[peer-metrics] refactor SystemMetrics to hold a reference to System

1 files changed, 66 insertions, 60 deletions

diff --git a/src/rpc/system_metrics.rs b/src/rpc/system_metrics.rs
index ee07672e..ad4aca2f 100644
--- a/src/rpc/system_metrics.rs
+++ b/src/rpc/system_metrics.rs
@@ -1,13 +1,12 @@
 use std::sync::{Arc, RwLock};
+use std::time::{Duration, Instant};
 
 use opentelemetry::{global, metrics::*, KeyValue};
 
-use crate::system::{ClusterHealth, ClusterHealthStatus, NodeStatus};
+use crate::system::{ClusterHealthStatus, System};
 
 /// TableMetrics reference all counter used for metrics
 pub struct SystemMetrics {
-	pub(crate) health: Arc<RwLock<Option<ClusterHealth>>>,
-
 	// Static values
 	pub(crate) _garage_build_info: ValueObserver<u64>,
 	pub(crate) _replication_factor: ValueObserver<u64>,
@@ -29,12 +28,25 @@ pub struct SystemMetrics {
 }
 
 impl SystemMetrics {
-	pub fn new(replication_factor: usize, local_status: Arc<RwLock<NodeStatus>>) -> Self {
+	pub fn new(system: Arc<System>) -> Self {
 		let meter = global::meter("garage_system");
-		let health = Arc::new(RwLock::new(None));
-		Self {
-			health: health.clone(),
 
+		let health_cache = RwLock::new((Instant::now(), system.health()));
+		let system2 = system.clone();
+		let get_health = Arc::new(move || {
+			{
+				let cache = health_cache.read().unwrap();
+				if cache.0 > Instant::now() - Duration::from_secs(1) {
+					return cache.1;
+				}
+			}
+
+			let health = system2.health();
+			*health_cache.write().unwrap() = (Instant::now(), health);
+			health
+		});
+
+		Self {
 			// Static values
 			_garage_build_info: meter
 				.u64_value_observer("garage_build_info", move |observer| {
@@ -48,19 +60,22 @@ impl SystemMetrics {
 				})
 				.with_description("Garage build info")
 				.init(),
-			_replication_factor: meter
-				.u64_value_observer("garage_replication_factor", move |observer| {
-					observer.observe(replication_factor as u64, &[])
-				})
-				.with_description("Garage replication factor setting")
-				.init(),
+			_replication_factor: {
+				let replication_factor = system.replication_factor;
+				meter
+					.u64_value_observer("garage_replication_factor", move |observer| {
+						observer.observe(replication_factor as u64, &[])
+					})
+					.with_description("Garage replication factor setting")
+					.init()
+			},
 
 			// Disk space values from System::local_status
 			_disk_avail: {
-				let status = local_status.clone();
+				let system = system.clone();
 				meter
 					.u64_value_observer("garage_local_disk_avail", move |observer| {
-						let st = status.read().unwrap();
+						let st = system.local_status.read().unwrap();
 						if let Some((avail, _total)) = st.data_disk_avail {
 							observer.observe(avail, &[KeyValue::new("volume", "data")]);
 						}
@@ -72,10 +87,10 @@ impl SystemMetrics {
 					.init()
 			},
 			_disk_total: {
-				let status = local_status.clone();
+				let system = system.clone();
 				meter
 					.u64_value_observer("garage_local_disk_total", move |observer| {
-						let st = status.read().unwrap();
+						let st = system.local_status.read().unwrap();
 						if let Some((_avail, total)) = st.data_disk_avail {
 							observer.observe(total, &[KeyValue::new("volume", "data")]);
 						}
@@ -87,98 +102,90 @@ impl SystemMetrics {
 					.init()
 			},
 
