Allow to specify multiple relay chain RPC urls for collator node (#1880)

* Allow specification of multiple urls for relay chain rpc nodes

* Add pooled RPC client basics

* Add list of clients to pooled client

* Improve

* Forward requests to dispatcher

* Switch clients on error

* Implement rotation logic

* Improve subscription handling

* Error handling cleanup

* Remove retry from rpc-client

* Improve naming

* Improve documentation

* Improve `ClientManager` abstraction

* Adjust zombienet test

* Add more comments

* fmt

* Apply reviewers comments

* Extract reconnection to extra method

* Add comment to reconnection method

* Clean up some dependencies

* Fix build

* fmt

* Provide alias for cli argument

* Apply review comments

* Rename P* to Relay*

* Improve zombienet test

* fmt

* Fix zombienet sleep

* Simplify zombienet test

* Reduce log clutter and fix starting position

* Do not distribute duplicated imported and finalized blocks

* fmt

* Apply code review suggestions

* Move building of relay chain interface to `cumulus-client-service`

* Refactoring to not push back into channel

* FMT
This commit is contained in:
Sebastian Kunert
2022-12-15 11:42:07 +01:00
committed by GitHub
parent 50f212d8c0
commit 7cab12e9d2
22 changed files with 720 additions and 428 deletions
@@ -19,7 +19,7 @@ use core::time::Duration;
use cumulus_primitives_core::{
relay_chain::{
v2::{CommittedCandidateReceipt, OccupiedCoreAssumption, SessionIndex, ValidatorId},
Hash as PHash, Header as PHeader, InboundHrmpMessage,
Hash as RelayHash, Header as RelayHeader, InboundHrmpMessage,
},
InboundDownwardMessage, ParaId, PersistedValidationData,
};
@@ -34,6 +34,7 @@ use std::pin::Pin;
pub use url::Url;
mod reconnecting_ws_client;
mod rpc_client;
pub use rpc_client::{create_client_and_start_worker, RelayChainRpcClient};
@@ -58,7 +59,7 @@ impl RelayChainInterface for RelayChainRpcInterface {
async fn retrieve_dmq_contents(
&self,
para_id: ParaId,
relay_parent: PHash,
relay_parent: RelayHash,
) -> RelayChainResult<Vec<InboundDownwardMessage>> {
self.rpc_client.parachain_host_dmq_contents(para_id, relay_parent).await
}
@@ -66,7 +67,7 @@ impl RelayChainInterface for RelayChainRpcInterface {
async fn retrieve_all_inbound_hrmp_channel_contents(
&self,
para_id: ParaId,
relay_parent: PHash,
relay_parent: RelayHash,
) -> RelayChainResult<BTreeMap<ParaId, Vec<InboundHrmpMessage>>> {
self.rpc_client
.parachain_host_inbound_hrmp_channels_contents(para_id, relay_parent)
@@ -75,7 +76,7 @@ impl RelayChainInterface for RelayChainRpcInterface {
async fn persisted_validation_data(
&self,
hash: PHash,
hash: RelayHash,
para_id: ParaId,
occupied_core_assumption: OccupiedCoreAssumption,
) -> RelayChainResult<Option<PersistedValidationData>> {
@@ -86,7 +87,7 @@ impl RelayChainInterface for RelayChainRpcInterface {
async fn candidate_pending_availability(
&self,
hash: PHash,
hash: RelayHash,
para_id: ParaId,
) -> RelayChainResult<Option<CommittedCandidateReceipt>> {
self.rpc_client
@@ -94,31 +95,31 @@ impl RelayChainInterface for RelayChainRpcInterface {
.await
}
async fn session_index_for_child(&self, hash: PHash) -> RelayChainResult<SessionIndex> {
async fn session_index_for_child(&self, hash: RelayHash) -> RelayChainResult<SessionIndex> {
self.rpc_client.parachain_host_session_index_for_child(hash).await
}
async fn validators(&self, block_id: PHash) -> RelayChainResult<Vec<ValidatorId>> {
async fn validators(&self, block_id: RelayHash) -> RelayChainResult<Vec<ValidatorId>> {
self.rpc_client.parachain_host_validators(block_id).await
}
async fn import_notification_stream(
&self,
) -> RelayChainResult<Pin<Box<dyn Stream<Item = PHeader> + Send>>> {
let imported_headers_stream = self.rpc_client.get_imported_heads_stream().await?;
) -> RelayChainResult<Pin<Box<dyn Stream<Item = RelayHeader> + Send>>> {
let imported_headers_stream = self.rpc_client.get_imported_heads_stream()?;
Ok(imported_headers_stream.boxed())
}
async fn finality_notification_stream(
&self,
) -> RelayChainResult<Pin<Box<dyn Stream<Item = PHeader> + Send>>> {
let imported_headers_stream = self.rpc_client.get_finalized_heads_stream().await?;
) -> RelayChainResult<Pin<Box<dyn Stream<Item = RelayHeader> + Send>>> {
let imported_headers_stream = self.rpc_client.get_finalized_heads_stream()?;
Ok(imported_headers_stream.boxed())
}
async fn best_block_hash(&self) -> RelayChainResult<PHash> {
async fn best_block_hash(&self) -> RelayChainResult<RelayHash> {
self.rpc_client.chain_get_head(None).await
}
@@ -132,7 +133,7 @@ impl RelayChainInterface for RelayChainRpcInterface {
async fn get_storage_by_key(
&self,
relay_parent: PHash,
relay_parent: RelayHash,
key: &[u8],
) -> RelayChainResult<Option<StorageValue>> {
let storage_key = StorageKey(key.to_vec());
@@ -144,7 +145,7 @@ impl RelayChainInterface for RelayChainRpcInterface {
async fn prove_read(
&self,
relay_parent: PHash,
relay_parent: RelayHash,
relevant_keys: &Vec<Vec<u8>>,
) -> RelayChainResult<StorageProof> {
let cloned = relevant_keys.clone();
@@ -167,8 +168,8 @@ impl RelayChainInterface for RelayChainRpcInterface {
/// 2. Check if the block is already in chain. If yes, succeed early.
/// 3. Wait for the block to be imported via subscription.
/// 4. If timeout is reached, we return an error.
async fn wait_for_block(&self, wait_for_hash: PHash) -> RelayChainResult<()> {
let mut head_stream = self.rpc_client.get_imported_heads_stream().await?;
