stricter ordering. refs #210

This commit is contained in:
Bryan Stitt 2023-11-06 10:06:32 -08:00
parent 0b35cbce3b
commit a4400bbd40
15 changed files with 77 additions and 627 deletions

@ -49,14 +49,17 @@ where
/// filter duplicates and send the rest to any subscribers
/// TODO: change this to be `send` and put a moka cache here instead of lru. then the de-dupe load will be spread across senders
pub async fn send(&self, item: T) {
self.total_unfiltered.fetch_add(1, Ordering::Relaxed);
// this is just a debug counter so Relaxed is probably fine
self.total_unfiltered.fetch_add(1, Ordering::AcqRel);
self.cache
.get_with(item.clone(), async {
self.total_filtered.fetch_add(1, Ordering::Relaxed);
// this is just a debug counter so Relaxed is probably fine
self.total_filtered.fetch_add(1, Ordering::AcqRel);
if let Ok(x) = self.broadcast_filtered_tx.send(item) {
self.total_broadcasts.fetch_add(x, Ordering::Relaxed);
// this is just a debug counter so Relaxed is probably fine
self.total_broadcasts.fetch_add(x, Ordering::AcqRel);
}
})
.await;
@ -75,15 +78,15 @@ where
f.debug_struct("DedupedBroadcaster")
.field(
"total_unfiltered",
&self.total_unfiltered.load(Ordering::Relaxed),
&self.total_unfiltered.load(Ordering::Acquire),
)
.field(
"total_filtered",
&self.total_filtered.load(Ordering::Relaxed),
&self.total_filtered.load(Ordering::Acquire),
)
.field(
"total_broadcasts",
&self.total_broadcasts.load(Ordering::Relaxed),
&self.total_broadcasts.load(Ordering::Acquire),
)
.field(
"subscriptions",
@ -105,15 +108,15 @@ where
state.serialize_field(
"total_unfiltered",
&self.total_unfiltered.load(Ordering::Relaxed),
&self.total_unfiltered.load(Ordering::Acquire),
)?;
state.serialize_field(
"total_filtered",
&self.total_filtered.load(Ordering::Relaxed),
&self.total_filtered.load(Ordering::Acquire),
)?;
state.serialize_field(
"total_broadcasts",
&self.total_broadcasts.load(Ordering::Relaxed),
&self.total_broadcasts.load(Ordering::Acquire),
)?;
state.serialize_field(
"subscriptions",
@ -153,8 +156,8 @@ mod tests {
yield_now().await;
assert_eq!(broadcaster.total_unfiltered.load(Ordering::Relaxed), 7);
assert_eq!(broadcaster.total_filtered.load(Ordering::Relaxed), 3);
assert_eq!(broadcaster.total_broadcasts.load(Ordering::Relaxed), 6);
assert_eq!(broadcaster.total_unfiltered.load(Ordering::Acquire), 7);
assert_eq!(broadcaster.total_filtered.load(Ordering::Acquire), 3);
assert_eq!(broadcaster.total_broadcasts.load(Ordering::Acquire), 6);
}
}

@ -52,6 +52,7 @@ impl RollingQuantileLatencyTask {
}
/// Update the estimate object atomically.
/// This is relaxed because we care more about speed than accuracy.
fn update(&mut self, rtt: f32) {
self.quantile.update(rtt);

@ -765,7 +765,7 @@ impl App {
// TODO: what interval? i don't think we use it
// i tried and failed to `impl JsonRpcClient for Web3ProxyApi`
// i tried and failed to set up ipc. http is already running, so lets just use that
let frontend_port = self.frontend_port.load(Ordering::Relaxed);
let frontend_port = self.frontend_port.load(Ordering::SeqCst);
if frontend_port == 0 {
panic!("frontend is not running. cannot create provider yet");
@ -1305,22 +1305,22 @@ impl App {
let (code, response) = match last_response {
Ok(response_data) => {
web3_request.error_response.store(false, Ordering::Relaxed);
web3_request.error_response.store(false, Ordering::Release);
// TODO: is it true that all jsonrpc errors are user errors?
web3_request
.user_error_response
.store(response_data.is_jsonrpc_err(), Ordering::Relaxed);
.store(response_data.is_jsonrpc_err(), Ordering::Release);
(StatusCode::OK, response_data)
}
Err(err) => {
// max tries exceeded. return the error
web3_request.error_response.store(true, Ordering::Relaxed);
web3_request.error_response.store(true, Ordering::Release);
web3_request
.user_error_response
.store(false, Ordering::Relaxed);
.store(false, Ordering::Release);
err.as_json_response_parts(web3_request.id())
}
@ -1523,7 +1523,7 @@ impl App {
// TODO: only charge for archive if it gave a result
web3_request
.archive_request
.store(true, atomic::Ordering::Relaxed);
.store(true, atomic::Ordering::Release);
// TODO: we don't actually want try_send_all. we want the first non-null, non-error response
self