-			// Health report from System::health()
+			// Health report from System::()
 			_cluster_healthy: {
-				let health = health.clone();
+				let get_health = get_health.clone();
 				meter
 					.u64_value_observer("cluster_healthy", move |observer| {
-						if let Some(h) = health.read().unwrap().as_ref() {
-							if h.status == ClusterHealthStatus::Healthy {
-								observer.observe(1, &[]);
-							} else {
-								observer.observe(0, &[]);
-							}
+						let h = get_health();
+						if h.status == ClusterHealthStatus::Healthy {
+							observer.observe(1, &[]);
+						} else {
+							observer.observe(0, &[]);
 						}
 					})
 					.with_description("Whether all storage nodes are connected")
 					.init()
 			},
 			_cluster_available: {
-				let health = health.clone();
+				let get_health = get_health.clone();
 				meter.u64_value_observer("cluster_available", move |observer| {
-					if let Some(h) = health.read().unwrap().as_ref() {
-						if h.status != ClusterHealthStatus::Unavailable {
-							observer.observe(1, &[]);
-						} else {
-							observer.observe(0, &[]);
-						}
+					let h = get_health();
+					if h.status != ClusterHealthStatus::Unavailable {
+						observer.observe(1, &[]);
+					} else {
+						observer.observe(0, &[]);
 					}
 				})
 				.with_description("Whether all requests can be served, even if some storage nodes are disconnected")
 				.init()
 			},
 			_known_nodes: {
-				let health = health.clone();
+				let get_health = get_health.clone();
 				meter
 					.u64_value_observer("cluster_known_nodes", move |observer| {
-						if let Some(h) = health.read().unwrap().as_ref() {
-							observer.observe(h.known_nodes as u64, &[]);
-						}
+						let h = get_health();
+						observer.observe(h.known_nodes as u64, &[]);
 					})
 					.with_description("Number of nodes already seen once in the cluster")
 					.init()
 			},
 			_connected_nodes: {
-				let health = health.clone();
+				let get_health = get_health.clone();
 				meter
 					.u64_value_observer("cluster_connected_nodes", move |observer| {
-						if let Some(h) = health.read().unwrap().as_ref() {
-							observer.observe(h.connected_nodes as u64, &[]);
-						}
+						let h = get_health();
+						observer.observe(h.connected_nodes as u64, &[]);
 					})
 					.with_description("Number of nodes currently connected")
 					.init()
 			},
 			_storage_nodes: {
-				let health = health.clone();
+				let get_health = get_health.clone();
 				meter
 					.u64_value_observer("cluster_storage_nodes", move |observer| {
-						if let Some(h) = health.read().unwrap().as_ref() {
-							observer.observe(h.storage_nodes as u64, &[]);
-						}
+						let h = get_health();
+						observer.observe(h.storage_nodes as u64, &[]);
 					})
 					.with_description("Number of storage nodes declared in the current layout")
 					.init()
 			},
 			_storage_nodes_ok: {
-				let health = health.clone();
+				let get_health = get_health.clone();
 				meter
 					.u64_value_observer("cluster_storage_nodes_ok", move |observer| {
-						if let Some(h) = health.read().unwrap().as_ref() {
-							observer.observe(h.storage_nodes_ok as u64, &[]);
-						}
+						let h = get_health();
+						observer.observe(h.storage_nodes_ok as u64, &[]);
 					})
 					.with_description("Number of storage nodes currently connected")
 					.init()
 			},
 			_partitions: {
-				let health = health.clone();
+				let get_health = get_health.clone();
 				meter
 					.u64_value_observer("cluster_partitions", move |observer| {
-						if let Some(h) = health.read().unwrap().as_ref() {
-							observer.observe(h.partitions as u64, &[]);
-						}
+						let h = get_health();
+						observer.observe(h.partitions as u64, &[]);
 					})
 					.with_description("Number of partitions in the layout")
 					.init()
 			},
 			_partitions_quorum: {
-				let health = health.clone();
+				let get_health = get_health.clone();
 				meter
 					.u64_value_observer("cluster_partitions_quorum", move |observer| {
-						if let Some(h) = health.read().unwrap().as_ref() {
-							observer.observe(h.partitions_quorum as u64, &[]);
-						}
+						let h = get_health();
+						observer.observe(h.partitions_quorum as u64, &[]);
 					})
 					.with_description(
 						"Number of partitions for which we have a quorum of connected nodes",
@@ -186,12 +193,11 @@ impl SystemMetrics {
 					.init()
 			},
 			_partitions_all_ok: {
-				let health = health.clone();
+				let get_health = get_health.clone();
 				meter
 					.u64_value_observer("cluster_partitions_all_ok", move |observer| {
-						if let Some(h) = health.read().unwrap().as_ref() {
-							observer.observe(h.partitions_all_ok as u64, &[]);
-						}
+						let h = get_health();
+						observer.observe(h.partitions_all_ok as u64, &[]);
 					})
 					.with_description(
 						"Number of partitions for which all storage nodes are connected",