async fn wait_for_block(&self, wait_for_hash: RelayHash) -> RelayChainResult<()> {
let mut head_stream = self.rpc_client.get_imported_heads_stream()?;
if self.rpc_client.chain_get_header(Some(wait_for_hash)).await?.is_some() {
return Ok(())
@@ -191,8 +192,8 @@ impl RelayChainInterface for RelayChainRpcInterface {
async fn new_best_notification_stream(
&self,
) -> RelayChainResult<Pin<Box<dyn Stream<Item = PHeader> + Send>>> {
let imported_headers_stream = self.rpc_client.get_best_heads_stream().await?;
) -> RelayChainResult<Pin<Box<dyn Stream<Item = RelayHeader> + Send>>> {
let imported_headers_stream = self.rpc_client.get_best_heads_stream()?;
Ok(imported_headers_stream.boxed())
}
}
@@ -0,0 +1,520 @@
// Copyright 2022 Parity Technologies (UK) Ltd.
// This file is part of Cumulus.
// Cumulus 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.
// Cumulus 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 Cumulus. If not, see <http://www.gnu.org/licenses/>.
use cumulus_primitives_core::relay_chain::{
BlockNumber as RelayBlockNumber, Header as RelayHeader,
};
use cumulus_relay_chain_interface::{RelayChainError, RelayChainResult};
use futures::{
channel::{
mpsc::{Receiver, Sender},
oneshot::Sender as OneshotSender,
},
future::BoxFuture,
stream::FuturesUnordered,
FutureExt, StreamExt,
};
use jsonrpsee::{
core::{
client::{Client as JsonRpcClient, ClientT, Subscription, SubscriptionClientT},
params::ArrayParams,
Error as JsonRpseeError, JsonValue,
},
rpc_params,
ws_client::WsClientBuilder,
};
use lru::LruCache;
use polkadot_service::TaskManager;
use std::{num::NonZeroUsize, sync::Arc};
use tokio::sync::mpsc::{
channel as tokio_channel, Receiver as TokioReceiver, Sender as TokioSender,
};
use url::Url;
const NOTIFICATION_CHANNEL_SIZE_LIMIT: usize = 20;
const LOG_TARGET: &str = "reconnecting-websocket-client";
/// Messages for communication between [`ReconnectingWsClient`] and [`ReconnectingWebsocketWorker`].
#[derive(Debug)]
enum RpcDispatcherMessage {
RegisterBestHeadListener(Sender<RelayHeader>),
RegisterImportListener(Sender<RelayHeader>),
RegisterFinalizationListener(Sender<RelayHeader>),
Request(String, ArrayParams, OneshotSender<Result<JsonValue, JsonRpseeError>>),
}
/// Frontend for performing websocket requests.
/// Requests and stream requests are forwarded to [`ReconnectingWebsocketWorker`].
#[derive(Debug, Clone)]
pub struct ReconnectingWsClient {
/// Channel to communicate with the RPC worker
to_worker_channel: TokioSender<RpcDispatcherMessage>,
}
impl ReconnectingWsClient {
/// Create a new websocket client frontend.
pub async fn new(urls: Vec<Url>, task_manager: &mut TaskManager) -> RelayChainResult<Self> {
tracing::debug!(target: LOG_TARGET, "Instantiating reconnecting websocket client");
let (worker, sender) = ReconnectingWebsocketWorker::new(urls).await;
task_manager
.spawn_essential_handle()
.spawn("relay-chain-rpc-worker", None, worker.run());
Ok(Self { to_worker_channel: sender })
}
}
impl ReconnectingWsClient {
/// Perform a request via websocket connection.
pub async fn request<R>(&self, method: &str, params: ArrayParams) -> Result<R, RelayChainError>
where
R: serde::de::DeserializeOwned,
{
let (tx, rx) = futures::channel::oneshot::channel();
let message = RpcDispatcherMessage::Request(method.into(), params, tx);
self.to_worker_channel.send(message).await.map_err(|err| {
RelayChainError::WorkerCommunicationError(format!(
"Unable to send message to RPC worker: {}",
err
))
})?;
let value = rx.await.map_err(|err| {
RelayChainError::WorkerCommunicationError(format!(
"Unexpected channel close on RPC worker side: {}",
err
))
})??;
serde_json::from_value(value)
.map_err(|_| RelayChainError::GenericError("Unable to deserialize value".to_string()))
}
/// Get a stream of new best relay chain headers
pub fn get_best_heads_stream(&self) -> Result<Receiver<RelayHeader>, RelayChainError> {
let (tx, rx) =
futures::channel::mpsc::channel::<RelayHeader>(NOTIFICATION_CHANNEL_SIZE_LIMIT);
self.send_register_message_to_worker(RpcDispatcherMessage::RegisterBestHeadListener(tx))?;
Ok(rx)
}
/// Get a stream of finalized relay chain headers
pub fn get_finalized_heads_stream(&self) -> Result<Receiver<RelayHeader>, RelayChainError> {
let (tx, rx) =
futures::channel::mpsc::channel::<RelayHeader>(NOTIFICATION_CHANNEL_SIZE_LIMIT);
self.send_register_message_to_worker(RpcDispatcherMessage::RegisterFinalizationListener(
tx,
))?;
Ok(rx)
}
/// Get a stream of all imported relay chain headers
pub fn get_imported_heads_stream(&self) -> Result<Receiver<RelayHeader>, RelayChainError> {
let (tx, rx) =
futures::channel::mpsc::channel::<RelayHeader>(NOTIFICATION_CHANNEL_SIZE_LIMIT);
self.send_register_message_to_worker(RpcDispatcherMessage::RegisterImportListener(tx))?;
Ok(rx)
}
fn send_register_message_to_worker(
&self,
message: RpcDispatcherMessage,
) -> Result<(), RelayChainError> {
self.to_worker_channel
.try_send(message)
.map_err(|e| RelayChainError::WorkerCommunicationError(e.to_string()))
}
}
/// Worker that should be used in combination with [`RelayChainRpcClient`]. Must be polled to distribute header notifications to listeners.
struct ReconnectingWebsocketWorker {
ws_urls: Vec<Url>,
/// Communication channel with the RPC client
client_receiver: TokioReceiver<RpcDispatcherMessage>,
/// Senders to distribute incoming header notifications to