@ -54,6 +54,7 @@ pub enum RateLimitResult {
#[derive(Copy, Clone, Debug, Hash, Eq, PartialEq)]
pub enum AuthorizationType {
/// TODO: sometimes localhost should be internal and other times it should be Frontend. make a better separatation
Internal,
Frontend,
}

@ -320,13 +320,13 @@ pub async fn serve(
axum::Server::from_tcp(listener)?
} else {
// TODO: allow only listening on localhost? top_config.app.host.parse()?
let addr = SocketAddr::from(([0, 0, 0, 0], app.frontend_port.load(Ordering::Relaxed)));
let addr = SocketAddr::from(([0, 0, 0, 0], app.frontend_port.load(Ordering::SeqCst)));
axum::Server::try_bind(&addr)?
};
#[cfg(not(feature = "listenfd"))]
let server_builder = {
let addr = SocketAddr::from(([0, 0, 0, 0], app.frontend_port.load(Ordering::Relaxed)));
let addr = SocketAddr::from(([0, 0, 0, 0], app.frontend_port.load(Ordering::SeqCst)));
axum::Server::try_bind(&addr)?
};
@ -358,7 +358,7 @@ pub async fn serve(
let port = server.local_addr().port();
info!("listening on port {}", port);
app.frontend_port.store(port, Ordering::Relaxed);
app.frontend_port.store(port, Ordering::SeqCst);
let server = server
// TODO: option to use with_connect_info. we want it in dev, but not when running behind a proxy, but not

@ -180,7 +180,7 @@ impl RequestBuilder {
if let Ok(x) = &x {
if self.archive_request {
x.archive_request.store(true, atomic::Ordering::Relaxed);
x.archive_request.store(true, atomic::Ordering::Release);
}
}
@ -279,7 +279,7 @@ impl Serialize for ValidatedRequest {
state.serialize_field(
"archive_request",
&self.archive_request.load(atomic::Ordering::Relaxed),
&self.archive_request.load(atomic::Ordering::Acquire),
)?;
state.serialize_field("chain_id", &self.chain_id)?;
@ -302,7 +302,7 @@ impl Serialize for ValidatedRequest {
state.serialize_field(
"response_bytes",
&self.response_bytes.load(atomic::Ordering::Relaxed),
&self.response_bytes.load(atomic::Ordering::Acquire),
)?;
state.end()
@ -510,7 +510,7 @@ impl ValidatedRequest {
#[inline]
pub fn min_block_needed(&self) -> Option<U64> {
if self.archive_request.load(atomic::Ordering::Relaxed) {
if self.archive_request.load(atomic::Ordering::Acquire) {
Some(U64::zero())
} else {
self.cache_mode.from_block().map(|x| x.num())
@ -564,16 +564,16 @@ impl ValidatedRequest {
let num_bytes = response.num_bytes();
self.response_bytes
.fetch_add(num_bytes, atomic::Ordering::Relaxed);
.fetch_add(num_bytes, atomic::Ordering::AcqRel);
self.response_millis.fetch_add(
self.start_instant.elapsed().as_millis() as u64,
atomic::Ordering::Relaxed,
atomic::Ordering::AcqRel,
);
// TODO: record first or last timestamp? really, we need multiple
self.response_timestamp
.store(Utc::now().timestamp(), atomic::Ordering::Relaxed);
.store(Utc::now().timestamp(), atomic::Ordering::Release);
// TODO: set user_error_response and error_response here instead of outside this function

@ -136,7 +136,8 @@ impl KafkaDebugLogger {
let payload =
serde_json::to_vec(&request).expect("requests should always serialize with rmp");
self.num_requests.fetch_add(1, atomic::Ordering::Relaxed);
// this is just a debug counter so Relaxed is probably fine
self.num_requests.fetch_add(1, atomic::Ordering::AcqRel);
self.background_log(payload)
}
@ -148,7 +149,8 @@ impl KafkaDebugLogger {
let payload =
serde_json::to_vec(&response).expect("requests should always serialize with rmp");
self.num_responses.fetch_add(1, atomic::Ordering::Relaxed);
// this is just a debug counter so Relaxed is probably fine
self.num_responses.fetch_add(1, atomic::Ordering::AcqRel);
self.background_log(payload)
}

