hello, actualy I'm testing my server. I'm making new connections faster than server is able to handle it. because of it server quickly is getting from system errno 24 (Too many open files). because of chaining async operation in asio design async_accept handler like below lead to server hang on. void AnldServer::handle_accept(const boost::system::error_code& e) { if (!e) { new_connection_->start(); new_connection_.reset(new AnldConnection(io_service_, request_handler_)); acceptor_.async_accept(new_connection_->socket(), boost::bind(&AnldServer::handle_accept, this, boost::asio::placeholders::error)); }else{ NDSLOG_NOTICE_S() << "Accept error: " << e << "\n"; } } Since we are not starting a new accept operation the io_service will run out of work to do and the server will hang on (if it was daemonised) or exit(if not). ofcourse server in production environment couldn't behaves like this. server must can handle such situation. my question is: how to fix it? just start async_accept without creating new_connection in case of error? like this for example? void AnldServer::handle_accept(const boost::system::error_code& e) { if (!e) { new_connection_->start(); new_connection_.reset(new AnldConnection(io_service_, request_handler_)); acceptor_.async_accept(new_connection_->socket(), boost::bind(&AnldServer::handle_accept, this, boost::asio::placeholders::error)); }else{ NDSLOG_NOTICE_S() << "Accept error: " << e << "\n"; acceptor_.async_accept(new_connection_->socket(), boost::bind(&AnldServer::handle_accept, this, boost::asio::placeholders::error)); } } I have just tested above code - it seems to work fine. but maybe someone have beter solution> or maybe it is good idea to add sleep(5) before async_accept in case of error? thanks, tom ---------------------------------------------------- Twoja rodzina na bliscy.pl Zobacz: http://klik.wp.pl/?adr=http%3A%2F%2Fcorto.www.wp.pl%2Fas%2Fbliscy.html&sid=628