imported_header_listeners: Vec<Sender<RelayHeader>>,
finalized_header_listeners: Vec<Sender<RelayHeader>>,
best_header_listeners: Vec<Sender<RelayHeader>>,
}
fn distribute_header(header: RelayHeader, senders: &mut Vec<Sender<RelayHeader>>) {
senders.retain_mut(|e| {
match e.try_send(header.clone()) {
// Receiver has been dropped, remove Sender from list.
Err(error) if error.is_disconnected() => false,
// Channel is full. This should not happen.
// TODO: Improve error handling here
// https://github.com/paritytech/cumulus/issues/1482
Err(error) => {
tracing::error!(target: LOG_TARGET, ?error, "Event distribution channel has reached its limit. This can lead to missed notifications.");
true
},
_ => true,
}
});
}
/// Manages the active websocket client.
/// Responsible for creating request futures, subscription streams
/// and reconnections.
#[derive(Debug)]
struct ClientManager {
urls: Vec<Url>,
active_client: Arc<JsonRpcClient>,
active_index: usize,
}
struct RelayChainSubscriptions {
import_subscription: Subscription<RelayHeader>,
finalized_subscription: Subscription<RelayHeader>,
best_subscription: Subscription<RelayHeader>,
}
/// Try to find a new RPC server to connect to.
async fn connect_next_available_rpc_server(
urls: &Vec<Url>,
starting_position: usize,
) -> Result<(usize, Arc<JsonRpcClient>), ()> {
tracing::debug!(target: LOG_TARGET, starting_position, "Connecting to RPC server.");
for (counter, url) in urls.iter().cycle().skip(starting_position).take(urls.len()).enumerate() {
let index = (starting_position + counter) % urls.len();
tracing::info!(
target: LOG_TARGET,
index,
?url,
"Trying to connect to next external relaychain node.",
);
if let Ok(ws_client) = WsClientBuilder::default().build(url).await {
return Ok((index, Arc::new(ws_client)))
};
}
Err(())
}
impl ClientManager {
pub async fn new(urls: Vec<Url>) -> Result<Self, ()> {
if urls.is_empty() {
return Err(())
}
let active_client = connect_next_available_rpc_server(&urls, 0).await?;
Ok(Self { urls, active_client: active_client.1, active_index: active_client.0 })
}
pub async fn connect_to_new_rpc_server(&mut self) -> Result<(), ()> {
let new_active =
connect_next_available_rpc_server(&self.urls, self.active_index + 1).await?;
self.active_client = new_active.1;
self.active_index = new_active.0;
Ok(())
}
async fn get_subscriptions(&self) -> Result<RelayChainSubscriptions, JsonRpseeError> {
let import_subscription = self
.active_client
.subscribe::<RelayHeader, _>(
"chain_subscribeAllHeads",
rpc_params![],
"chain_unsubscribeAllHeads",
)
.await
.map_err(|e| {
tracing::error!(
target: LOG_TARGET,
?e,
"Unable to open `chain_subscribeAllHeads` subscription."
);
e
})?;
let best_subscription = self
.active_client
.subscribe::<RelayHeader, _>(
"chain_subscribeNewHeads",
rpc_params![],
"chain_unsubscribeNewHeads",
)
.await
.map_err(|e| {
tracing::error!(
target: LOG_TARGET,
?e,
"Unable to open `chain_subscribeNewHeads` subscription."
);
e
})?;
let finalized_subscription = self
.active_client
.subscribe::<RelayHeader, _>(
"chain_subscribeFinalizedHeads",
rpc_params![],
"chain_unsubscribeFinalizedHeads",
)
.await
.map_err(|e| {
tracing::error!(
target: LOG_TARGET,
?e,
"Unable to open `chain_subscribeFinalizedHeads` subscription."
);
e
})?;
Ok(RelayChainSubscriptions {
import_subscription,
best_subscription,
finalized_subscription,
})
}
/// Create a request future that performs an RPC request and sends the results to the caller.
/// In case of a dead websocket connection, it returns the original request parameters to
/// enable retries.
fn create_request(
&self,
method: String,
params: ArrayParams,
response_sender: OneshotSender<Result<JsonValue, JsonRpseeError>>,
) -> BoxFuture<'static, Result<(), RpcDispatcherMessage>> {
let future_client = self.active_client.clone();
async move {
let resp = future_client.request(&method, params.clone()).await;
// We should only return the original request in case
// the websocket connection is dead and requires a restart.
// Other errors should be forwarded to the request caller.
if let Err(JsonRpseeError::RestartNeeded(_)) = resp {
return Err(RpcDispatcherMessage::Request(method, params, response_sender))
}
if let Err(err) = response_sender.send(resp) {
tracing::debug!(
target: LOG_TARGET,
?err,
"Recipient no longer interested in request result"
);
}
Ok(())
}
.boxed()
}
}
enum ConnectionStatus {
Connected,
ReconnectRequired(Option<RpcDispatcherMessage>),
}
impl ReconnectingWebsocketWorker {
/// Create new worker. Returns the worker and a channel to register new listeners.
async fn new(
urls: Vec<Url>,
) -> (ReconnectingWebsocketWorker, TokioSender<RpcDispatcherMessage>) {
let (tx, rx) = tokio_channel(100);
let worker = ReconnectingWebsocketWorker {
ws_urls: urls,
client_receiver: rx,
imported_header_listeners: Vec::new(),
finalized_header_listeners: Vec::new(),
best_header_listeners: Vec::new(),
};
(worker, tx)
}
/// Reconnect via [`ClientManager`] and provide new notification streams.
async fn handle_reconnect(
&mut self,
client_manager: &mut ClientManager,
pending_requests: &mut FuturesUnordered<
BoxFuture<'static, Result<(), RpcDispatcherMessage>>,
>,
first_failed_request: Option<RpcDispatcherMessage>,
) -> Result<RelayChainSubscriptions, String> {
let mut requests_to_retry = Vec::new();
if let Some(req @ RpcDispatcherMessage::Request(_, _, _)) = first_failed_request {
requests_to_retry.push(req);
}