@ -22,7 +22,7 @@ pub async fn serve(
.layer(Extension(app.clone()));
// note: the port here might be 0
let port = app.prometheus_port.load(Ordering::Relaxed);
let port = app.prometheus_port.load(Ordering::SeqCst);
// TODO: config for the host?
let addr = SocketAddr::from(([0, 0, 0, 0], port));
@ -34,7 +34,7 @@ pub async fn serve(
let port = server.local_addr().port();
info!("prometheus listening on port {}", port);
app.prometheus_port.store(port, Ordering::Relaxed);
app.prometheus_port.store(port, Ordering::SeqCst);
server
.with_graceful_shutdown(async move {

@ -672,7 +672,7 @@ impl ConsensusFinder {
0 => {}
1 => {
for rpc in self.rpc_heads.keys() {
rpc.tier.store(1, atomic::Ordering::Relaxed)
rpc.tier.store(1, atomic::Ordering::Release)
}
}
_ => {
@ -752,7 +752,7 @@ impl ConsensusFinder {
trace!("{} - p50_sec: {}, tier {}", rpc, median_latency_sec, tier);
rpc.tier.store(tier, atomic::Ordering::Relaxed);
rpc.tier.store(tier, atomic::Ordering::Release);
}
}
}
@ -810,7 +810,7 @@ impl ConsensusFinder {
HashMap::with_capacity(num_known);
for (rpc, rpc_head) in self.rpc_heads.iter() {
if !rpc.healthy.load(atomic::Ordering::Relaxed) {
if !rpc.healthy.load(atomic::Ordering::Acquire) {
// TODO: should unhealthy servers get a vote? they were included in minmax_block. i think that is enough
continue;
}
@ -878,14 +878,14 @@ impl ConsensusFinder {
pub fn best_tier(&self) -> Option<u32> {
self.rpc_heads
.iter()
.map(|(x, _)| x.tier.load(atomic::Ordering::Relaxed))
.map(|(x, _)| x.tier.load(atomic::Ordering::Acquire))
.min()
}
pub fn worst_tier(&self) -> Option<u32> {
self.rpc_heads
.iter()
.map(|(x, _)| x.tier.load(atomic::Ordering::Relaxed))
.map(|(x, _)| x.tier.load(atomic::Ordering::Acquire))
.max()
}
}

