Add API to get single cluster status

Author: sbernauer

CHANGELOG.md

@@ -6,8 +6,8 @@ All notable changes to this project will be documented in this file.
 
 ### Added
 
-- Support activating and deactivation Trino clusters via API calls to `/admin/activate-cluster/{cluster_name}` and `/admin/deactivate-cluster/{cluster_name}` respectively. For this to work you need to authenticate yourself at trino-lb via basic auth ([#95]).
-- Expose cluster statistics at `/admin/cluster-status` ([#95]).
+- Support activating and deactivation Trino clusters via API calls to `/admin/clusters/{cluster_name}/activate` and `/admin/clusters/{cluster_name}/deactivate` respectively. For this to work you need to authenticate yourself at trino-lb via basic auth ([#95]).
+- Expose cluster statistics at `/admin/clusters/{cluster_name}/status` and `/admin/clusters/status` ([#95]).
 
 ### Changed
 

docs/design.md

@@ -88,10 +88,10 @@ A deactivated cluster continues to process already running queries, but no new q
 
 To safely update a running Trino cluster you can
 
-1. Deactivate the Trino cluster using `POST /admin/deactivate-cluster/{cluster_name}`
-2. Wait until all running queries have finished. Tip: You can use `GET /admin/cluster-status` to fetch the current query counter or talk to the Trino cluster directly.
+1. Deactivate the Trino cluster using `POST /admin/clusters/{cluster_name}/deactivate`
+2. Wait until all running queries have finished. Tip: You can use `GET /admin/clusters/{cluster_name}/status` (or `GET /admin/clusters/status`) to fetch the current query counter or talk to the Trino cluster directly.
 3. Safely do modifications to the Trino cluster without any user impact
-4. Once you are done with your changes re-activate the Trino cluster again using `POST /admin/activate-cluster/{cluster_name}`
+4. Once you are done with your changes re-activate the Trino cluster again using `POST /admin/clusters/{cluster_name}/activate`
 
 Important: To activate/deactivate a Trino cluster you need to authenticate against trino-lb.
 Currently only Basic Auth is supported, you can configure the credentials like this:

trino-lb/src/http_server/admin/mod.rs

@@ -35,12 +35,24 @@ pub enum Error {
     #[snafu(display("Unknown Trino cluster {cluster:?}"))]
     UnknownCluster { cluster: TrinoClusterName },
 
+    #[snafu(display("Failed to get cluster state for cluster {cluster:?} from persistence"))]
+    GetClusterStateFromPersistence {
+        source: trino_lb_persistence::Error,
+        cluster: TrinoClusterName,
+    },
+
     #[snafu(display("Failed to set cluster state for cluster {cluster:?} in persistence"))]
     SetClusterStateInPersistence {
         source: trino_lb_persistence::Error,
         cluster: TrinoClusterName,
     },
 
+    #[snafu(display("Failed to get the query counter for cluster {cluster:?} from persistence"))]
+    GetQueryCounterForGroup {
+        source: trino_lb_persistence::Error,
+        cluster: TrinoClusterName,
+    },
+
     #[snafu(display("Failed to get all cluster states"))]
     GetAllClusterStates {
         source: cluster_group_manager::Error,
@@ -54,7 +66,9 @@ impl IntoResponse for Error {
             Error::NoAdminAuthenticationMethodDefined => StatusCode::UNAUTHORIZED,
             Error::InvalidAdminCredentials => StatusCode::UNAUTHORIZED,
             Error::UnknownCluster { .. } => StatusCode::NOT_FOUND,
+            Error::GetClusterStateFromPersistence { .. } => StatusCode::INTERNAL_SERVER_ERROR,
             Error::SetClusterStateInPersistence { .. } => StatusCode::INTERNAL_SERVER_ERROR,
+            Error::GetQueryCounterForGroup { .. } => StatusCode::INTERNAL_SERVER_ERROR,
             Error::GetAllClusterStates { .. } => StatusCode::INTERNAL_SERVER_ERROR,
         };
         (status_code, format!("{self}")).into_response()
@@ -64,7 +78,7 @@ impl IntoResponse for Error {
 /// (Re)-Activates a Trino Cluster, so that it receives new queries.
 ///
 /// This is useful for maintenance actions (in combination with deactivation).
-#[instrument(name = "POST /admin/activate-cluster/{cluster_name}", skip(state))]
+#[instrument(name = "POST /admin/clusters/{cluster_name}/activate", skip(state))]
 pub async fn post_activate_cluster(
     TypedHeader(Authorization(basic_auth)): TypedHeader<Authorization<Basic>>,
     State(state): State<Arc<AppState>>,
@@ -81,7 +95,7 @@ pub async fn post_activate_cluster(
 /// Deactivate a Trino Cluster, so that it doesn't receive any new queries.
 ///
 /// This is useful for maintenance actions (in combination with activation).
-#[instrument(name = "POST /admin/deactivate-cluster/{cluster_name}", skip(state))]
+#[instrument(name = "POST /admin/clusters/{cluster_name}/deactivate", skip(state))]
 pub async fn post_deactivate_cluster(
     TypedHeader(Authorization(basic_auth)): TypedHeader<Authorization<Basic>>,
     State(state): State<Arc<AppState>>,
@@ -95,16 +109,55 @@ pub async fn post_deactivate_cluster(
     set_cluster_activation(state, basic_auth, &cluster_name, false).await
 }
 