// At this point, all pending requests will return an error since the
// websocket connection is dead. So draining the pending requests should be fast.
while !pending_requests.is_empty() {
if let Some(Err(req)) = pending_requests.next().await {
requests_to_retry.push(req);
}
}
if client_manager.connect_to_new_rpc_server().await.is_err() {
return Err(format!("Unable to find valid external RPC server, shutting down."))
};
for item in requests_to_retry.into_iter() {
if let RpcDispatcherMessage::Request(method, params, response_sender) = item {
pending_requests.push(client_manager.create_request(
method,
params,
response_sender,
));
};
}
client_manager.get_subscriptions().await.map_err(|e| {
format!("Not able to create streams from newly connected RPC server, shutting down. err: {:?}", e)
})
}
/// Run this worker to drive notification streams.
/// The worker does the following:
/// - Listen for [`RpcDispatcherMessage`], perform requests and register new listeners for the notification streams
/// - Distribute incoming import, best head and finalization notifications to registered listeners.
/// If an error occurs during sending, the receiver has been closed and we remove the sender from the list.
/// - Find a new valid RPC server to connect to in case the websocket connection is terminated.
/// If the worker is not able to connec to an RPC server from the list, the worker shuts down.
async fn run(mut self) {
let mut pending_requests = FuturesUnordered::new();
let urls = std::mem::take(&mut self.ws_urls);
let Ok(mut client_manager) = ClientManager::new(urls).await else {
tracing::error!(target: LOG_TARGET, "No valid RPC url found. Stopping RPC worker.");
return;
};
let Ok(mut subscriptions) = client_manager.get_subscriptions().await else {
tracing::error!(target: LOG_TARGET, "Unable to fetch subscriptions on initial connection.");
return;
};
let mut imported_blocks_cache =
LruCache::new(NonZeroUsize::new(40).expect("40 is nonzero; qed."));
let mut should_reconnect = ConnectionStatus::Connected;
let mut last_seen_finalized_num: RelayBlockNumber = 0;
loop {
// This branch is taken if the websocket connection to the current RPC server is closed.
if let ConnectionStatus::ReconnectRequired(maybe_failed_request) = should_reconnect {
match self
.handle_reconnect(
&mut client_manager,
&mut pending_requests,
maybe_failed_request,
)
.await
{
Ok(new_subscriptions) => {
subscriptions = new_subscriptions;
},
Err(message) => {
tracing::error!(
target: LOG_TARGET,
message,
"Unable to reconnect, stopping worker."
);
return
},
}
should_reconnect = ConnectionStatus::Connected;
}
tokio::select! {
evt = self.client_receiver.recv() => match evt {
Some(RpcDispatcherMessage::RegisterBestHeadListener(tx)) => {
self.best_header_listeners.push(tx);
},
Some(RpcDispatcherMessage::RegisterImportListener(tx)) => {
self.imported_header_listeners.push(tx)
},
Some(RpcDispatcherMessage::RegisterFinalizationListener(tx)) => {
self.finalized_header_listeners.push(tx)
},
Some(RpcDispatcherMessage::Request(method, params, response_sender)) => {
pending_requests.push(client_manager.create_request(method, params, response_sender));
},
None => {
tracing::error!(target: LOG_TARGET, "RPC client receiver closed. Stopping RPC Worker.");
return;
}
},
should_retry = pending_requests.next(), if !pending_requests.is_empty() => {
if let Some(Err(req)) = should_retry {
should_reconnect = ConnectionStatus::ReconnectRequired(Some(req));
}
},
import_event = subscriptions.import_subscription.next() => {
match import_event {
Some(Ok(header)) => {
let hash = header.hash();
if imported_blocks_cache.contains(&hash) {
tracing::debug!(
target: LOG_TARGET,
number = header.number,
?hash,
"Duplicate imported block header. This might happen after switching to a new RPC node. Skipping distribution."
);
continue;
}
imported_blocks_cache.put(hash, ());
distribute_header(header, &mut self.imported_header_listeners);
},
None => {
tracing::error!(target: LOG_TARGET, "Subscription closed.");
should_reconnect = ConnectionStatus::ReconnectRequired(None);
},
Some(Err(error)) => {
tracing::error!(target: LOG_TARGET, ?error, "Error in RPC subscription.");
should_reconnect = ConnectionStatus::ReconnectRequired(None);
},
}
},
best_header_event = subscriptions.best_subscription.next() => {
match best_header_event {
Some(Ok(header)) => distribute_header(header, &mut self.best_header_listeners),
None => {
tracing::error!(target: LOG_TARGET, "Subscription closed.");
should_reconnect = ConnectionStatus::ReconnectRequired(None);
},
Some(Err(error)) => {
tracing::error!(target: LOG_TARGET, ?error, "Error in RPC subscription.");
should_reconnect = ConnectionStatus::ReconnectRequired(None);
},
}
}
finalized_event = subscriptions.finalized_subscription.next() => {
match finalized_event {
Some(Ok(header)) if header.number > last_seen_finalized_num => {
last_seen_finalized_num = header.number;
distribute_header(header, &mut self.finalized_header_listeners);
},
Some(Ok(header)) => {
tracing::debug!(
target: LOG_TARGET,
number = header.number,
last_seen_finalized_num,
"Duplicate finalized block header. This might happen after switching to a new RPC node. Skipping distribution."
);
},
None => {
tracing::error!(target: LOG_TARGET, "Subscription closed.");
should_reconnect = ConnectionStatus::ReconnectRequired(None);
},
Some(Err(error)) => {
tracing::error!(target: LOG_TARGET, ?error, "Error in RPC subscription.");