@ -155,7 +155,7 @@ impl Web3Rpc {
let backup = config.backup;
let block_data_limit: AtomicU64 = config.block_data_limit.into();
let automatic_block_limit = (block_data_limit.load(atomic::Ordering::Relaxed) == 0)
let automatic_block_limit = (block_data_limit.load(atomic::Ordering::Acquire) == 0)
&& block_and_rpc_sender.is_some();
// have a sender for tracking hard limit anywhere. we use this in case we
@ -287,7 +287,7 @@ impl Web3Rpc {
head_block = head_block.min(max_block);
}
let tier = self.tier.load(atomic::Ordering::Relaxed);
let tier = self.tier.load(atomic::Ordering::Acquire);
let backup = self.backup;
@ -354,7 +354,7 @@ impl Web3Rpc {
let request_scaling = 0.01;
// TODO: what ordering?
let active_requests =
self.active_requests.load(atomic::Ordering::Relaxed) as f32 * request_scaling + 1.0;
self.active_requests.load(atomic::Ordering::Acquire) as f32 * request_scaling + 1.0;
peak_latency.mul_f32(active_requests)
}
@ -439,7 +439,7 @@ impl Web3Rpc {
}
self.block_data_limit
.store(limit, atomic::Ordering::Relaxed);
.store(limit, atomic::Ordering::Release);
}
if limit == Some(u64::MAX) {
@ -763,7 +763,7 @@ impl Web3Rpc {
.await
.web3_context("failed check_provider")
{
self.healthy.store(false, atomic::Ordering::Relaxed);
self.healthy.store(false, atomic::Ordering::Release);
return Err(err);
}
@ -791,14 +791,14 @@ impl Web3Rpc {
break;
}
new_total_requests = rpc.internal_requests.load(atomic::Ordering::Relaxed)
+ rpc.external_requests.load(atomic::Ordering::Relaxed);
new_total_requests = rpc.internal_requests.load(atomic::Ordering::Acquire)
+ rpc.external_requests.load(atomic::Ordering::Acquire);
let detailed_healthcheck = new_total_requests - old_total_requests < 5;
// TODO: if this fails too many times, reset the connection
if let Err(err) = rpc.check_health(detailed_healthcheck, error_handler).await {
rpc.healthy.store(false, atomic::Ordering::Relaxed);
rpc.healthy.store(false, atomic::Ordering::Release);
// TODO: different level depending on the error handler
// TODO: if rate limit error, set "retry_at"
@ -808,7 +808,7 @@ impl Web3Rpc {
error!(?err, "health check on {} failed", rpc);
}
} else {
rpc.healthy.store(true, atomic::Ordering::Relaxed);
rpc.healthy.store(true, atomic::Ordering::Release);
}
// TODO: should we count the requests done inside this health check
@ -833,7 +833,7 @@ impl Web3Rpc {
true
};
self.healthy.store(initial_check, atomic::Ordering::Relaxed);
self.healthy.store(initial_check, atomic::Ordering::Release);
tokio::spawn(f)
} else {
@ -849,7 +849,7 @@ impl Web3Rpc {
// TODO: if this fails too many times, reset the connection
if let Err(err) = rpc.check_provider().await {
rpc.healthy.store(false, atomic::Ordering::Relaxed);
rpc.healthy.store(false, atomic::Ordering::Release);
// TODO: if rate limit error, set "retry_at"
if rpc.backup {
@ -858,7 +858,7 @@ impl Web3Rpc {
error!(?err, "provider check on {} failed", rpc);
}
} else {
rpc.healthy.store(true, atomic::Ordering::Relaxed);
rpc.healthy.store(true, atomic::Ordering::Release);
}
sleep(Duration::from_secs(health_sleep_seconds)).await;
@ -904,7 +904,7 @@ impl Web3Rpc {
let (first_exit, _, _) = select_all(futures).await;
// mark unhealthy
self.healthy.store(false, atomic::Ordering::Relaxed);
self.healthy.store(false, atomic::Ordering::Release);
debug!(?first_exit, "subscriptions on {} exited", self);
@ -1163,7 +1163,7 @@ impl Web3Rpc {
// TODO: if websocket is reconnecting, return an error?
if !allow_unhealthy {
if !(self.healthy.load(atomic::Ordering::Relaxed)) {
if !(self.healthy.load(atomic::Ordering::Acquire)) {
return Ok(OpenRequestResult::Failed);
}
@ -1384,17 +1384,17 @@ impl Serialize for Web3Rpc {
state.serialize_field(
"external_requests",
&self.external_requests.load(atomic::Ordering::Relaxed),
&self.external_requests.load(atomic::Ordering::Acquire),
)?;
state.serialize_field(
"internal_requests",
&self.internal_requests.load(atomic::Ordering::Relaxed),
&self.internal_requests.load(atomic::Ordering::Acquire),
)?;
state.serialize_field(
"active_requests",
&self.active_requests.load(atomic::Ordering::Relaxed),
&self.active_requests.load(atomic::Ordering::Acquire),
)?;
{
@ -1423,7 +1423,7 @@ impl Serialize for Web3Rpc {
state.serialize_field("weighted_latency_ms", &weighted_latency_ms)?;
}
{
let healthy = self.healthy.load(atomic::Ordering::Relaxed);
let healthy = self.healthy.load(atomic::Ordering::Acquire);
state.serialize_field("healthy", &healthy)?;
}
@ -1446,7 +1446,7 @@ impl fmt::Debug for Web3Rpc {
f.field("backup", &self.backup);
f.field("tier", &self.tier.load(atomic::Ordering::Relaxed));
f.field("tier", &self.tier.load(atomic::Ordering::Acquire));
f.field("weighted_ms", &self.weighted_peak_latency().as_millis());

@ -145,7 +145,7 @@ impl Drop for OpenRequestHandle {
fn drop(&mut self) {
self.rpc
.active_requests
.fetch_sub(1, atomic::Ordering::Relaxed);
.fetch_sub(1, atomic::Ordering::AcqRel);
}
}
@ -159,7 +159,7 @@ impl OpenRequestHandle {
// TODO: attach a unique id to this? customer requests have one, but not internal queries
// TODO: what ordering?!
rpc.active_requests
.fetch_add(1, std::sync::atomic::Ordering::Relaxed);
.fetch_add(1, std::sync::atomic::Ordering::AcqRel);
let error_handler = error_handler.unwrap_or_default();
@ -306,14 +306,16 @@ impl OpenRequestHandle {
match &authorization.authorization_type {
AuthorizationType::Frontend => {
// this is just a debug counter, so Relaxed is probably fine
self.rpc
.external_requests
.fetch_add(1, std::sync::atomic::Ordering::Relaxed);
.fetch_add(1, std::sync::atomic::Ordering::AcqRel);
}
AuthorizationType::Internal => {
// this is just a debug counter, so Relaxed is probably fine
self.rpc
.internal_requests
.fetch_add(1, std::sync::atomic::Ordering::Relaxed);
.fetch_add(1, std::sync::atomic::Ordering::AcqRel);
}
}

