mirror of
https://github.com/pchuan98/codex.git
synced 2026-07-01 00:31:56 +08:00
Allow large remote app-server resume responses (#19920)
## Why Remote TUI resume uses the app-server websocket client. That client inherited tungstenite's default `16 MiB` frame limit, so a large saved session could make `thread/resume` return a single JSON-RPC response frame that the client rejected before the TUI could deserialize or render it. Fixes #19837 ## What Changed - Configure the remote app-server websocket client with a bounded `128 MiB` max frame/message size. - Preserve the concrete remote worker exit reason when completing pending requests after a transport/read failure instead of replacing it with a generic channel-closed error. - Add a regression test that sends a single `>16 MiB` JSON-RPC response frame and verifies the typed request succeeds. Note: This isn't a perfect fix. It really just moves the limit to a much larger value. I looked at a bunch of other potential fixes (both server-side and client-side), and they all involved significant complexity, had backward-compatibility impact, or impacted performance of common use cases. This simple fix should address the vast majority of remote use cases. ## Verification I reproed the problem locally using a long rollout. Verified that fix addresses connection drop.
This commit is contained in:
committed by
GitHub
Unverified
parent
fc2a69107c
commit
92fb848065
@@ -1396,6 +1396,55 @@ mod tests {
|
||||
client.shutdown().await.expect("shutdown should complete");
|
||||
}
|
||||
|
||||
#[tokio::test]
|
||||
async fn remote_typed_request_accepts_large_single_frame_response() {
|
||||
let padding = "x".repeat((17 << 20) + 1024);
|
||||
let websocket_url = start_test_remote_server(move |mut websocket| async move {
|
||||
expect_remote_initialize(&mut websocket).await;
|
||||
let JSONRPCMessage::Request(request) = read_websocket_message(&mut websocket).await
|
||||
else {
|
||||
panic!("expected account/read request");
|
||||
};
|
||||
assert_eq!(request.method, "account/read");
|
||||
write_websocket_message(
|
||||
&mut websocket,
|
||||
JSONRPCMessage::Response(JSONRPCResponse {
|
||||
id: request.id,
|
||||
result: serde_json::json!({
|
||||
"account": null,
|
||||
"requiresOpenaiAuth": false,
|
||||
"padding": padding,
|
||||
}),
|
||||
}),
|
||||
)
|
||||
.await;
|
||||
websocket.close(None).await.expect("close should succeed");
|
||||
})
|
||||
.await;
|
||||
let client = RemoteAppServerClient::connect(test_remote_connect_args(websocket_url))
|
||||
.await
|
||||
.expect("remote client should connect");
|
||||
|
||||
let response: GetAccountResponse = client
|
||||
.request_typed(ClientRequest::GetAccount {
|
||||
request_id: RequestId::Integer(1),
|
||||
params: codex_app_server_protocol::GetAccountParams {
|
||||
refresh_token: false,
|
||||
},
|
||||
})
|
||||
.await
|
||||
.expect("large typed request should succeed");
|
||||
assert_eq!(
|
||||
response,
|
||||
GetAccountResponse {
|
||||
account: None,
|
||||
requires_openai_auth: false,
|
||||
}
|
||||
);
|
||||
|
||||
client.shutdown().await.expect("shutdown should complete");
|
||||
}
|
||||
|
||||
#[tokio::test]
|
||||
async fn remote_connect_includes_auth_header_when_configured() {
|
||||
let auth_token = "remote-bearer-token".to_string();
|
||||
|
||||
@@ -45,16 +45,18 @@ use tokio::sync::oneshot;
|
||||
use tokio::time::timeout;
|
||||
use tokio_tungstenite::MaybeTlsStream;
|
||||
use tokio_tungstenite::WebSocketStream;
|
||||
use tokio_tungstenite::connect_async;
|
||||
use tokio_tungstenite::connect_async_with_config;
|
||||
use tokio_tungstenite::tungstenite::Message;
|
||||
use tokio_tungstenite::tungstenite::client::IntoClientRequest;
|
||||
use tokio_tungstenite::tungstenite::http::HeaderValue;
|
||||
use tokio_tungstenite::tungstenite::http::header::AUTHORIZATION;
|
||||
use tokio_tungstenite::tungstenite::protocol::WebSocketConfig;
|
||||
use tracing::warn;
|
||||
use url::Url;
|
||||
|
||||
const CONNECT_TIMEOUT: Duration = Duration::from_secs(10);
|
||||
const INITIALIZE_TIMEOUT: Duration = Duration::from_secs(10);
|
||||
const REMOTE_APP_SERVER_MAX_WEBSOCKET_MESSAGE_SIZE: usize = 128 << 20;
|
||||
|
||||
#[derive(Debug, Clone)]
|
||||
pub struct RemoteAppServerConnectArgs {
|
||||
@@ -170,20 +172,32 @@ impl RemoteAppServerClient {
|
||||
request.headers_mut().insert(AUTHORIZATION, header_value);
|
||||
}
|
||||
ensure_rustls_crypto_provider();
|
||||
let stream = timeout(CONNECT_TIMEOUT, connect_async(request))
|
||||
.await
|
||||
.map_err(|_| {
|
||||
IoError::new(
|
||||
ErrorKind::TimedOut,
|
||||
format!("timed out connecting to remote app server at `{websocket_url}`"),
|
||||
)
|
||||
})?