should_reconnect = ConnectionStatus::ReconnectRequired(None);
},
}
}
}
}
}
}
@@ -14,7 +14,7 @@
// You should have received a copy of the GNU General Public License
// along with Cumulus. If not, see <http://www.gnu.org/licenses/>.
use backoff::{future::retry_notify, ExponentialBackoff};
use crate::reconnecting_ws_client::ReconnectingWsClient;
use cumulus_primitives_core::{
relay_chain::{
v2::{
@@ -23,211 +23,49 @@ use cumulus_primitives_core::{
PvfCheckStatement, ScrapedOnChainVotes, SessionIndex, SessionInfo, ValidationCode,
ValidationCodeHash, ValidatorId, ValidatorIndex, ValidatorSignature,
},
CandidateHash, Hash as PHash, Header as PHeader, InboundHrmpMessage,
CandidateHash, Hash as RelayHash, Header as RelayHeader, InboundHrmpMessage,
},
InboundDownwardMessage, ParaId, PersistedValidationData,
};
use cumulus_relay_chain_interface::{RelayChainError, RelayChainResult};
use futures::{
channel::mpsc::{Receiver, Sender},
StreamExt,
};
use jsonrpsee::{
core::{
client::{Client as JsonRpcClient, ClientT, Subscription, SubscriptionClientT},
params::ArrayParams,
Error as JsonRpseeError,
},
rpc_params,
ws_client::WsClientBuilder,
};
use futures::channel::mpsc::Receiver;
use jsonrpsee::{core::params::ArrayParams, rpc_params};
use parity_scale_codec::{Decode, Encode};
use polkadot_service::{BlockNumber, TaskManager};
use sc_client_api::StorageData;
use sc_rpc_api::{state::ReadProof, system::Health};
use serde::de::DeserializeOwned;
use sp_api::RuntimeVersion;
use sp_consensus_babe::Epoch;
use sp_core::sp_std::collections::btree_map::BTreeMap;
use sp_runtime::DeserializeOwned;
use sp_storage::StorageKey;
use std::sync::Arc;
use tokio::sync::mpsc::{
channel as tokio_channel, Receiver as TokioReceiver, Sender as TokioSender,
};
pub use url::Url;
const LOG_TARGET: &str = "relay-chain-rpc-client";
const NOTIFICATION_CHANNEL_SIZE_LIMIT: usize = 20;
/// Client that maps RPC methods and deserializes results
#[derive(Clone)]
pub struct RelayChainRpcClient {
/// Websocket client to make calls
ws_client: Arc<JsonRpcClient>,
/// Retry strategy that should be used for requests and subscriptions
retry_strategy: ExponentialBackoff,
/// Channel to communicate with the RPC worker
to_worker_channel: TokioSender<NotificationRegisterMessage>,
}
/// Worker messages to register new notification listeners
#[derive(Clone, Debug)]
pub enum NotificationRegisterMessage {
RegisterBestHeadListener(Sender<PHeader>),
RegisterImportListener(Sender<PHeader>),
RegisterFinalizationListener(Sender<PHeader>),
}
/// Worker that should be used in combination with [`RelayChainRpcClient`]. Must be polled to distribute header notifications to listeners.
struct RpcStreamWorker {
// Communication channel with the RPC client
client_receiver: TokioReceiver<NotificationRegisterMessage>,
// Senders to distribute incoming header notifications to
imported_header_listeners: Vec<Sender<PHeader>>,
finalized_header_listeners: Vec<Sender<PHeader>>,
best_header_listeners: Vec<Sender<PHeader>>,
// Incoming notification subscriptions
rpc_imported_header_subscription: Subscription<PHeader>,
rpc_finalized_header_subscription: Subscription<PHeader>,
rpc_best_header_subscription: Subscription<PHeader>,
ws_client: ReconnectingWsClient,
}
/// Entry point to create [`RelayChainRpcClient`] and start a worker that distributes notifications.
pub async fn create_client_and_start_worker(
url: Url,
urls: Vec<Url>,
task_manager: &mut TaskManager,
) -> RelayChainResult<RelayChainRpcClient> {
tracing::info!(target: LOG_TARGET, url = %url.to_string(), "Initializing RPC Client");
let ws_client = WsClientBuilder::default().build(url.as_str()).await?;
let ws_client = ReconnectingWsClient::new(urls, task_manager).await?;
let best_head_stream = RelayChainRpcClient::subscribe_new_best_heads(&ws_client).await?;
let finalized_head_stream = RelayChainRpcClient::subscribe_finalized_heads(&ws_client).await?;
let imported_head_stream = RelayChainRpcClient::subscribe_imported_heads(&ws_client).await?;
let (worker, sender) =
RpcStreamWorker::new(imported_head_stream, best_head_stream, finalized_head_stream);
let client = RelayChainRpcClient::new(ws_client, sender).await?;
task_manager
.spawn_essential_handle()
.spawn("relay-chain-rpc-worker", None, worker.run());
let client = RelayChainRpcClient::new(ws_client).await?;
Ok(client)
}
fn handle_event_distribution(
event: Option<Result<PHeader, JsonRpseeError>>,
senders: &mut Vec<Sender<PHeader>>,
) -> Result<(), String> {
match event {
Some(Ok(header)) => {
senders.retain_mut(|e| {
match e.try_send(header.clone()) {
// Receiver has been dropped, remove Sender from list.
Err(error) if error.is_disconnected() => false,
// Channel is full. This should not happen.
// TODO: Improve error handling here
// https://github.com/paritytech/cumulus/issues/1482
Err(error) => {
tracing::error!(target: LOG_TARGET, ?error, "Event distribution channel has reached its limit. This can lead to missed notifications.");
true
},
_ => true,
}
});
Ok(())
},
None => Err("RPC Subscription closed.".to_string()),
Some(Err(err)) => Err(format!("Error in RPC subscription: {}", err)),
}
}
impl RpcStreamWorker {
/// Create new worker. Returns the worker and a channel to register new listeners.
fn new(
import_sub: Subscription<PHeader>,
best_sub: Subscription<PHeader>,
finalized_sub: Subscription<PHeader>,
) -> (RpcStreamWorker, TokioSender<NotificationRegisterMessage>) {
