|
Boost : |
From: JH (jupiter.hce_at_[hidden])
Date: 2019-06-27 12:57:05
Hi Gavin,
>From other Internet discussions, it seems there is no way to switch an
existing TCP socket connection from one interface to another,
particularly, the NAT mapping is linked to 4-tuple, the only solution
is terminate the connection by deleting the socket, then create a new
socket and connection again. If you have different view could you
please elaborate how Linux OS
can take care of routing the data appropriately?
Thank you very much.
Kind regards,
- jupiter
On 6/25/19, JH <jupiter.hce_at_[hidden]> wrote:
> Hi Gavin,
>
> I should clarify it is a client socket connection.
>
> On 6/25/19, Gavin Lambert via Boost <boost_at_[hidden]> wrote:
>> On 25/06/2019 17:43, JH wrote:
>>> I have an application using ASIO TCP/IP SSL socket connection in an
>>> embedded device, the device has two network interfaces, when one is
>>> down, another network interface takes over, which requires application
>>> TCP SSL socket connection to be established binding to a new network
>>> interface. What will be an appropriate implementation to switch
>>> application TCP connection from one to another interface?
>>
>> The simplest option is to not bind to a network interface in the first
>> place. Provided that only at most one is up at a given moment, the OS
>> will take care of routing the data appropriately.
>
> Did you mean to remove bind interface in client code, right? Sorry
> where I should to remove the bind to a network interface? Here is my
> client connection code:
>
> void AsyncClientSslConnection::InitialConnection(void) {
> boost::asio::ip::tcp::resolver::query query(this->mHost, this->mPort);
> boost::asio::ip::tcp::resolver::iterator iterator =
> this->mResolver.resolve(query);
> this->mSocket.set_verify_mode(boost::asio::ssl::verify_peer);
>
> this->mSocket.set_verify_callback(boost::bind(&AsyncClientSslConnection::VerifyCertificate,
> this, _1, _2));
> boost::asio::async_connect(this->mSocket.lowest_layer(), iterator,
> boost::bind(&AsyncClientSslConnection::HandleConnect, this,
> boost::asio::placeholders::error));
> }
>
> At the moment, when one interface is down, the client TCP socket keeps
> sending the data to server, but the remote server could no longer
> receive it.
>
>> You're never allowed to have both interfaces up at the same time to the
>> same network, unless the OS itself has special routing to handle
>> failover -- in which case again, you don't need anything fancy in the
>> application.
>
> You right, there would be only one interface has IP address, the OS is
> Linux Ubuntu and Yocto OS, I guess both should be able to handle
> failover, the key is where the binding statement?
>
> Thank you very much.
>
> - j
>
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk