PVF: Add worker check during tests and benches

polkadot/cli/src/command.rs

@@ -50,7 +50,7 @@ impl SubstrateCli for Cli {
 
 	fn impl_version() -> String {
 		let commit_hash = env!("SUBSTRATE_CLI_COMMIT_HASH");
-		format!("{NODE_VERSION}-{commit_hash}")
+		format!("{}-{commit_hash}", NODE_VERSION)
 	}
 
 	fn description() -> String {

polkadot/node/core/pvf/Cargo.toml

@@ -12,6 +12,7 @@ cfg-if = "1.0"
 futures = "0.3.21"
 futures-timer = "3.0.2"
 gum = { package = "tracing-gum", path = "../../gum" }
+is_executable = "1.0.1"
 libc = "0.2.139"
 pin-project = "1.0.9"
 rand = "0.8.5"
@@ -34,19 +35,23 @@ sp-maybe-compressed-blob = { path = "../../../../substrate/primitives/maybe-comp
 polkadot-node-core-pvf-prepare-worker = { path = "prepare-worker", optional = true }
 polkadot-node-core-pvf-execute-worker = { path = "execute-worker", optional = true }
 
-[build-dependencies]
-substrate-build-script-utils = { path = "../../../../substrate/utils/build-script-utils" }
-
 [dev-dependencies]
 assert_matches = "1.4.0"
+criterion = { version = "0.4.0", default-features = false, features = ["cargo_bench_support", "async_tokio"] }
 hex-literal = "0.4.1"
 polkadot-node-core-pvf-common = { path = "common", features = ["test-utils"] }
-# For the puppet worker, depend on ourselves with the test-utils feature.
+# For benches and integration tests, depend on ourselves with the test-utils
+# feature.
 polkadot-node-core-pvf = { path = ".", features = ["test-utils"] }
+rococo-runtime = { path = "../../../runtime/rococo" }
 
 adder = { package = "test-parachain-adder", path = "../../../parachain/test-parachains/adder" }
 halt = { package = "test-parachain-halt", path = "../../../parachain/test-parachains/halt" }
 
+[[bench]]
+name = "host_rococo_runtime"
+harness = false
+
 [features]
 ci-only-tests = []
 jemalloc-allocator = [ "polkadot-node-core-pvf-common/jemalloc-allocator" ]

