-
Notifications
You must be signed in to change notification settings - Fork 1.1k
Open
Description
Bug Report
server
let svc = ServiceBuilder::new()
.concurrency_limit(100_000)
.load_shed();
let mut server = Server::builder()
.tls_config(tls_config)
.unwrap()
.concurrency_limit_per_connection(256)
.tcp_keepalive(Some(Duration::from_secs(60)))
.http2_keepalive_interval(Some(Duration::from_secs(20)))
.http2_keepalive_timeout(Some(Duration::from_secs(60)))
.max_concurrent_streams(Some(512))
.layer(svc)
.layer(SessionMiddlewareLayer::new(db, redis_connection_manager));
....
let listener = TcpListener::bind(SocketAddr::V4(SocketAddrV4::new(Ipv4Addr::UNSPECIFIED, CONFIG.port))).await.map_err(|e| {
eprintln!("Failed to bind TCP listener: {}", e);
e.to_string()
})?;
let incoming = TcpIncoming::from(listener)
.inspect_err(|e| {
if let Some(raw_os_error) = e.raw_os_error() {
if raw_os_error == libc::EMFILE {
println!("FD exhausted (EMFILE): {}", raw_os_error);
}
}
// if let Some(io) = e.downcast_ref::<std::io::Error>() {
// if io.raw_os_error() == Some(libc::EMFILE) {
// println!("FD exhausted (EMFILE): {io}");
// }
// }
});
server.serve_with_incoming(incoming)
.await.map_err(|e| {
eprintln!("gRPC Server Error: {}", e);
e.to_string()
})client
final channel = ClientChannel(
actualHost,
port: actualPort,
options: ChannelOptions(
credentials: MyChannelCredentials.secure(selfSecurityContext()),
codecRegistry: CodecRegistry(codecs: [GzipCodec(), IdentityCodec()]),
keepAlive: ClientKeepAliveOptions(
pingInterval: Duration(seconds: 20),
permitWithoutCalls: true,
),
),
);
final options = CallOptions(providers: [_metadataProvider]);
final interceptors = [_ErrorHandlingInterceptor(() => onUnauthenticated)];Version
tonic v0.14.2
tonic-prost v0.14.2
tonic v0.14.2 (*)
tonic-prost-build v0.14.2
tonic-build v0.14.0
Platform
Windows x64
Description
gRPC Error (code: 2, codeName: UNKNOWN, message: HTTP/2 error: Connection error: Connection is being forcefully terminated. (errorCode: 10), details: null, rawResponse: null, trailers: {})
This error occurs when using persistent connections (server streams) or during server-side breakpoint debugging. When deployed to production, I sometimes get a large number of FD exhausted (EMFILE) messages, but the cause is unknown. Is there any way to prevent interruptions and avoid the FD exhausted (EMFILE) message?
Metadata
Metadata
Assignees
Labels
No labels