PVF: Move PVF workers into separate crate (#7101)

* Move PVF workers into separate crate

* Fix indentation

* Fix compilation errors

* Fix more compilation errors

* Rename `worker.rs` files, make host interface to worker more clear

* Fix more compilation errors

* Fix more compilation errors

* Add link to issue

* Address review comments

* Update comment
This commit is contained in:
Marcin S
2023-04-21 12:40:09 +02:00
committed by GitHub
parent ac09a84115
commit e277f95b3b
42 changed files with 878 additions and 627 deletions
-148
View File
@@ -1,148 +0,0 @@
// 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/>.
use super::TestHost;
use adder::{hash_state, BlockData, HeadData};
use parity_scale_codec::{Decode, Encode};
use polkadot_parachain::primitives::{
BlockData as GenericBlockData, HeadData as GenericHeadData, RelayChainBlockNumber,
ValidationParams,
};
#[tokio::test]
async fn execute_good_block_on_parent() {
let parent_head = HeadData { number: 0, parent_hash: [0; 32], post_state: hash_state(0) };
let block_data = BlockData { state: 0, add: 512 };
let host = TestHost::new();
let ret = host
.validate_candidate(
adder::wasm_binary_unwrap(),
ValidationParams {
parent_head: GenericHeadData(parent_head.encode()),
block_data: GenericBlockData(block_data.encode()),
relay_parent_number: 1,
relay_parent_storage_root: Default::default(),
},
Default::default(),
)
.await
.unwrap();
let new_head = HeadData::decode(&mut &ret.head_data.0[..]).unwrap();
assert_eq!(new_head.number, 1);
assert_eq!(new_head.parent_hash, parent_head.hash());
assert_eq!(new_head.post_state, hash_state(512));
}
#[tokio::test]
async fn execute_good_chain_on_parent() {
let mut number = 0;
let mut parent_hash = [0; 32];
let mut last_state = 0;
let host = TestHost::new();
for add in 0..10 {
let parent_head = HeadData { number, parent_hash, post_state: hash_state(last_state) };
let block_data = BlockData { state: last_state, add };
let ret = host
.validate_candidate(
adder::wasm_binary_unwrap(),
ValidationParams {
parent_head: GenericHeadData(parent_head.encode()),
block_data: GenericBlockData(block_data.encode()),
relay_parent_number: number as RelayChainBlockNumber + 1,
relay_parent_storage_root: Default::default(),
},
Default::default(),
)
.await
.unwrap();
let new_head = HeadData::decode(&mut &ret.head_data.0[..]).unwrap();
assert_eq!(new_head.number, number + 1);
assert_eq!(new_head.parent_hash, parent_head.hash());
assert_eq!(new_head.post_state, hash_state(last_state + add));
number += 1;
parent_hash = new_head.hash();
last_state += add;
}
}
#[tokio::test]
async fn execute_bad_block_on_parent() {
let parent_head = HeadData { number: 0, parent_hash: [0; 32], post_state: hash_state(0) };
let block_data = BlockData {
state: 256, // start state is wrong.
add: 256,
};
let host = TestHost::new();
let _ret = host
.validate_candidate(
adder::wasm_binary_unwrap(),
ValidationParams {
parent_head: GenericHeadData(parent_head.encode()),
block_data: GenericBlockData(block_data.encode()),
relay_parent_number: 1,
relay_parent_storage_root: Default::default(),
},
Default::default(),
)
.await
.unwrap_err();
}
#[tokio::test]
async fn stress_spawn() {
let host = std::sync::Arc::new(TestHost::new());
async fn execute(host: std::sync::Arc<TestHost>) {
let parent_head = HeadData { number: 0, parent_hash: [0; 32], post_state: hash_state(0) };
let block_data = BlockData { state: 0, add: 512 };
let ret = host
.validate_candidate(
adder::wasm_binary_unwrap(),
ValidationParams {
parent_head: GenericHeadData(parent_head.encode()),
block_data: GenericBlockData(block_data.encode()),
relay_parent_number: 1,
relay_parent_storage_root: Default::default(),
},
Default::default(),
)
.await
.unwrap();
let new_head = HeadData::decode(&mut &ret.head_data.0[..]).unwrap();
assert_eq!(new_head.number, 1);
assert_eq!(new_head.parent_hash, parent_head.hash());
assert_eq!(new_head.post_state, hash_state(512));
}
futures::future::join_all((0..100).map(|_| execute(host.clone()))).await;
}
-297
View File
@@ -1,297 +0,0 @@
// 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/>.
use assert_matches::assert_matches;