polkadot/node/core/pvf/benches/host_rococo_runtime.rs

@@ -0,0 +1,220 @@
+// Copyright (C) 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/>.
+
+//! Benchmarks for preparation and execution through the host.
+//!
+//! # Examples
+//!
+//! $ cargo bench
+//! $ cargo bench prepare
+//! $ cargo bench execute
+
+use criterion::{criterion_group, criterion_main, BatchSize, Criterion, SamplingMode};
+use parity_scale_codec::Encode;
+use polkadot_node_core_pvf::{
+	start, testing, Config, Metrics, PrepareError, PrepareJobKind, PrepareStats, PvfPrepData,
+	ValidationError, ValidationHost,
+};
+use polkadot_parachain_primitives::primitives::{BlockData, ValidationParams, ValidationResult};
+use polkadot_primitives::ExecutorParams;
+use std::time::Duration;
+use tokio::{runtime::Handle, sync::Mutex};
+
+const TEST_EXECUTION_TIMEOUT: Duration = Duration::from_secs(3);
+const TEST_PREPARATION_TIMEOUT: Duration = Duration::from_secs(30);
+
+struct TestHost {
+	host: Mutex<ValidationHost>,
+}
+
+impl TestHost {
+	fn new_with_config<F>(handle: &Handle, f: F) -> Self
+	where
+		F: FnOnce(&mut Config),
+	{
+		let (prepare_worker_path, execute_worker_path) = testing::get_and_check_worker_paths();
+
+		let cache_dir = tempfile::tempdir().unwrap();
+		let mut config = Config::new(
+			cache_dir.path().to_owned(),
+			None,
+			prepare_worker_path,
+			execute_worker_path,
+		);
+		f(&mut config);
+		let (host, task) = start(config, Metrics::default());
+		let _ = handle.spawn(task);
+		Self { host: Mutex::new(host) }
+	}
+
+	async fn precheck_pvf(
+		&self,
+		code: &[u8],
+		executor_params: ExecutorParams,
+	) -> Result<PrepareStats, PrepareError> {
+		let (result_tx, result_rx) = futures::channel::oneshot::channel();
+
+		let code = sp_maybe_compressed_blob::decompress(code, 16 * 1024 * 1024)
+			.expect("Compression works");
+
+		self.host
+			.lock()
+			.await
+			.precheck_pvf(
+				PvfPrepData::from_code(
+					code.into(),
+					executor_params,
+					TEST_PREPARATION_TIMEOUT,
+					PrepareJobKind::Prechecking,
+				),
+				result_tx,
+			)
+			.await
+			.unwrap();
+		result_rx.await.unwrap()
+	}
+
+	async fn validate_candidate(
+		&self,
+		code: &[u8],
+		params: ValidationParams,
+		executor_params: ExecutorParams,
+	) -> Result<ValidationResult, ValidationError> {
+		let (result_tx, result_rx) = futures::channel::oneshot::channel();
+
+		let code = sp_maybe_compressed_blob::decompress(code, 16 * 1024 * 1024)
+			.expect("Compression works");
+
+		self.host
+			.lock()
+			.await
+			.execute_pvf(
+				PvfPrepData::from_code(
+					code.into(),
+					executor_params,
+					TEST_PREPARATION_TIMEOUT,
+					PrepareJobKind::Compilation,
+				),
+				TEST_EXECUTION_TIMEOUT,
+				params.encode(),
+				polkadot_node_core_pvf::Priority::Normal,
+				result_tx,
+			)
+			.await
+			.unwrap();
+		result_rx.await.unwrap()
+	}
+}
+
+fn host_prepare_rococo_runtime(c: &mut Criterion) {
+	polkadot_node_core_pvf_common::sp_tracing::try_init_simple();
+
+	let rt = tokio::runtime::Runtime::new().unwrap();
+
+	let blob = rococo_runtime::WASM_BINARY.unwrap();
+	let pvf = match sp_maybe_compressed_blob::decompress(&blob, 64 * 1024 * 1024) {
+		Ok(code) => PvfPrepData::from_code(
+			code.into_owned(),
+			ExecutorParams::default(),
+			Duration::from_secs(360),
+			PrepareJobKind::Compilation,
+		),
+		Err(e) => {
+			panic!("Cannot decompress blob: {:?}", e);
+		},
+	};
+
+	let mut group = c.benchmark_group("prepare rococo");
+	group.sampling_mode(SamplingMode::Flat);
+	group.sample_size(20);
+	group.measurement_time(Duration::from_secs(240));
+	group.bench_function("host: prepare Rococo runtime", |b| {
+		b.to_async(&rt).iter_batched(
+			|| {
+				(
+					TestHost::new_with_config(rt.handle(), |cfg| {
+						cfg.prepare_workers_hard_max_num = 1;
+					}),
+					pvf.clone().code(),
+				)
+			},
+			|(host, pvf_code)| async move {
+				// `PvfPrepData` is designed to be cheap to clone, so cloning shouldn't affect the
+				// benchmark accuracy.
+				let _stats = host.precheck_pvf(&pvf_code, Default::default()).await.unwrap();
+			},
+			BatchSize::SmallInput,
+		)
+	});
+	group.finish();
+}
+
+fn host_execute_rococo_runtime(c: &mut Criterion) {
+	polkadot_node_core_pvf_common::sp_tracing::try_init_simple();
+
+	let rt = tokio::runtime::Runtime::new().unwrap();
+
+	let host = TestHost::new_with_config(rt.handle(), |cfg| {
+		cfg.execute_workers_max_num = 1;
+	});
+
+	let blob = rococo_runtime::WASM_BINARY.unwrap();
+	let pvf = match sp_maybe_compressed_blob::decompress(&blob, 64 * 1024 * 1024) {
+		Ok(code) => PvfPrepData::from_code(
+			code.into_owned(),
+			ExecutorParams::default(),
+			Duration::from_secs(360),
+			PrepareJobKind::Compilation,
+		),
+		Err(e) => {
+			panic!("Cannot decompress blob: {:?}", e);
+		},
+	};
+
+	// Prepare beforehand, so that the benchmark only measures execution.
+	rt.block_on(async {
+		let _stats = host.precheck_pvf(&pvf.code(), Default::default()).await.unwrap();
+	});
+
+	let mut group = c.benchmark_group("execute rococo");
+	group.sampling_mode(SamplingMode::Flat);
+	group.sample_size(20);
+	group.measurement_time(Duration::from_secs(240));
+	group.bench_function("host: prepare Rococo runtime", |b| {
+		b.to_async(&rt).iter(|| async {
+			// `PvfPrepData` is designed to be cheap to clone, so cloning shouldn't affect the
+			// benchmark accuracy.
+			let _result = host
+				.validate_candidate(
+					&pvf.code(),
+					ValidationParams {
+						block_data: BlockData(Vec::new()),
+						parent_head: Default::default(),
+						relay_parent_number: 1,
+						relay_parent_storage_root: Default::default(),
+					},
+					Default::default(),
+				)
+				.await
+				.unwrap();
+		})
+	});
+	group.finish();
+}
+
+criterion_group!(prepare, host_prepare_rococo_runtime);
+criterion_group!(execute, host_execute_rococo_runtime);
+criterion_main!(prepare, execute);

