Switch to latest Jaeger and improve the spans (#2216)

* Switch to latest Jaeger and improve the spans

* Update node/jaeger/src/lib.rs

Co-authored-by: Robert Habermeier <rphmeier@gmail.com>

* Use better span in bitfield signing

* Update node/core/bitfield-signing/src/lib.rs

Co-authored-by: Andronik Ordian <write@reusable.software>

Co-authored-by: Robert Habermeier <rphmeier@gmail.com>
Co-authored-by: Andronik Ordian <write@reusable.software>

Cargo.lock

@@ -3477,9 +3477,9 @@ dependencies = [
 
 [[package]]
 name = "mick-jaeger"
-version = "0.1.2"
+version = "0.1.4"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "4247b181b96e62eacb5a1e7a2f1a39a78b945cb0434c0fceaae4ed1882555957"
+checksum = "c023c3f16109e7f33aa451f773fd61070e265b4977d0b6e344a51049296dd7df"
 dependencies = [
  "futures 0.3.8",
  "rand 0.7.3",

node/core/bitfield-signing/src/lib.rs

@@ -78,7 +78,7 @@ async fn get_core_availability(
 	span: &jaeger::JaegerSpan,
 ) -> Result<bool, Error> {
 	if let CoreState::Occupied(core) = core {
-		let _span = span.child("query chunk availability");
+		let _span = span.child("query-chunk-availability");
 
 		let (tx, rx) = oneshot::channel();
 		sender
@@ -246,12 +246,12 @@ impl JobTrait for BitfieldSigningJob {
 			let _timer = metrics.time_run();
 
 			drop(_span);
-			let _span = span.child("availablity");
+			let span_availability = span.child("availability");
 
 			let bitfield =
 				match construct_availability_bitfield(
 					relay_parent,
-					&span,
+					&span_availability,
 					validator.index(),
 					&mut sender,
 				).await
@@ -265,7 +265,7 @@ impl JobTrait for BitfieldSigningJob {
 				Ok(bitfield) => bitfield,
 			};
 
-			drop(_span);
+			drop(span_availability);
 			let _span = span.child("signing");
 
 			let signed_bitfield = validator

node/jaeger/Cargo.toml

@@ -7,7 +7,7 @@ description = "Polkadot Jaeger primitives"
 
 [dependencies]
 async-std = "1.8.0"
-mick-jaeger = "0.1.2"
+mick-jaeger = "0.1.4"
 lazy_static = "1.4"
 parking_lot = "0.11.1"
 polkadot-primitives = { path = "../../primitives" }

node/jaeger/src/lib.rs

@@ -183,6 +183,14 @@ impl JaegerSpan {
 			Self::Disabled => {},
 		}
 	}
+
+	/// Adds the `FollowsFrom` relationship to this span with respect to the given one.
+	pub fn add_follows_from(&mut self, other: &Self) {
+		match (self, other) {
+			(Self::Enabled(ref mut inner), Self::Enabled(ref other_inner)) => inner.add_follows_from(&other_inner),
+			_ => {},
+		}
+	}
 }
 
 impl std::fmt::Debug for JaegerSpan {

node/network/availability-distribution/src/lib.rs

@@ -45,7 +45,7 @@ use polkadot_subsystem::messages::{
 	NetworkBridgeMessage, RuntimeApiMessage, RuntimeApiRequest,
 };
 use polkadot_subsystem::{
-	jaeger, errors::{ChainApiError, RuntimeApiError},
+	jaeger, errors::{ChainApiError, RuntimeApiError}, PerLeafSpan,
 	ActiveLeavesUpdate, FromOverseer, OverseerSignal, SpawnedSubsystem, Subsystem, SubsystemContext, SubsystemError,
 };
 use std::collections::{HashMap, HashSet};
@@ -191,12 +191,14 @@ impl PerCandidate {
 	}
 }
 
-#[derive(Debug, Clone, Default)]
+#[derive(Debug)]
 struct PerRelayParent {
 	/// Set of `K` ancestors for this relay parent.
 	ancestors: Vec<Hash>,
 	/// Live candidates, according to this relay parent.
 	live_candidates: HashSet<CandidateHash>,
+	/// The span that belongs to this relay parent.
+	span: PerLeafSpan,
 }
 
 impl ProtocolState {
@@ -216,7 +218,7 @@ impl ProtocolState {
 		)
 	}
 
-	#[tracing::instrument(level = "trace", skip(candidates), fields(subsystem = LOG_TARGET))]
+	#[tracing::instrument(level = "trace", skip(candidates, span), fields(subsystem = LOG_TARGET))]
 	fn add_relay_parent(
 		&mut self,
 		relay_parent: Hash,
@@ -224,10 +226,13 @@ impl ProtocolState {
 		validator_index: Option<ValidatorIndex>,
 		candidates: HashMap<CandidateHash, FetchedLiveCandidate>,
 		ancestors: Vec<Hash>,
+		span: PerLeafSpan,
 	) {
-		let per_relay_parent = self.per_relay_parent.entry(relay_parent).or_default();
-		per_relay_parent.ancestors = ancestors;
-		per_relay_parent.live_candidates.extend(candidates.keys().cloned());
+		let per_relay_parent = self.per_relay_parent.entry(relay_parent).or_insert_with(|| PerRelayParent {
+			span,
+			ancestors,
+			live_candidates: candidates.keys().cloned().collect(),
+		});
 