@ -1,559 +0,0 @@
use super::one::Web3Rpc;
use crate::errors::{Web3ProxyError, Web3ProxyErrorContext, Web3ProxyResult};
use crate::frontend::authorization::{Authorization, AuthorizationType};
use crate::globals::{global_db_conn, DB_CONN};
use crate::jsonrpc::{
self, JsonRpcErrorData, JsonRpcResultData, ResponsePayload, ValidatedRequest,
};
use anyhow::Context;
use chrono::Utc;
use derive_more::From;
use entities::revert_log;
use entities::sea_orm_active_enums::Method;
use ethers::providers::ProviderError;
use ethers::types::{Address, Bytes};
use futures::Future;
use http::StatusCode;
use migration::sea_orm::{self, ActiveEnum, ActiveModelTrait};
use nanorand::Rng;
use serde_json::json;
use std::pin::Pin;
use std::sync::atomic;
use std::sync::Arc;
use tokio::time::{Duration, Instant};
use tracing::{debug, error, info, trace, warn, Level};
#[derive(From)]
pub enum OpenRequestResult {
Handle(OpenRequestHandle),
/// Unable to start a request. Retry at the given time.
RetryAt(Instant),
/// The rpc are not synced, but they should be soon.
/// You should wait for the given block number.
/// TODO: should this return an OpenRequestHandle? that might recurse
Lagged(Pin<Box<dyn Future<Output = Web3ProxyResult<Arc<Web3Rpc>>> + Send>>),
/// Unable to start a request because no servers are synced or the necessary data has been pruned
Failed,
}
/// Make RPC requests through this handle and drop it when you are done.
/// Opening this handle checks rate limits. Developers, try to keep opening a handle and using it as close together as possible
pub struct OpenRequestHandle {
web3_request: Arc<ValidatedRequest>,
error_handler: RequestErrorHandler,
rpc: Arc<Web3Rpc>,
}
/// Depending on the context, RPC errors require different handling.
#[derive(Copy, Clone, Debug, Default)]
pub enum RequestErrorHandler {
/// Log at the trace level. Use when errors are expected.
#[default]
TraceLevel,
/// Log at the debug level. Use when errors are expected.
DebugLevel,
/// Log at the info level. Use when errors are expected.
InfoLevel,
/// Log at the error level. Use when errors are bad.
ErrorLevel,
/// Log at the warn level. Use when errors do not cause problems.
WarnLevel,
/// Potentially save the revert. Users can tune how often this happens
Save,
}
// TODO: second param could be skipped since we don't need it here
#[derive(serde::Deserialize, serde::Serialize)]
struct EthCallParams((EthCallFirstParams, Option<serde_json::Value>));
#[derive(serde::Deserialize, serde::Serialize)]
struct EthCallFirstParams {
to: Option<Address>,
data: Option<Bytes>,
}
impl std::fmt::Debug for OpenRequestHandle {
fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
f.debug_struct("OpenRequestHandle")
.field("method", &self.web3_request.inner.method())
.field("rpc", &self.rpc.name)
.finish_non_exhaustive()
}
}
impl From<Level> for RequestErrorHandler {
fn from(level: Level) -> Self {
match level {
Level::DEBUG => RequestErrorHandler::DebugLevel,
Level::ERROR => RequestErrorHandler::ErrorLevel,
Level::INFO => RequestErrorHandler::InfoLevel,
Level::TRACE => RequestErrorHandler::TraceLevel,
Level::WARN => RequestErrorHandler::WarnLevel,
}
}
}
impl Authorization {
/// Save a RPC call that return "execution reverted" to the database.
async fn save_revert(
self: Arc<Self>,
method: Method,
params: EthCallFirstParams,
) -> Web3ProxyResult<()> {
let rpc_key_id = match self.checks.rpc_secret_key_id {
Some(rpc_key_id) => rpc_key_id.into(),
None => {
// trace!(?self, "cannot save revert without rpc_key_id");
return Ok(());
}
};
let db_conn = global_db_conn()?;
// TODO: should the database set the timestamp?
// we intentionally use "now" and not the time the request started
// why? because we aggregate stats and setting one in the past could cause confusion
let timestamp = Utc::now();
let to = params.to.unwrap_or_else(Address::zero).as_bytes().to_vec();
let call_data = params.data.map(|x| x.to_string());
let rl = revert_log::ActiveModel {
rpc_key_id: sea_orm::Set(rpc_key_id),
method: sea_orm::Set(method),
to: sea_orm::Set(to),
call_data: sea_orm::Set(call_data),
timestamp: sea_orm::Set(timestamp),
..Default::default()
};
let rl = rl
.save(&db_conn)
.await
.web3_context("Failed saving new revert log")?;
// TODO: what log level and format?
trace!(revert_log=?rl);
// TODO: return something useful
Ok(())
}
}
impl Drop for OpenRequestHandle {
fn drop(&mut self) {
self.rpc
.active_requests
.fetch_sub(1, atomic::Ordering::Relaxed);
}
}
impl OpenRequestHandle {
pub async fn new(
web3_request: Arc<ValidatedRequest>,
rpc: Arc<Web3Rpc>,
error_handler: Option<RequestErrorHandler>,
) -> Self {
// TODO: take request_id as an argument?
// TODO: attach a unique id to this? customer requests have one, but not internal queries
// TODO: what ordering?!
rpc.active_requests
.fetch_add(1, std::sync::atomic::Ordering::Relaxed);
let error_handler = error_handler.unwrap_or_default();
Self {
web3_request,
error_handler,
rpc,
}
}
pub fn connection_name(&self) -> String {
self.rpc.name.clone()
}
#[inline]
pub fn clone_connection(&self) -> Arc<Web3Rpc> {
self.rpc.clone()
}
pub fn rate_limit_for(&self, duration: Duration) {
if self.rpc.backup {