-/// Get the status of the Trino clusters
-#[instrument(name = "GET /admin/cluster-status", skip(state))]
+/// Get the status of a single Trino clusters
+#[instrument(name = "/admin/clusters/{cluster_name}/status", skip(state))]
 pub async fn get_cluster_status(
     State(state): State<Arc<AppState>>,
-) -> Result<Json<BTreeMap<TrinoClusterName, ClusterStats>>, Error> {
+    Path(cluster_name): Path<TrinoClusterName>,
+) -> Result<Json<ClusterStats>, Error> {
     state
         .metrics
         .http_counter
         .add(1, &[KeyValue::new("resource", "get_cluster_status")]);
 
+    ensure!(
+        state.config.is_cluster_in_config(&cluster_name),
+        UnknownClusterSnafu {
+            cluster: cluster_name
+        }
+    );
+
+    let cluster_state = state
+        .persistence
+        .get_cluster_state(&cluster_name)
+        .await
+        .context(GetClusterStateFromPersistenceSnafu {
+            cluster: &cluster_name,
+        })?;
+    let cluster_query_counter = state
+        .persistence
+        .get_cluster_query_count(&cluster_name)
+        .await
+        .context(GetQueryCounterForGroupSnafu {
+            cluster: &cluster_name,
+        })?;
+
+    Ok(Json(ClusterStats {
+        state: cluster_state,
+        query_counter: cluster_query_counter,
+    }))
+}
+
+/// Get the status of all Trino clusters
+#[instrument(name = "/admin/clusters/status", skip(state))]
+pub async fn get_all_cluster_status(
+    State(state): State<Arc<AppState>>,
+) -> Result<Json<BTreeMap<TrinoClusterName, ClusterStats>>, Error> {
+    state
+        .metrics
+        .http_counter
+        .add(1, &[KeyValue::new("resource", "get_all_cluster_status")]);
+
     let cluster_stats = state
         .cluster_group_manager
         .get_all_cluster_stats()

trino-lb/src/http_server/mod.rs

@@ -123,14 +123,18 @@ pub async fn start_http_server(
             delete(v1::statement::delete_trino_executing_statement),
         )
         .route(
-            "/admin/activate-cluster/{cluster_name}",
+            "/admin/clusters/{cluster_name}/activate",
             post(admin::post_activate_cluster),
         )
         .route(
-            "/admin/deactivate-cluster/{cluster_name}",
+            "/admin/clusters/{cluster_name}/deactivate",
             post(admin::post_deactivate_cluster),
         )
-        .route("/admin/cluster-status", get(admin::get_cluster_status))
+        .route(
+            "/admin/clusters/{cluster_name}/status",
+            get(admin::get_cluster_status),
+        )
+        .route("/admin/clusters/status", get(admin::get_all_cluster_status))
         .route("/ui/index.html", get(ui::index::get_ui_index))
         .route("/ui/query.html", get(ui::query::get_ui_query))
         .layer(TraceLayer::new_for_http())