Add logging to PVF and other related parts (#3596)

Co-authored-by: Bastian Köcher <bkchr@users.noreply.github.com>

polkadot/node/core/pvf/src/prepare/pool.rs

@@ -194,6 +194,7 @@ fn handle_to_pool(
 ) {
 	match to_pool {
 		ToPool::Spawn => {
+			tracing::debug!(target: LOG_TARGET, "spawning a new prepare worker");
 			mux.push(spawn_worker_task(program_path.to_owned(), spawn_timeout).boxed());
 		},
 		ToPool::StartWork { worker, code, artifact_path, background_priority } => {
@@ -224,6 +225,7 @@ fn handle_to_pool(
 			}
 		},
 		ToPool::Kill(worker) => {
+			tracing::debug!(target: LOG_TARGET, ?worker, "killing prepare worker");
 			// It may be absent if it were previously already removed by `purge_dead`.
 			let _ = spawned.remove(worker);
 		},

polkadot/node/core/pvf/src/prepare/queue.rs

@@ -212,6 +212,13 @@ async fn handle_to_queue(queue: &mut Queue, to_queue: ToQueue) -> Result<(), Fat
 }
 
 async fn handle_enqueue(queue: &mut Queue, priority: Priority, pvf: Pvf) -> Result<(), Fatal> {
+	tracing::debug!(
+		target: LOG_TARGET,
+		validation_code_hash = ?pvf.code_hash,
+		?priority,
+		"PVF is enqueued for preparation.",
+	);
+
 	let artifact_id = pvf.as_artifact_id();
 	if never!(
 		queue.artifact_id_to_job.contains_key(&artifact_id),
@@ -254,8 +261,14 @@ async fn handle_amend(
 	artifact_id: ArtifactId,
 ) -> Result<(), Fatal> {
 	if let Some(&job) = queue.artifact_id_to_job.get(&artifact_id) {
-		let mut job_data: &mut JobData = &mut queue.jobs[job];
+		tracing::debug!(
+			target: LOG_TARGET,
+			validation_code_hash = ?artifact_id.code_hash,
+			?priority,
+			"amending preparation priority.",
+		);
 
+		let mut job_data: &mut JobData = &mut queue.jobs[job];
 		if job_data.priority < priority {
 			// The new priority is higher. We should do two things:
 			// - if the worker was already spawned with the background prio and the new one is not
@@ -349,6 +362,14 @@ async fn handle_worker_concluded(
 
 	queue.artifact_id_to_job.remove(&artifact_id);
 
+	tracing::debug!(
+		target: LOG_TARGET,
+		validation_code_hash = ?artifact_id.code_hash,
+		?worker,
+		?rip,
+		"prepare worker concluded",
+	);
+
 	reply(&mut queue.from_queue_tx, FromQueue::Prepared(artifact_id))?;
 
 	// Figure out what to do with the worker.
@@ -380,8 +401,9 @@ async fn handle_worker_concluded(
 }
 
 async fn handle_worker_rip(queue: &mut Queue, worker: Worker) -> Result<(), Fatal> {
-	let worker_data = queue.workers.remove(worker);
+	tracing::debug!(target: LOG_TARGET, ?worker, "prepare worker ripped");
 
+	let worker_data = queue.workers.remove(worker);
 	if let Some(WorkerData { job: Some(job), .. }) = worker_data {
 		// This is an edge case where the worker ripped after we sent assignment but before it
 		// was received by the pool.

polkadot/node/core/pvf/src/prepare/worker.rs

@@ -103,6 +103,7 @@ pub async fn start_work(
 		// We may potentially overwrite the artifact in rare cases where the worker didn't make
 		// it to report back the result.
 
+		#[derive(Debug)]
 		enum Selected {
 			Done,
 			IoErr,
@@ -170,7 +171,15 @@ pub async fn start_work(
 			Selected::IoErr | Selected::Deadline => {
 				let bytes = Artifact::DidntMakeIt.serialize();
 				// best effort: there is nothing we can do here if the write fails.
-				let _ = async_std::fs::write(&artifact_path, &bytes).await;
+				if let Err(err) = async_std::fs::write(&artifact_path, &bytes).await {
+					tracing::warn!(
+						target: LOG_TARGET,
+						worker_pid = %pid,
+						"preparation didn't make it, because of `{:?}`: {:?}",
+						selected,
+						err,
+					);
+				}
 				Outcome::DidntMakeIt
 			},
 		}