use parity_scale_codec::Encode as _;
use polkadot_node_core_pvf::{
start, Config, InvalidCandidate, Metrics, PvfPrepData, ValidationError, ValidationHost,
JOB_TIMEOUT_WALL_CLOCK_FACTOR,
};
use polkadot_parachain::primitives::{BlockData, ValidationParams, ValidationResult};
use polkadot_primitives::{ExecutorParam, ExecutorParams};
use std::time::Duration;
use tokio::sync::Mutex;
mod adder;
mod worker_common;
const PUPPET_EXE: &str = env!("CARGO_BIN_EXE_puppet_worker");
const TEST_EXECUTION_TIMEOUT: Duration = Duration::from_secs(3);
const TEST_PREPARATION_TIMEOUT: Duration = Duration::from_secs(3);
struct TestHost {
cache_dir: tempfile::TempDir,
host: Mutex<ValidationHost>,
}
impl TestHost {
fn new() -> Self {
Self::new_with_config(|_| ())
}
fn new_with_config<F>(f: F) -> Self
where
F: FnOnce(&mut Config),
{
let cache_dir = tempfile::tempdir().unwrap();
let program_path = std::path::PathBuf::from(PUPPET_EXE);
let mut config = Config::new(cache_dir.path().to_owned(), program_path);
f(&mut config);
let (host, task) = start(config, Metrics::default());
let _ = tokio::task::spawn(task);
Self { cache_dir, host: Mutex::new(host) }
}
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),
TEST_EXECUTION_TIMEOUT,
params.encode(),
polkadot_node_core_pvf::Priority::Normal,
result_tx,
)
.await
.unwrap();
result_rx.await.unwrap()
}
}
#[tokio::test]
async fn terminates_on_timeout() {
let host = TestHost::new();
let start = std::time::Instant::now();
let result = host
.validate_candidate(
halt::wasm_binary_unwrap(),
ValidationParams {
block_data: BlockData(Vec::new()),
parent_head: Default::default(),
relay_parent_number: 1,
relay_parent_storage_root: Default::default(),
},
Default::default(),
)
.await;
match result {
Err(ValidationError::InvalidCandidate(InvalidCandidate::HardTimeout)) => {},
r => panic!("{:?}", r),
}
let duration = std::time::Instant::now().duration_since(start);
assert!(duration >= TEST_EXECUTION_TIMEOUT);
assert!(duration < TEST_EXECUTION_TIMEOUT * JOB_TIMEOUT_WALL_CLOCK_FACTOR);
}
#[tokio::test]
async fn ensure_parallel_execution() {
// Run some jobs that do not complete, thus timing out.
let host = TestHost::new();
let execute_pvf_future_1 = host.validate_candidate(
halt::wasm_binary_unwrap(),
ValidationParams {
block_data: BlockData(Vec::new()),
parent_head: Default::default(),
relay_parent_number: 1,
relay_parent_storage_root: Default::default(),
},
Default::default(),
);
let execute_pvf_future_2 = host.validate_candidate(
halt::wasm_binary_unwrap(),
ValidationParams {
block_data: BlockData(Vec::new()),
parent_head: Default::default(),
relay_parent_number: 1,
relay_parent_storage_root: Default::default(),
},
Default::default(),
);
let start = std::time::Instant::now();
let (res1, res2) = futures::join!(execute_pvf_future_1, execute_pvf_future_2);
assert_matches!(
(res1, res2),
(
Err(ValidationError::InvalidCandidate(InvalidCandidate::HardTimeout)),
Err(ValidationError::InvalidCandidate(InvalidCandidate::HardTimeout))
)
);
// Total time should be < 2 x TEST_EXECUTION_TIMEOUT (two workers run in parallel).
let duration = std::time::Instant::now().duration_since(start);
let max_duration = 2 * TEST_EXECUTION_TIMEOUT;
assert!(
duration < max_duration,
"Expected duration {}ms to be less than {}ms",
duration.as_millis(),
max_duration.as_millis()
);
}
#[tokio::test]
async fn execute_queue_doesnt_stall_if_workers_died() {
let host = TestHost::new_with_config(|cfg| {
cfg.execute_workers_max_num = 5;
});
// Here we spawn 8 validation jobs for the `halt` PVF and share those between 5 workers. The
// first five jobs should timeout and the workers killed. For the next 3 jobs a new batch of
// workers should be spun up.
let start = std::time::Instant::now();
futures::future::join_all((0u8..=8).map(|_| {
host.validate_candidate(
halt::wasm_binary_unwrap(),
ValidationParams {
block_data: BlockData(Vec::new()),
parent_head: Default::default(),
relay_parent_number: 1,
relay_parent_storage_root: Default::default(),
},
Default::default(),
)
}))
.await;
// Total time should be >= 2 x TEST_EXECUTION_TIMEOUT (two separate sets of workers that should
// both timeout).