|
||||
.map(|(stream, _response)| stream)
|
||||
.map_err(|err| {
|
||||
IoError::other(format!(
|
||||
"failed to connect to remote app server at `{websocket_url}`: {err}"
|
||||
))
|
||||
})?;
|
||||
// Remote resume responses can legitimately carry large thread histories.
|
||||
// Keep a bounded cap, but raise it above tungstenite's 16 MiB frame default.
|
||||
let websocket_config = WebSocketConfig::default()
|
||||
.max_frame_size(Some(REMOTE_APP_SERVER_MAX_WEBSOCKET_MESSAGE_SIZE))
|
||||
.max_message_size(Some(REMOTE_APP_SERVER_MAX_WEBSOCKET_MESSAGE_SIZE));
|
||||
let stream = timeout(
|
||||
CONNECT_TIMEOUT,
|
||||
connect_async_with_config(
|
||||
request,
|
||||
Some(websocket_config),
|
||||
/*disable_nagle*/ false,
|
||||
),
|
||||
)
|
||||
.await
|
||||
.map_err(|_| {
|
||||
IoError::new(
|
||||
ErrorKind::TimedOut,
|
||||
format!("timed out connecting to remote app server at `{websocket_url}`"),
|
||||
)
|
||||
})?
|
||||
.map(|(stream, _response)| stream)
|
||||
.map_err(|err| {
|
||||
IoError::other(format!(
|
||||
"failed to connect to remote app server at `{websocket_url}`: {err}"
|
||||
))
|
||||
})?;
|
||||
let mut stream = stream;
|
||||
let pending_events = initialize_remote_connection(
|
||||
&mut stream,
|
||||
@@ -198,6 +212,7 @@ impl RemoteAppServerClient {
|
||||
let worker_handle = tokio::spawn(async move {
|
||||
let mut pending_requests =
|
||||
HashMap::<RequestId, oneshot::Sender<IoResult<RequestResult>>>::new();
|
||||
let mut worker_exit_error: Option<(ErrorKind, String)> = None;
|
||||
loop {
|
||||
tokio::select! {
|
||||
command = command_rx.recv() => {
|
||||
@@ -224,17 +239,19 @@ impl RemoteAppServerClient {
|
||||
.await
|
||||
{
|
||||
let err_message = err.to_string();
|
||||
let message = format!(
|
||||
"remote app server at `{websocket_url}` write failed: {err_message}"
|
||||
);
|
||||
if let Some(response_tx) = pending_requests.remove(&request_id) {
|
||||
let _ = response_tx.send(Err(err));
|
||||
}
|
||||
let _ = deliver_event(
|
||||
&event_tx,
|
||||
AppServerEvent::Disconnected {
|
||||
message: format!(
|
||||
"remote app server at `{websocket_url}` write failed: {err_message}"
|
||||
),
|
||||
message: message.clone(),
|
||||
},
|
||||
);
|
||||
worker_exit_error = Some((ErrorKind::BrokenPipe, message));
|
||||
break;
|
||||
}
|
||||
}
|
||||
@@ -351,28 +368,34 @@ impl RemoteAppServerClient {
|
||||
.await
|
||||
{
|
||||
let err_message = reject_err.to_string();
|
||||
let message = format!(
|
||||
"remote app server at `{websocket_url}` write failed: {err_message}"
|
||||
);
|
||||
let _ = deliver_event(
|
||||
&event_tx,
|
||||
AppServerEvent::Disconnected {
|
||||
message: format!(
|
||||
"remote app server at `{websocket_url}` write failed: {err_message}"
|
||||
),
|
||||
message: message.clone(),
|
||||
},
|
||||
);
|
||||
worker_exit_error =
|
||||
Some((ErrorKind::BrokenPipe, message));
|
||||
break;
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
Err(err) => {
|
||||
let message = format!(
|
||||
"remote app server at `{websocket_url}` sent invalid JSON-RPC: {err}"
|
||||
);
|
||||
let _ = deliver_event(
|
||||
&event_tx,
|
||||
AppServerEvent::Disconnected {
|
||||
message: format!(
|
||||
"remote app server at `{websocket_url}` sent invalid JSON-RPC: {err}"
|
||||
),
|
||||
message: message.clone(),
|
||||
},
|
||||
);
|
||||
worker_exit_error =
|
||||
Some((ErrorKind::InvalidData, message));
|
||||
break;
|
||||
}
|
||||
}
|
||||
@@ -383,14 +406,19 @@ impl RemoteAppServerClient {
|
||||
.map(|frame| frame.reason.to_string())
|
||||
.filter(|reason| !reason.is_empty())
|
||||
.unwrap_or_else(|| "connection closed".to_string());
|
||||
let message = format!(
|
||||
"remote app server at `{websocket_url}` disconnected: {reason}"
|
||||
);
|
||||
let _ = deliver_event(
|
||||
&event_tx,
|
||||
AppServerEvent::Disconnected {
|
||||
message: format!(
|
||||
"remote app server at `{websocket_url}` disconnected: {reason}"
|
||||
),
|
||||
message: message.clone(),
|
||||
},
|
||||
);
|
||||
worker_exit_error = Some((
|
||||
ErrorKind::ConnectionAborted,
|
||||
message,
|
||||
));
|
||||
break;
|
||||
}
|
||||
Some(Ok(Message::Binary(_)))
|
||||
@@ -398,25 +426,29 @@ impl RemoteAppServerClient {
|
||||
| Some(Ok(Message::Pong(_)))
|
||||
| Some(Ok(Message::Frame(_))) => {}
|
||||
Some(Err(err)) => {
|
||||
let message = format!(
|
||||
"remote app server at `{websocket_url}` transport failed: {err}"
|
||||
);
|
||||
let _ = deliver_event(
|
||||
&event_tx,
|
||||
AppServerEvent::Disconnected {
|
||||
message: format!(
|
||||
"remote app server at `{websocket_url}` transport failed: {err}"
|
||||
),
|
||||
message: message.clone(),
|
||||
},
|
||||
);
|
||||
worker_exit_error = Some((ErrorKind::InvalidData, message));
|
||||
break;
|
||||
}
|
||||
None => {
|
||||
let message = format!(
|
||||
"remote app server at `{websocket_url}` closed the connection"
|
||||
);
|
||||
let _ = deliver_event(
|
||||
&event_tx,
|
||||
AppServerEvent::Disconnected {
|
||||
message: format!(
|
||||
"remote app server at `{websocket_url}` closed the connection"
|
||||
),
|
||||
message: message.clone(),
|
||||
},
|
||||
);
|
||||
worker_exit_error = Some((ErrorKind::UnexpectedEof, message));
|
||||
break;
|
||||
}
|
||||
}
|
||||
@@ -424,12 +456,14 @@ impl RemoteAppServerClient {
|
||||
}
|
||||
}
|
||||
|
||||
let err = IoError::new(
|
||||
ErrorKind::BrokenPipe,
|
||||
"remote app-server worker channel is closed",
|
||||
);
|
||||
let (err_kind, err_message) = worker_exit_error.unwrap_or_else(|| {
|
||||
(
|
||||
ErrorKind::BrokenPipe,
|
||||
"remote app-server worker channel is closed".to_string(),
|
||||
)
|
||||
});
|
||||
for (_, response_tx) in pending_requests {
|
||||
let _ = response_tx.send(Err(IoError::new(err.kind(), err.to_string())));
|
||||
let _ = response_tx.send(Err(IoError::new(err_kind, err_message.clone())));
|
||||
}
|
||||
});
|
||||
|
||||
|
||||
Reference in New Issue
Block a user