Add node authority status metric (paritytech#4699)

* Check authority status on active leaves update

Signed-off-by: Andrei Sandu <andrei-mihail@parity.io>

* cargo changes

Signed-off-by: Andrei Sandu <andrei-mihail@parity.io>

* Fix tests

Signed-off-by: Andrei Sandu <andrei-mihail@parity.io>

* Add metric for authority status

Signed-off-by: Andrei Sandu <andrei-mihail@parity.io>

* Revert "Fix tests"

This reverts commit 5bd56bb.

* Revert "cargo changes"

This reverts commit ffea18f.

* Revert "Check authority status on active leaves update"

This reverts commit 55a30ac.

* Test fixups

Signed-off-by: Andrei Sandu <andrei-mihail@parity.io>

* fix

Signed-off-by: Andrei Sandu <andrei-mihail@parity.io>

* update

Signed-off-by: Andrei Sandu <andrei-mihail@parity.io>

* undo damage

Signed-off-by: Andrei Sandu <andrei-mihail@parity.io>

* dont update status on runtime errors

Signed-off-by: Andrei Sandu <andrei-mihail@parity.io>

* Fix tests

Signed-off-by: Andrei Sandu <andrei-mihail@parity.io>

* fix inconsistency

Signed-off-by: Andrei Sandu <andrei-mihail@parity.io>

* Review feedback

Signed-off-by: Andrei Sandu <andrei-mihail@parity.io>

* Dont derive primitive Default

Signed-off-by: Andrei Sandu <andrei-mihail@parity.io>

* add dummy_session_info helper

Signed-off-by: Andrei Sandu <andrei-mihail@parity.io>

* unset parachain validator status if no longer authority

Signed-off-by: Andrei Sandu <andrei-mihail@parity.io>

* update

Signed-off-by: Andrei Sandu <andrei-mihail@parity.io>

* damn

Signed-off-by: Andrei Sandu <andrei-mihail@parity.io>

* 🤦

Signed-off-by: Andrei Sandu <andrei-mihail@parity.io>

node/network/gossip-support/src/lib.rs

@@ -49,14 +49,18 @@ use polkadot_node_subsystem::{
 		RuntimeApiRequest,
 	},
 	overseer, ActiveLeavesUpdate, FromOverseer, OverseerSignal, SpawnedSubsystem, SubsystemContext,
-	SubsystemError,
+	SubsystemError, SubsystemSender,
 };
 use polkadot_node_subsystem_util as util;
 use polkadot_primitives::v1::{AuthorityDiscoveryId, Hash, SessionIndex};
 
 #[cfg(test)]
 mod tests;
 
+mod metrics;
+
+use metrics::Metrics;
+
 const LOG_TARGET: &str = "parachain::gossip-support";
 // How much time should we wait to reissue a connection request
 // since the last authority discovery resolution failure.
@@ -104,14 +108,21 @@ pub struct GossipSupport<AD> {
 	connected_authorities_by_peer_id: HashMap<PeerId, HashSet<AuthorityDiscoveryId>>,
 	/// Authority discovery service.
 	authority_discovery: AD,
+
+	/// Subsystem metrics.
+	metrics: Metrics,
 }
 
 impl<AD> GossipSupport<AD>
 where
 	AD: AuthorityDiscovery,
 {
 	/// Create a new instance of the [`GossipSupport`] subsystem.
-	pub fn new(keystore: SyncCryptoStorePtr, authority_discovery: AD) -> Self {
+	pub fn new(keystore: SyncCryptoStorePtr, authority_discovery: AD, metrics: Metrics) -> Self {
+		// Initialize metrics to `0`.
+		metrics.on_is_not_authority();
+		metrics.on_is_not_parachain_validator();
+
 		Self {
 			keystore,
 			last_session_index: None,
@@ -121,6 +132,7 @@ where
 			connected_authorities: HashMap::new(),
 			connected_authorities_by_peer_id: HashMap::new(),
 			authority_discovery,
+			metrics,
 		}
 	}
 
@@ -223,10 +235,60 @@ where
 
 				if is_new_session {
 					update_gossip_topology(ctx, our_index, all_authorities, relay_parent).await?;
+					self.update_authority_status_metrics(leaf, ctx.sender()).await?;
 				}
 			}
 		}
+		Ok(())
+	}
 
