
+
 struct HeartbeatSender {
     handle: JoinHandle<Never>,
     db: PlaneDatabase,
@@ -243,6 +256,7 @@ impl ControllerServer {
                 "/b/:backend/status-stream",
                 get(handle_backend_status_stream),
             )
+            .route("/health", get(health))
             .layer(cors_public.clone());
 
         let app = Router::new()