let (tx, rx) = tokio_channel(100);
let worker = RpcStreamWorker {
client_receiver: rx,
imported_header_listeners: Vec::new(),
finalized_header_listeners: Vec::new(),
best_header_listeners: Vec::new(),
rpc_imported_header_subscription: import_sub,
rpc_best_header_subscription: best_sub,
rpc_finalized_header_subscription: finalized_sub,
};
(worker, tx)
}
/// Run this worker to drive notification streams.
/// The worker does two things:
/// 1. Listen for `NotificationRegisterMessage` and register new listeners for the notification streams
/// 2. Distribute incoming import, best head and finalization notifications to registered listeners.
/// If an error occurs during sending, the receiver has been closed and we remove the sender from the list.
pub async fn run(mut self) {
let mut import_sub = self.rpc_imported_header_subscription.fuse();
let mut best_head_sub = self.rpc_best_header_subscription.fuse();
let mut finalized_sub = self.rpc_finalized_header_subscription.fuse();
loop {
tokio::select! {
evt = self.client_receiver.recv() => match evt {
Some(NotificationRegisterMessage::RegisterBestHeadListener(tx)) => {
self.best_header_listeners.push(tx);
},
Some(NotificationRegisterMessage::RegisterImportListener(tx)) => {
self.imported_header_listeners.push(tx)
},
Some(NotificationRegisterMessage::RegisterFinalizationListener(tx)) => {
self.finalized_header_listeners.push(tx)
},
None => {
tracing::error!(target: LOG_TARGET, "RPC client receiver closed. Stopping RPC Worker.");
return;
}
},
import_event = import_sub.next() => {
if let Err(err) = handle_event_distribution(import_event, &mut self.imported_header_listeners) {
tracing::error!(target: LOG_TARGET, err, "Encountered error while processing imported header notification. Stopping RPC Worker.");
return;
}
},
best_header_event = best_head_sub.next() => {
if let Err(err) = handle_event_distribution(best_header_event, &mut self.best_header_listeners) {
tracing::error!(target: LOG_TARGET, err, "Encountered error while processing best header notification. Stopping RPC Worker.");
return;
}
}
finalized_event = finalized_sub.next() => {
if let Err(err) = handle_event_distribution(finalized_event, &mut self.finalized_header_listeners) {
tracing::error!(target: LOG_TARGET, err, "Encountered error while processing finalized header notification. Stopping RPC Worker.");
return;
}
}
}
}
}
}
impl RelayChainRpcClient {
/// Initialize new RPC Client.
async fn new(
ws_client: JsonRpcClient,
sender: TokioSender<NotificationRegisterMessage>,
) -> RelayChainResult<Self> {
let client = RelayChainRpcClient {
to_worker_channel: sender,
ws_client: Arc::new(ws_client),
retry_strategy: ExponentialBackoff::default(),
};
async fn new(ws_client: ReconnectingWsClient) -> RelayChainResult<Self> {
let client = RelayChainRpcClient { ws_client };
Ok(client)
}
@@ -236,7 +74,7 @@ impl RelayChainRpcClient {
pub async fn call_remote_runtime_function<R: Decode>(
&self,
method_name: &str,
hash: PHash,
hash: RelayHash,
payload: Option<impl Encode>,
) -> RelayChainResult<R> {
let payload_bytes =
@@ -286,34 +124,23 @@ impl RelayChainRpcClient {
) -> Result<R, RelayChainError>
where
R: DeserializeOwned + std::fmt::Debug,
OR: Fn(&jsonrpsee::core::Error),
OR: Fn(&RelayChainError),
{
retry_notify(
self.retry_strategy.clone(),
|| async {
self.ws_client.request(method, params.clone()).await.map_err(|err| match err {
JsonRpseeError::Transport(_) =>
backoff::Error::Transient { err, retry_after: None },
_ => backoff::Error::Permanent(err),
})
},
|error, dur| tracing::trace!(target: LOG_TARGET, %error, ?dur, "Encountered transport error, retrying."),
)
.await
.map_err(|err| {
self.ws_client.request(method, params).await.map_err(|err| {
trace_error(&err);
RelayChainError::RpcCallError(method.to_string(), err)})
RelayChainError::RpcCallError(method.to_string())
})
}
/// Returns information regarding the current epoch.
pub async fn babe_api_current_epoch(&self, at: PHash) -> Result<Epoch, RelayChainError> {
pub async fn babe_api_current_epoch(&self, at: RelayHash) -> Result<Epoch, RelayChainError> {
self.call_remote_runtime_function("BabeApi_current_epoch", at, None::<()>).await
}
/// Old method to fetch v1 session info.
pub async fn parachain_host_session_info_before_version_2(
&self,
at: PHash,
at: RelayHash,
index: SessionIndex,
) -> Result<Option<OldV1SessionInfo>, RelayChainError> {
self.call_remote_runtime_function(
@@ -327,8 +154,8 @@ impl RelayChainRpcClient {
/// Scrape dispute relevant from on-chain, backing votes and resolved disputes.
pub async fn parachain_host_on_chain_votes(
&self,
at: PHash,
) -> Result<Option<ScrapedOnChainVotes<PHash>>, RelayChainError> {
at: RelayHash,
) -> Result<Option<ScrapedOnChainVotes<RelayHash>>, RelayChainError> {
self.call_remote_runtime_function("ParachainHost_on_chain_votes", at, None::<()>)
.await
}
@@ -336,7 +163,7 @@ impl RelayChainRpcClient {
/// Returns code hashes of PVFs that require pre-checking by validators in the active set.
pub async fn parachain_host_pvfs_require_precheck(
&self,
at: PHash,
at: RelayHash,
) -> Result<Vec<ValidationCodeHash>, RelayChainError> {
self.call_remote_runtime_function("ParachainHost_pvfs_require_precheck", at, None::<()>)
.await
@@ -345,7 +172,7 @@ impl RelayChainRpcClient {
/// Submits a PVF pre-checking statement into the transaction pool.
pub async fn parachain_host_submit_pvf_check_statement(