debug!(?duration, "rate limited on {}!", self.rpc);
} else {
warn!(?duration, "rate limited on {}!", self.rpc);
}
// TODO: use send_if_modified to be sure we only send if our value is greater
self.rpc
.hard_limit_until
.as_ref()
.unwrap()
.send_replace(Instant::now() + duration);
}
/// Just get the response from the provider without any extra handling.
/// This lets us use the try operator which makes it much easier to read
async fn _request<R: JsonRpcResultData + serde::Serialize>(
&self,
) -> Web3ProxyResult<jsonrpc::SingleResponse<R>> {
// TODO: replace ethers-rs providers with our own that supports streaming the responses
// TODO: replace ethers-rs providers with our own that handles "id" being null
if let (Some(url), Some(ref client)) = (self.rpc.http_url.clone(), &self.rpc.http_client) {
// prefer the http provider
let request = self
.web3_request
.inner
.jsonrpc_request()
.context("there should always be a request here")?;
let response = client.post(url).json(request).send().await?;
if response.status() == StatusCode::TOO_MANY_REQUESTS {
// TODO: how much should we actually rate limit?
self.rate_limit_for(Duration::from_secs(1));
}
let response = response.error_for_status()?;
jsonrpc::SingleResponse::read_if_short(response, 1024, &self.web3_request).await
} else if let Some(p) = self.rpc.ws_provider.load().as_ref() {
// use the websocket provider if no http provider is available
let method = self.web3_request.inner.method();
let params = self.web3_request.inner.params();
// some ethers::ProviderError need to be converted to JsonRpcErrorData. the rest to Web3ProxyError
let response = match p.request::<_, R>(method, params).await {
Ok(x) => jsonrpc::ParsedResponse::from_result(x, self.web3_request.id()),
Err(provider_error) => match JsonRpcErrorData::try_from(&provider_error) {
Ok(x) => jsonrpc::ParsedResponse::from_error(x, self.web3_request.id()),
Err(ProviderError::HTTPError(error)) => {
if let Some(status_code) = error.status() {
if status_code == StatusCode::TOO_MANY_REQUESTS {
// TODO: how much should we actually rate limit?
self.rate_limit_for(Duration::from_secs(1));
}
}
return Err(provider_error.into());
}
Err(err) => {
warn!(?err, "error from {}", self.rpc);
return Err(provider_error.into());
}
},
};
Ok(response.into())
} else {
// this must be a test
Err(anyhow::anyhow!("no provider configured!").into())
}
}
pub fn error_handler(&self) -> RequestErrorHandler {
if let RequestErrorHandler::Save = self.error_handler {
let method = self.web3_request.inner.method();
// TODO: should all these be Trace or Debug or a mix?
// TODO: this list should come from config. other methods might be desired
if !["eth_call", "eth_estimateGas"].contains(&method) {
// trace!(%method, "skipping save on revert");
RequestErrorHandler::TraceLevel
} else if DB_CONN.read().is_ok() {
let log_revert_chance = self.web3_request.authorization.checks.log_revert_chance;
if log_revert_chance == 0 {
// trace!(%method, "no chance. skipping save on revert");
RequestErrorHandler::TraceLevel
} else if log_revert_chance == u16::MAX {
// trace!(%method, "gaurenteed chance. SAVING on revert");
self.error_handler
} else if nanorand::tls_rng().generate_range(0u16..u16::MAX) < log_revert_chance {
// trace!(%method, "missed chance. skipping save on revert");
RequestErrorHandler::TraceLevel
} else {
// trace!("Saving on revert");
// TODO: is always logging at debug level fine?
self.error_handler
}
} else {
// trace!(%method, "no database. skipping save on revert");
RequestErrorHandler::TraceLevel
}
} else {
self.error_handler
}
}
/// Send a web3 request
/// By having the request method here, we ensure that the rate limiter was called and connection counts were properly incremented
/// depending on how things are locked, you might need to pass the provider in
/// we take self to ensure this function only runs once
/// This does some inspection of the response to check for non-standard errors and rate limiting to try to give a Web3ProxyError instead of an Ok
pub async fn request<R: JsonRpcResultData + serde::Serialize>(
self,
) -> Web3ProxyResult<jsonrpc::SingleResponse<R>> {
// TODO: use tracing spans
// TODO: including params in this log is way too verbose
// trace!(rpc=%self.rpc, %method, "request");
trace!("requesting from {}", self.rpc);
let authorization = &self.web3_request.authorization;
match &authorization.authorization_type {
AuthorizationType::Frontend => {
self.rpc
.external_requests
.fetch_add(1, std::sync::atomic::Ordering::Relaxed);
}
AuthorizationType::Internal => {
self.rpc
.internal_requests
.fetch_add(1, std::sync::atomic::Ordering::Relaxed);
}
}
// we used to fetch_add the active_request count here, but sometimes a request is made without going through this function (like with subscriptions)
// we generally don't want to use the try operator. we might need to log errors
let start = Instant::now();
let mut response = self._request().await;
// measure successes and errors
// originally i thought we wouldn't want errors, but I think it's a more accurate number including all requests
let latency = start.elapsed();
// we used to fetch_sub the active_request count here, but sometimes the handle is dropped without request being called!