+	async fn update_authority_status_metrics(
+		&mut self,
+		leaf: Hash,
+		sender: &mut impl SubsystemSender,
+	) -> Result<(), util::Error> {
+		if let Some(session_info) = util::request_session_info(
+			leaf,
+			self.last_session_index
+				.expect("Last session index is always set on every session index change"),
+			sender,
+		)
+		.await
+		.await??
+		{
+			let maybe_index = match ensure_i_am_an_authority(
+				&self.keystore,
+				&session_info.discovery_keys,
+			)
+			.await
+			{
+				Ok(index) => {
+					self.metrics.on_is_authority();
+					Some(index)
+				},
+				Err(util::Error::NotAValidator) => {
+					self.metrics.on_is_not_authority();
+					self.metrics.on_is_not_parachain_validator();
+					None
+				},
+				// Don't update on runtime errors.
+				Err(_) => None,
+			};
+
+			if let Some(validator_index) = maybe_index {
+				// The subset of authorities participating in parachain consensus.
+				let parachain_validators_this_session = session_info.validators;
+
+				// First `maxValidators` entries are the parachain validators. We'll check
+				// if our index is in this set to avoid searching for the keys.
+				// https://github.com/paritytech/polkadot/blob/a52dca2be7840b23c19c153cf7e110b1e3e475f8/runtime/parachains/src/configuration.rs#L148
+				if validator_index < parachain_validators_this_session.len() {
+					self.metrics.on_is_parachain_validator();
+				} else {
+					self.metrics.on_is_not_parachain_validator();
+				}
+			}
+		}
 		Ok(())
 	}
 

node/network/gossip-support/src/metrics.rs

@@ -0,0 +1,90 @@
+// Copyright 2021 Parity Technologies (UK) Ltd.
+// This file is part of Polkadot.
+
+// Polkadot is free software: you can redistribute it and/or modify
+// it under the terms of the GNU General Public License as published by
+// the Free Software Foundation, either version 3 of the License, or
+// (at your option) any later version.
+
+// Polkadot is distributed in the hope that it will be useful,
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+// GNU General Public License for more details.
+
+// You should have received a copy of the GNU General Public License
+// along with Polkadot.  If not, see <http://www.gnu.org/licenses/>.
+
+use polkadot_node_subsystem_util::{
+	metrics,
+	metrics::{
+		prometheus,
+		prometheus::{Gauge, PrometheusError, Registry, U64},
+	},
+};
+
+/// Dispute Distribution metrics.
+#[derive(Clone, Default)]
+pub struct Metrics(Option<MetricsInner>);
+
+#[derive(Clone)]
+struct MetricsInner {
+	/// Tracks authority status for producing relay chain blocks.
+	is_authority: Gauge<U64>,
+	/// Tracks authority status for parachain approval checking.
+	is_parachain_validator: Gauge<U64>,
+}
+
+impl Metrics {
+	/// Dummy constructor for testing.
+	#[cfg(test)]
+	pub fn new_dummy() -> Self {
+		Self(None)
+	}
+
+	/// Set the `relaychain validator` metric.
+	pub fn on_is_authority(&self) {
+		if let Some(metrics) = &self.0 {
+			metrics.is_authority.set(1);
+		}
+	}
+
+	/// Unset the `relaychain validator` metric.
+	pub fn on_is_not_authority(&self) {
+		if let Some(metrics) = &self.0 {
+			metrics.is_authority.set(0);
+		}
+	}
+
+	/// Set the `parachain validator` metric.
+	pub fn on_is_parachain_validator(&self) {
+		if let Some(metrics) = &self.0 {
+			metrics.is_parachain_validator.set(1);
+		}
+	}
+
+	/// Unset the `parachain validator` metric.
+	pub fn on_is_not_parachain_validator(&self) {
+		if let Some(metrics) = &self.0 {
+			metrics.is_parachain_validator.set(0);
+		}
+	}
+}
+
+impl metrics::Metrics for Metrics {
+	fn try_register(registry: &Registry) -> Result<Self, PrometheusError> {
+		let metrics = MetricsInner {
+			is_authority: prometheus::register(
+				Gauge::new("polkadot_node_is_authority", "Tracks the node authority status across sessions. \
+				An authority is any node that is a potential block producer in a session.")?,
+				registry,
+			)?,
+			is_parachain_validator: prometheus::register(
+				Gauge::new("polkadot_node_is_parachain_validator", 
+				"Tracks the node parachain validator status across sessions. Parachain validators are a \
+				subset of authorities that perform approval checking of all parachain candidates in a session.")?,
+				registry,
+			)?,
+		};
+		Ok(Metrics(Some(metrics)))
+	}
+}