&self,
at: PHash,
at: RelayHash,
stmt: PvfCheckStatement,
signature: ValidatorSignature,
) -> Result<(), RelayChainError> {
@@ -371,9 +198,9 @@ impl RelayChainRpcClient {
pub async fn state_get_read_proof(
&self,
storage_keys: Vec<StorageKey>,
at: Option<PHash>,
) -> Result<ReadProof<PHash>, RelayChainError> {
let params = rpc_params!(storage_keys, at);
at: Option<RelayHash>,
) -> Result<ReadProof<RelayHash>, RelayChainError> {
let params = rpc_params![storage_keys, at];
self.request("state_getReadProof", params).await
}
@@ -381,17 +208,17 @@ impl RelayChainRpcClient {
pub async fn state_get_storage(
&self,
storage_key: StorageKey,
at: Option<PHash>,
at: Option<RelayHash>,
) -> Result<Option<StorageData>, RelayChainError> {
let params = rpc_params!(storage_key, at);
let params = rpc_params![storage_key, at];
self.request("state_getStorage", params).await
}
/// Get hash of the n-th block in the canon chain.
///
/// By default returns latest block hash.
pub async fn chain_get_head(&self, at: Option<u64>) -> Result<PHash, RelayChainError> {
let params = rpc_params!(at);
pub async fn chain_get_head(&self, at: Option<u64>) -> Result<RelayHash, RelayChainError> {
let params = rpc_params![at];
self.request("chain_getHead", params).await
}
@@ -400,7 +227,7 @@ impl RelayChainRpcClient {
/// should be the successor of the number of the block.
pub async fn parachain_host_validator_groups(
&self,
at: PHash,
at: RelayHash,
) -> Result<(Vec<Vec<ValidatorIndex>>, GroupRotationInfo), RelayChainError> {
self.call_remote_runtime_function("ParachainHost_validator_groups", at, None::<()>)
.await
@@ -409,7 +236,7 @@ impl RelayChainRpcClient {
/// Get a vector of events concerning candidates that occurred within a block.
pub async fn parachain_host_candidate_events(
&self,
at: PHash,
at: RelayHash,
) -> Result<Vec<CandidateEvent>, RelayChainError> {
self.call_remote_runtime_function("ParachainHost_candidate_events", at, None::<()>)
.await
@@ -418,7 +245,7 @@ impl RelayChainRpcClient {
/// Checks if the given validation outputs pass the acceptance criteria.
pub async fn parachain_host_check_validation_outputs(
&self,
at: PHash,
at: RelayHash,
para_id: ParaId,
outputs: CandidateCommitments,
) -> Result<bool, RelayChainError> {
@@ -435,9 +262,9 @@ impl RelayChainRpcClient {
/// data hash against an expected one and yields `None` if they're not equal.
pub async fn parachain_host_assumed_validation_data(
&self,
at: PHash,
at: RelayHash,
para_id: ParaId,
expected_hash: PHash,
expected_hash: RelayHash,
) -> Result<Option<(PersistedValidationData, ValidationCodeHash)>, RelayChainError> {
self.call_remote_runtime_function(
"ParachainHost_persisted_assumed_validation_data",
@@ -448,7 +275,7 @@ impl RelayChainRpcClient {
}
/// Get hash of last finalized block.
pub async fn chain_get_finalized_head(&self) -> Result<PHash, RelayChainError> {
pub async fn chain_get_finalized_head(&self) -> Result<RelayHash, RelayChainError> {
self.request("chain_getFinalizedHead", rpc_params![]).await
}
@@ -456,8 +283,8 @@ impl RelayChainRpcClient {
pub async fn chain_get_block_hash(
&self,
block_number: Option<polkadot_service::BlockNumber>,
) -> Result<Option<PHash>, RelayChainError> {
let params = rpc_params!(block_number);
) -> Result<Option<RelayHash>, RelayChainError> {
let params = rpc_params![block_number];
self.request("chain_getBlockHash", params).await
}
@@ -468,7 +295,7 @@ impl RelayChainRpcClient {
/// and the para already occupies a core.
pub async fn parachain_host_persisted_validation_data(
&self,
at: PHash,
at: RelayHash,
para_id: ParaId,
occupied_core_assumption: OccupiedCoreAssumption,
) -> Result<Option<PersistedValidationData>, RelayChainError> {
@@ -483,7 +310,7 @@ impl RelayChainRpcClient {
/// Get the validation code from its hash.
pub async fn parachain_host_validation_code_by_hash(
&self,
at: PHash,
at: RelayHash,
validation_code_hash: ValidationCodeHash,
) -> Result<Option<ValidationCode>, RelayChainError> {
self.call_remote_runtime_function(
@@ -498,15 +325,15 @@ impl RelayChainRpcClient {
/// Cores are either free or occupied. Free cores can have paras assigned to them.
pub async fn parachain_host_availability_cores(
&self,
at: PHash,
) -> Result<Vec<CoreState<PHash, BlockNumber>>, RelayChainError> {
at: RelayHash,
) -> Result<Vec<CoreState<RelayHash, BlockNumber>>, RelayChainError> {
self.call_remote_runtime_function("ParachainHost_availability_cores", at, None::<()>)
.await
}
/// Get runtime version
pub async fn runtime_version(&self, at: PHash) -> Result<RuntimeVersion, RelayChainError> {
let params = rpc_params!(at);
pub async fn runtime_version(&self, at: RelayHash) -> Result<RuntimeVersion, RelayChainError> {
let params = rpc_params![at];
self.request("state_getRuntimeVersion", params).await
}
@@ -514,7 +341,7 @@ impl RelayChainRpcClient {
/// This is a staging method! Do not use on production runtimes!
pub async fn parachain_host_staging_get_disputes(
&self,
at: PHash,
at: RelayHash,
) -> Result<Vec<(SessionIndex, CandidateHash, DisputeState<BlockNumber>)>, RelayChainError> {
self.call_remote_runtime_function("ParachainHost_staging_get_disputes", at, None::<()>)
.await
@@ -522,7 +349,7 @@ impl RelayChainRpcClient {
pub async fn authority_discovery_authorities(
&self,
at: PHash,
at: RelayHash,
) -> Result<Vec<sp_authority_discovery::AuthorityId>, RelayChainError> {