polkadot/node/core/pvf/common/Cargo.toml

@@ -37,6 +37,5 @@ tempfile = "3.3.0"
 
 [features]
 # This feature is used to export test code to other crates without putting it in the production build.
-# Also used for building the puppet worker.
 test-utils = []
 jemalloc-allocator = []

polkadot/node/core/pvf/common/src/worker/mod.rs

@@ -35,9 +35,14 @@ use tokio::{io, runtime::Runtime};
 /// spawning the desired worker.
 #[macro_export]
 macro_rules! decl_worker_main {
-	($expected_command:expr, $entrypoint:expr, $worker_version:expr $(,)*) => {
+	($expected_command:expr, $entrypoint:expr, $worker_version:expr, $worker_version_hash:expr $(,)*) => {
+		fn get_full_version() -> String {
+			format!("{}-{}", $worker_version, $worker_version_hash)
+		}
+
 		fn print_help(expected_command: &str) {
 			println!("{} {}", expected_command, $worker_version);
+			println!("commit: {}", $worker_version_hash);
 			println!();
 			println!("PVF worker that is called by polkadot.");
 		}
@@ -67,6 +72,11 @@ macro_rules! decl_worker_main {
 					println!("{}", $worker_version);
 					return
 				},
+				// Useful for debugging. --version is used for version checks.
+				"--full-version" => {
+					println!("{}", get_full_version());
+					return
+				},
 
 				"--check-can-enable-landlock" => {
 					#[cfg(target_os = "linux")]

polkadot/node/core/pvf/src/artifacts.rs

@@ -141,6 +141,7 @@ impl ArtifactPathId {
 	}
 }
 
+#[derive(Debug)]
 pub enum ArtifactState {
 	/// The artifact is ready to be used by the executor.
 	///

polkadot/node/core/pvf/src/host.rs

@@ -446,7 +446,8 @@ async fn handle_to_host(
 /// This tries to prepare the PVF by compiling the WASM blob within a timeout set in
 /// `PvfPrepData`.
 ///
-/// If the prepare job failed previously, we may retry it under certain conditions.
+/// We don't retry artifacts that previously failed preparation. We don't expect multiple
+/// pre-checking requests.
 async fn handle_precheck_pvf(
 	artifacts: &mut Artifacts,
 	prepare_queue: &mut mpsc::Sender<prepare::ToQueue>,
@@ -464,8 +465,7 @@ async fn handle_precheck_pvf(
 			ArtifactState::Preparing { waiting_for_response, num_failures: _ } =>
 				waiting_for_response.push(result_sender),
 			ArtifactState::FailedToProcess { error, .. } => {
-				// Do not retry failed preparation if another pre-check request comes in. We do not
-				// retry pre-checking, anyway.
+				// Do not retry an artifact that previously failed preparation.
 				let _ = result_sender.send(PrepareResult::Err(error.clone()));
 			},
 		}
@@ -764,7 +764,7 @@ async fn handle_prepare_done(
 			let last_time_failed = SystemTime::now();
 			let num_failures = *num_failures + 1;
 
-			gum::warn!(
+			gum::error!(
 				target: LOG_TARGET,
 				?artifact_id,
 				time_failed = ?last_time_failed,
@@ -846,7 +846,7 @@ async fn sweeper_task(mut sweeper_rx: mpsc::Receiver<PathBuf>) {
 				gum::trace!(
 					target: LOG_TARGET,
 					?result,
-					"Sweeping the artifact file {}",
+					"Sweeped the artifact file {}",
 					condemned.display(),
 				);
 			},