 		// register the relation of relay_parent to candidate..
 		for (receipt_hash, fetched) in candidates {
@@ -256,6 +261,11 @@ impl ProtocolState {
 				}
 			};
 
+			// Create some span that will make it able to switch between the candidate and relay parent span.
+			let mut span = per_relay_parent.span.child("live-candidate");
+			span.add_string_tag("candidate-hash", &format!("{:?}", receipt_hash));
+
+			candidate_entry.span.add_follows_from(&span);
 			candidate_entry.live_in.insert(relay_parent);
 		}
 	}
@@ -365,7 +375,9 @@ where
 	let view = state.view.clone();
 
 	// add all the relay parents and fill the cache
-	for added in view.difference(&old_view) {
+	for (added, span) in view.span_per_head().iter().filter(|v| !old_view.contains(&v.0)) {
+		let span = PerLeafSpan::new(span.clone(), "availability-distribution");
+
 		let validators = query_validators(ctx, *added).await?;
 		let validator_index = obtain_our_validator_index(&validators, keystore.clone()).await;
 		let (candidates, ancestors)
@@ -377,6 +389,7 @@ where
 			validator_index,
 			candidates,
 			ancestors,
+			span,
 		);
 	}
 
@@ -436,7 +449,6 @@ where
 					"Retrieved chunk from availability storage",
 				);
 
-
 				let msg = AvailabilityGossipMessage {
 					candidate_hash,
 					erasure_chunk,

node/network/availability-distribution/src/tests.rs

@@ -1005,9 +1005,14 @@ fn clean_up_receipts_cache_unions_ancestors_and_view() {
 	state.per_relay_parent.insert(hash_a, PerRelayParent {
 		ancestors: vec![hash_b],
 		live_candidates: HashSet::new(),
+		span: PerLeafSpan::new(Arc::new(jaeger::JaegerSpan::Disabled), "test"),
 	});
 
-	state.per_relay_parent.insert(hash_c, PerRelayParent::default());
+	state.per_relay_parent.insert(hash_c, PerRelayParent {
+		ancestors: Vec::new(),
+		live_candidates: HashSet::new(),
+		span: PerLeafSpan::new(Arc::new(jaeger::JaegerSpan::Disabled), "test"),
+	});
 
 	state.clean_up_live_under_cache();
 
@@ -1030,11 +1035,13 @@ fn remove_relay_parent_only_removes_per_candidate_if_final() {
 	state.per_relay_parent.insert(hash_a, PerRelayParent {
 		ancestors: vec![],
 		live_candidates: std::iter::once(candidate_hash_a).collect(),
+		span: PerLeafSpan::new(Arc::new(jaeger::JaegerSpan::Disabled), "test"),
 	});
 
 	state.per_relay_parent.insert(hash_b, PerRelayParent {
 		ancestors: vec![],
 		live_candidates: std::iter::once(candidate_hash_a).collect(),
+		span: PerLeafSpan::new(Arc::new(jaeger::JaegerSpan::Disabled), "test"),
 	});
 
 	state.per_candidate.insert(candidate_hash_a, {
@@ -1079,6 +1086,7 @@ fn add_relay_parent_includes_all_live_candidates() {
 		None,
 		candidates,
 		vec![ancestor_a],
+		PerLeafSpan::new(Arc::new(jaeger::JaegerSpan::Disabled), "test"),
 	);
 
 	assert!(

node/overseer/src/lib.rs

@@ -1510,7 +1510,7 @@ where
 
 		for (hash, number) in std::mem::take(&mut self.leaves) {
 			let _ = self.active_leaves.insert(hash, number);
-			let span = self.on_head_activated(&hash);
+			let span = self.on_head_activated(&hash, None);
 			update.activated.push((hash, span));
 		}
 
@@ -1589,7 +1589,7 @@ where
 			}
 		};
 
-		let span = self.on_head_activated(&block.hash);
+		let span = self.on_head_activated(&block.hash, Some(block.parent_hash));
 		let mut update = ActiveLeavesUpdate::start_work(block.hash, span);
 
 		if let Some(number) = self.active_leaves.remove(&block.parent_hash) {
@@ -1705,7 +1705,7 @@ where
 	}
 
 	#[tracing::instrument(level = "trace", skip(self), fields(subsystem = LOG_TARGET))]
-	fn on_head_activated(&mut self, hash: &Hash) -> Arc<JaegerSpan> {
+	fn on_head_activated(&mut self, hash: &Hash, parent_hash: Option<Hash>) -> Arc<JaegerSpan> {
 		self.metrics.on_head_activated();
 		if let Some(listeners) = self.activation_external_listeners.remove(hash) {
 			for listener in listeners {
@@ -1714,7 +1714,13 @@ where
 			}
 		}
 
-		let span = Arc::new(jaeger::hash_span(hash, "leave activated"));
+		let mut span = jaeger::hash_span(hash, "leaf-activated");
+
+		if let Some(parent_span) = parent_hash.and_then(|h| self.span_per_active_leaf.get(&h)) {
+			span.add_follows_from(&*parent_span);
+		}
+
+		let span = Arc::new(span);
 		self.span_per_active_leaf.insert(*hash, span.clone());
 		span
 	}