let duration = std::time::Instant::now().duration_since(start);
let max_duration = 2 * TEST_EXECUTION_TIMEOUT;
assert!(
duration >= max_duration,
"Expected duration {}ms to be greater than or equal to {}ms",
duration.as_millis(),
max_duration.as_millis()
);
}
#[tokio::test]
async fn execute_queue_doesnt_stall_with_varying_executor_params() {
let host = TestHost::new_with_config(|cfg| {
cfg.execute_workers_max_num = 2;
});
let executor_params_1 = ExecutorParams::default();
let executor_params_2 = ExecutorParams::from(&[ExecutorParam::StackLogicalMax(1024)][..]);
// Here we spawn 6 validation jobs for the `halt` PVF and share those between 2 workers. Every
// 3rd job will have different set of executor parameters. All the workers should be killed
// and in this case the queue should respawn new workers with needed executor environment
// without waiting. The jobs will be executed in 3 batches, each running two jobs in parallel,
// and execution time would be roughly 3 * TEST_EXECUTION_TIMEOUT
let start = std::time::Instant::now();
futures::future::join_all((0u8..6).map(|i| {
host.validate_candidate(
halt::wasm_binary_unwrap(),
ValidationParams {
block_data: BlockData(Vec::new()),
parent_head: Default::default(),
relay_parent_number: 1,
relay_parent_storage_root: Default::default(),
},
match i % 3 {
0 => executor_params_1.clone(),
_ => executor_params_2.clone(),
},
)
}))
.await;
let duration = std::time::Instant::now().duration_since(start);
let min_duration = 3 * TEST_EXECUTION_TIMEOUT;
let max_duration = 4 * TEST_EXECUTION_TIMEOUT;
assert!(
duration >= min_duration,
"Expected duration {}ms to be greater than or equal to {}ms",
duration.as_millis(),
min_duration.as_millis()
);
assert!(
duration <= max_duration,
"Expected duration {}ms to be less than or equal to {}ms",
duration.as_millis(),
max_duration.as_millis()
);
}
// Test that deleting a prepared artifact does not lead to a dispute when we try to execute it.
#[tokio::test]
async fn deleting_prepared_artifact_does_not_dispute() {
let host = TestHost::new();
let cache_dir = host.cache_dir.path().clone();
let result = host
.validate_candidate(
halt::wasm_binary_unwrap(),
ValidationParams {
block_data: BlockData(Vec::new()),
parent_head: Default::default(),
relay_parent_number: 1,
relay_parent_storage_root: Default::default(),
},
Default::default(),
)
.await;
match result {
Err(ValidationError::InvalidCandidate(InvalidCandidate::HardTimeout)) => {},
r => panic!("{:?}", r),
}
// Delete the prepared artifact.
{
// Get the artifact path (asserting it exists).
let mut cache_dir: Vec<_> = std::fs::read_dir(cache_dir).unwrap().collect();
assert_eq!(cache_dir.len(), 1);
let artifact_path = cache_dir.pop().unwrap().unwrap();
// Delete the artifact.
std::fs::remove_file(artifact_path.path()).unwrap();
}
// Try to validate again, artifact should get recreated.
let result = host
.validate_candidate(
halt::wasm_binary_unwrap(),
ValidationParams {
block_data: BlockData(Vec::new()),
parent_head: Default::default(),
relay_parent_number: 1,
relay_parent_storage_root: Default::default(),
},
Default::default(),
)
.await;
match result {
Err(ValidationError::InvalidCandidate(InvalidCandidate::HardTimeout)) => {},
r => panic!("{:?}", r),
}
}
@@ -1,48 +0,0 @@
// 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/>.
use crate::PUPPET_EXE;
use polkadot_node_core_pvf::testing::worker_common::{spawn_with_program_path, SpawnErr};
use std::time::Duration;
// Test spawning a program that immediately exits with a failure code.
#[tokio::test]
async fn spawn_immediate_exit() {
let result =
spawn_with_program_path("integration-test", PUPPET_EXE, &["exit"], Duration::from_secs(2))
.await;
assert!(matches!(result, Err(SpawnErr::AcceptTimeout)));
}
#[tokio::test]
async fn spawn_timeout() {
let result =
spawn_with_program_path("integration-test", PUPPET_EXE, &["sleep"], Duration::from_secs(2))
.await;
assert!(matches!(result, Err(SpawnErr::AcceptTimeout)));
}
#[tokio::test]
async fn should_connect() {
let _ = spawn_with_program_path(
"integration-test",
PUPPET_EXE,
&["prepare-worker"],
Duration::from_secs(2),
)
.await
.unwrap();
}