node/network/gossip-support/src/tests.rs

@@ -126,7 +126,11 @@ async fn get_other_authorities_addrs_map() -> HashMap<AuthorityDiscoveryId, Hash
 }
 
 fn make_subsystem() -> GossipSupport<MockAuthorityDiscovery> {
-	GossipSupport::new(make_ferdie_keystore(), MOCK_AUTHORITY_DISCOVERY.clone())
+	GossipSupport::new(
+		make_ferdie_keystore(),
+		MOCK_AUTHORITY_DISCOVERY.clone(),
+		Metrics::new_dummy(),
+	)
 }
 
 fn test_harness<T: Future<Output = VirtualOverseer>, AD: AuthorityDiscovery>(
@@ -230,6 +234,7 @@ fn issues_a_connection_request_on_new_session() {
 				tx.send(Ok(1)).unwrap();
 			}
 		);
+
 		assert_matches!(
 			overseer_recv(overseer).await,
 			AllMessages::RuntimeApi(RuntimeApiMessage::Request(
@@ -254,6 +259,17 @@ fn issues_a_connection_request_on_new_session() {
 
 		test_neighbors(overseer).await;
 
+		assert_matches!(
+			overseer_recv(overseer).await,
+			AllMessages::RuntimeApi(RuntimeApiMessage::Request(
+				relay_parent,
+				RuntimeApiRequest::SessionInfo(1, sender),
+			)) => {
+				assert_eq!(relay_parent, hash);
+				sender.send(Ok(None)).unwrap();
+			}
+		);
+
 		virtual_overseer
 	});
 
@@ -296,6 +312,7 @@ fn issues_a_connection_request_on_new_session() {
 				tx.send(Ok(2)).unwrap();
 			}
 		);
+
 		assert_matches!(
 			overseer_recv(overseer).await,
 			AllMessages::RuntimeApi(RuntimeApiMessage::Request(
@@ -320,6 +337,17 @@ fn issues_a_connection_request_on_new_session() {
 
 		test_neighbors(overseer).await;
 
+		assert_matches!(
+			overseer_recv(overseer).await,
+			AllMessages::RuntimeApi(RuntimeApiMessage::Request(
+				relay_parent,
+				RuntimeApiRequest::SessionInfo(2, sender),
+			)) => {
+				assert_eq!(relay_parent, hash);
+				sender.send(Ok(None)).unwrap();
+			}
+		);
+
 		virtual_overseer
 	});
 	assert_eq!(state.last_session_index, Some(2));
@@ -378,6 +406,7 @@ fn issues_a_connection_request_when_last_request_was_mostly_unresolved() {
 					tx.send(Ok(1)).unwrap();
 				}
 			);
+
 			assert_matches!(
 				overseer_recv(overseer).await,
 				AllMessages::RuntimeApi(RuntimeApiMessage::Request(
@@ -406,6 +435,17 @@ fn issues_a_connection_request_when_last_request_was_mostly_unresolved() {
 
 			test_neighbors(overseer).await;
 
+			assert_matches!(
+				overseer_recv(overseer).await,
+				AllMessages::RuntimeApi(RuntimeApiMessage::Request(
+					relay_parent,
+					RuntimeApiRequest::SessionInfo(1, sender),
+				)) => {
+					assert_eq!(relay_parent, hash);
+					sender.send(Ok(None)).unwrap();
+				}
+			);
+
 			virtual_overseer
 		})
 	};

node/service/src/overseer.rs

@@ -264,6 +264,7 @@ where
 		.gossip_support(GossipSupportSubsystem::new(
 			keystore.clone(),
 			authority_discovery_service.clone(),
+			Metrics::register(registry)?,
 		))
 		.dispute_coordinator(if disputes_enabled {
 			DisputeCoordinatorSubsystem::new(