trace!(
"response from {} for {}: {:?}",
self.rpc,
self.web3_request,
response,
);
// TODO: move this to a helper function?
// true if we got a jsonrpc result. a jsonrpc error or other error is false.
// TODO: counters for errors vs jsonrpc vs success?
let response_is_success = match &response {
Ok(jsonrpc::SingleResponse::Parsed(x)) => {
matches!(&x.payload, ResponsePayload::Success { .. })
}
Ok(jsonrpc::SingleResponse::Stream(..)) => true,
Err(_) => false,
};
if response_is_success {
// only track latency for successful requests
tokio::spawn(async move {
self.rpc.peak_latency.as_ref().unwrap().report(latency);
self.rpc.median_latency.as_ref().unwrap().record(latency);
// TODO: app-wide median and peak latency?
});
} else {
// only save reverts for some types of calls
// TODO: do something special for eth_sendRawTransaction too
// we do **NOT** use self.error_handler here because it might have been modified
let error_handler = self.error_handler();
enum ResponseType {
Error,
Revert,
RateLimited,
}
let response_type: ResponseType = match &response {
Ok(jsonrpc::SingleResponse::Parsed(x)) => match &x.payload {
ResponsePayload::Success { .. } => unreachable!(),
ResponsePayload::Error { error } => {
trace!(?error, "jsonrpc error data");
if error.message.starts_with("execution reverted") {
ResponseType::Revert
} else if error.code == StatusCode::TOO_MANY_REQUESTS.as_u16() as i64 {
ResponseType::RateLimited
} else {
// TODO! THIS HAS TOO MANY FALSE POSITIVES! Theres another spot in the code that checks for things.
// if error.message.contains("limit") || error.message.contains("request") {
// self.rate_limit_for(Duration::from_secs(1));
// }
match error.code {
-32000 => {
if error.message.contains("MDBX_PANIC:") {
response = Err(Web3ProxyError::MdbxPanic(
self.connection_name(),
error.message.clone(),
));
} else {
// TODO: regex?
let archive_prefixes = [
"header not found",
"header for hash not found",
"missing trie node",
];
for prefix in archive_prefixes {
if error.message.starts_with(prefix) {
// TODO: what error?
response = Err(Web3ProxyError::ArchiveRequired {
min: self.web3_request.min_block_needed(),
max: self.web3_request.max_block_needed(),
});
break;
}
}
}
ResponseType::Error
}
-32001 => {
if error.message == "Exceeded the quota usage" {
ResponseType::RateLimited
} else {
ResponseType::Error
}
}
-32005 => {
if error.message == "rate limit exceeded" {
ResponseType::RateLimited
} else {
ResponseType::Error
}
}
-32601 => {
let error_msg = error.message.as_ref();
// sometimes a provider does not support all rpc methods
// we check other connections rather than returning the error
// but sometimes the method is something that is actually unsupported,
// so we save the response here to return it later
// some providers look like this
if (error_msg.starts_with("the method")
&& error_msg.ends_with("is not available"))
|| error_msg == "Method not found"
{
let method = self.web3_request.inner.method().to_string();
response =
Err(Web3ProxyError::MethodNotFound(method.into()))
}
ResponseType::Error
}
_ => ResponseType::Error,
}
}
}
},
Ok(jsonrpc::SingleResponse::Stream(..)) => unreachable!(),
Err(_) => ResponseType::Error,
};
if matches!(response_type, ResponseType::RateLimited) {
// TODO: how long?
self.rate_limit_for(Duration::from_secs(1));
}
match error_handler {
RequestErrorHandler::DebugLevel => {
// TODO: think about this revert check more. sometimes we might want reverts logged so this needs a flag
if matches!(response_type, ResponseType::Revert) {
trace!(
rpc=%self.rpc,
%self.web3_request,
?response,
"revert",
);
} else {
debug!(
rpc=%self.rpc,
%self.web3_request,
?response,
"bad response",
);
}
}
RequestErrorHandler::InfoLevel => {
info!(
rpc=%self.rpc,
%self.web3_request,
?response,
"bad response",
);
}
RequestErrorHandler::TraceLevel => {
trace!(
rpc=%self.rpc,
%self.web3_request,
?response,
"bad response",
);
}
RequestErrorHandler::ErrorLevel => {
// TODO: only include params if not running in release mode
error!(
rpc=%self.rpc,
%self.web3_request,
?response,
"bad response",
);
}
RequestErrorHandler::WarnLevel => {
// TODO: only include params if not running in release mode
warn!(
rpc=%self.rpc,
%self.web3_request,
?response,
"bad response",
);
}
RequestErrorHandler::Save => {
trace!(
rpc=%self.rpc,
%self.web3_request,
?response,
"bad response",
);
// TODO: do not unwrap! (doesn't matter much since we check method as a string above)
// TODO: open this up for even more methods
let method: Method =
Method::try_from_value(&self.web3_request.inner.method().to_string())
.unwrap();
// TODO: i don't think this prsing is correct
match serde_json::from_value::<EthCallParams>(json!(self
.web3_request
.inner
.params()))
{
Ok(params) => {
// spawn saving to the database so we don't slow down the request
// TODO: log if this errors
// TODO: aren't the method and params already saved? this should just need the response
let f = authorization.clone().save_revert(method, params.0 .0);
tokio::spawn(f);
}
Err(err) => {
warn!(
%self.web3_request,
?response,
?err,
"failed parsing eth_call params. unable to save revert",
);
}
}
}
}
}
response
}
}

