api: add clocks to /health

2 files changed, 20 insertions, 3 deletions

diff --git a/src/api/health.rs b/src/api/health.rs
index b1e4523..3bd374e 100644
--- a/src/api/health.rs
+++ b/src/api/health.rs
@@ -1,15 +1,32 @@
 //! `GET /health` and `GET /ready`: unauthenticated, for supervisors and load balancers.
-//! They say whether the node is up and whether it serves the store, nothing more.
+//! They say whether the node is up and whether it serves the store, and `/health` on a
+//! node whether its clock is off, nothing more.
+
+use std::time::Duration;
 
 use axum::{extract::State, http::StatusCode};
 
 use crate::state::AppState;
 
-/// The HTTP API is up.
+/// The HTTP API is up (a relay's).
 pub(crate) async fn health() -> &'static str {
     "ok\n"
 }
 
+/// The HTTP API is up (a node's); a second line while its clock is off from its peers'
+/// (`swarm::skew`), which doesn't make it less up.
+pub(crate) async fn node_health(State(state): State<AppState>) -> String {
+    let skew = crate::swarm::skew::measure(&state).await;
+    match skew.offset_ms {
+        Some(offset) if skew.off => format!(
+            "ok\nclock off: {} {} most of its peers'\n",
+            crate::units::format_duration(Duration::from_millis(offset.unsigned_abs())),
+            if offset < 0 { "behind" } else { "ahead of" }
+        ),
+        _ => "ok\n".to_owned(),
+    }
+}
+
 /// The node serves reads and writes: ready, not joining or rejoining.
 pub(crate) async fn ready(State(state): State<AppState>) -> (StatusCode, &'static str) {
     if state.is_ready().await {
diff --git a/src/api/mod.rs b/src/api/mod.rs
index fb860c7..528df35 100644
--- a/src/api/mod.rs
+++ b/src/api/mod.rs
@@ -77,7 +77,7 @@ pub(crate) fn router(state: AppState, auth_state: AuthState, debug_api: bool) ->
             SwaggerUi::new("/swagger-ui").config(utoipa_swagger_ui::Config::from("/openapi.yaml")),
         )
         .route("/openapi.yaml", get(openapi_spec))
-        .route("/health", get(health::health))
+        .route("/health", get(health::node_health))
         .route("/ready", get(health::ready))
         // Outermost, so authentication failures get the error body too.
         .layer(middleware::from_fn(json_errors))