self.call_remote_runtime_function("AuthorityDiscoveryApi_authorities", at, None::<()>)
.await
@@ -534,7 +361,7 @@ impl RelayChainRpcClient {
/// and the para already occupies a core.
pub async fn parachain_host_validation_code(
&self,
at: PHash,
at: RelayHash,
para_id: ParaId,
occupied_core_assumption: OccupiedCoreAssumption,
) -> Result<Option<ValidationCode>, RelayChainError> {
@@ -549,7 +376,7 @@ impl RelayChainRpcClient {
/// Fetch the hash of the validation code used by a para, making the given `OccupiedCoreAssumption`.
pub async fn parachain_host_validation_code_hash(
&self,
at: PHash,
at: RelayHash,
para_id: ParaId,
occupied_core_assumption: OccupiedCoreAssumption,
) -> Result<Option<ValidationCodeHash>, RelayChainError> {
@@ -564,7 +391,7 @@ impl RelayChainRpcClient {
/// Get the session info for the given session, if stored.
pub async fn parachain_host_session_info(
&self,
at: PHash,
at: RelayHash,
index: SessionIndex,
) -> Result<Option<SessionInfo>, RelayChainError> {
self.call_remote_runtime_function("ParachainHost_session_info", at, Some(index))
@@ -574,9 +401,9 @@ impl RelayChainRpcClient {
/// Get header at specified hash
pub async fn chain_get_header(
&self,
hash: Option<PHash>,
) -> Result<Option<PHeader>, RelayChainError> {
let params = rpc_params!(hash);
hash: Option<RelayHash>,
) -> Result<Option<RelayHeader>, RelayChainError> {
let params = rpc_params![hash];
self.request("chain_getHeader", params).await
}
@@ -584,7 +411,7 @@ impl RelayChainRpcClient {
/// assigned to occupied cores in `availability_cores` and `None` otherwise.
pub async fn parachain_host_candidate_pending_availability(
&self,
at: PHash,
at: RelayHash,
para_id: ParaId,
) -> Result<Option<CommittedCandidateReceipt>, RelayChainError> {
self.call_remote_runtime_function(
@@ -600,7 +427,7 @@ impl RelayChainRpcClient {
/// This can be used to instantiate a `SigningContext`.
pub async fn parachain_host_session_index_for_child(
&self,
at: PHash,
at: RelayHash,
) -> Result<SessionIndex, RelayChainError> {
self.call_remote_runtime_function("ParachainHost_session_index_for_child", at, None::<()>)
.await
@@ -609,7 +436,7 @@ impl RelayChainRpcClient {
/// Get the current validators.
pub async fn parachain_host_validators(
&self,
at: PHash,
at: RelayHash,
) -> Result<Vec<ValidatorId>, RelayChainError> {
self.call_remote_runtime_function("ParachainHost_validators", at, None::<()>)
.await
@@ -620,7 +447,7 @@ impl RelayChainRpcClient {
pub async fn parachain_host_inbound_hrmp_channels_contents(
&self,
para_id: ParaId,
at: PHash,
at: RelayHash,
) -> Result<BTreeMap<ParaId, Vec<InboundHrmpMessage>>, RelayChainError> {
self.call_remote_runtime_function(
"ParachainHost_inbound_hrmp_channels_contents",
@@ -634,81 +461,24 @@ impl RelayChainRpcClient {
pub async fn parachain_host_dmq_contents(
&self,
para_id: ParaId,
at: PHash,
at: RelayHash,
) -> Result<Vec<InboundDownwardMessage>, RelayChainError> {
self.call_remote_runtime_function("ParachainHost_dmq_contents", at, Some(para_id))
.await
}
/// Get a stream of all imported relay chain headers
pub async fn get_imported_heads_stream(&self) -> Result<Receiver<PHeader>, RelayChainError> {
let (tx, rx) = futures::channel::mpsc::channel::<PHeader>(NOTIFICATION_CHANNEL_SIZE_LIMIT);
self.send_register_message_to_worker(NotificationRegisterMessage::RegisterImportListener(
tx,
))?;
Ok(rx)
pub fn get_imported_heads_stream(&self) -> Result<Receiver<RelayHeader>, RelayChainError> {
self.ws_client.get_imported_heads_stream()
}
/// Get a stream of new best relay chain headers
pub async fn get_best_heads_stream(&self) -> Result<Receiver<PHeader>, RelayChainError> {
let (tx, rx) = futures::channel::mpsc::channel::<PHeader>(NOTIFICATION_CHANNEL_SIZE_LIMIT);
self.send_register_message_to_worker(
NotificationRegisterMessage::RegisterBestHeadListener(tx),
)?;
Ok(rx)
pub fn get_best_heads_stream(&self) -> Result<Receiver<RelayHeader>, RelayChainError> {
self.ws_client.get_best_heads_stream()
}
/// Get a stream of finalized relay chain headers
pub async fn get_finalized_heads_stream(&self) -> Result<Receiver<PHeader>, RelayChainError> {
let (tx, rx) = futures::channel::mpsc::channel::<PHeader>(NOTIFICATION_CHANNEL_SIZE_LIMIT);
self.send_register_message_to_worker(
NotificationRegisterMessage::RegisterFinalizationListener(tx),
)?;
Ok(rx)
}
fn send_register_message_to_worker(
&self,
message: NotificationRegisterMessage,
) -> Result<(), RelayChainError> {
self.to_worker_channel
.try_send(message)
.map_err(|e| RelayChainError::WorkerCommunicationError(e.to_string()))
}
async fn subscribe_imported_heads(
ws_client: &JsonRpcClient,
) -> Result<Subscription<PHeader>, RelayChainError> {
Ok(ws_client
.subscribe::<PHeader, _>(
"chain_subscribeAllHeads",
rpc_params![],
"chain_unsubscribeAllHeads",
)
.await?)
}
async fn subscribe_finalized_heads(
ws_client: &JsonRpcClient,
) -> Result<Subscription<PHeader>, RelayChainError> {
Ok(ws_client
.subscribe::<PHeader, _>(
"chain_subscribeFinalizedHeads",
rpc_params![],
"chain_unsubscribeFinalizedHeads",
)
.await?)
}
async fn subscribe_new_best_heads(
ws_client: &JsonRpcClient,
) -> Result<Subscription<PHeader>, RelayChainError> {
Ok(ws_client
.subscribe::<PHeader, _>(
"chain_subscribeNewHeads",
rpc_params![],
"chain_unsubscribeFinalizedHeads",
)
.await?)
pub fn get_finalized_heads_stream(&self) -> Result<Receiver<RelayHeader>, RelayChainError> {
self.ws_client.get_finalized_heads_stream()
}
}