@ -551,20 +551,20 @@ impl RpcQueryStats {
// TODO: do this without a clone
let authorization = metadata.authorization.clone();
let archive_request = metadata.archive_request.load(Ordering::Relaxed);
let archive_request = metadata.archive_request.load(Ordering::Acquire);
// TODO: do this without cloning. we can take their vec
let backend_rpcs_used = metadata.backend_rpcs_used();
let request_bytes = metadata.inner.num_bytes() as u64;
let response_bytes = metadata.response_bytes.load(Ordering::Relaxed);
let response_bytes = metadata.response_bytes.load(Ordering::Acquire);
let mut error_response = metadata.error_response.load(Ordering::Relaxed);
let mut response_millis = metadata.response_millis.load(Ordering::Relaxed);
let mut error_response = metadata.error_response.load(Ordering::Acquire);
let mut response_millis = metadata.response_millis.load(Ordering::Acquire);
let user_error_response = metadata.user_error_response.load(Ordering::Relaxed);
let user_error_response = metadata.user_error_response.load(Ordering::Acquire);
let response_timestamp = match metadata.response_timestamp.load(Ordering::Relaxed) {
let response_timestamp = match metadata.response_timestamp.load(Ordering::Acquire) {
0 => {
// no response timestamp!
if !error_response {

@ -154,7 +154,7 @@ impl TestApp {
})
};
let mut frontend_port = frontend_port_arc.load(Ordering::Relaxed);
let mut frontend_port = frontend_port_arc.load(Ordering::SeqCst);
let start = Instant::now();
while frontend_port == 0 {
// we have to give it some time because it might have to do migrations
@ -163,7 +163,7 @@ impl TestApp {
}
sleep(Duration::from_millis(10)).await;
frontend_port = frontend_port_arc.load(Ordering::Relaxed);
frontend_port = frontend_port_arc.load(Ordering::SeqCst);
}
let proxy_endpoint = format!("http://127.0.0.1:{}", frontend_port);