Server crashes during restart

[atimin]

If the server restart and there are connected clients, it crushes sometimes with the following message in the log:

19-02-06 15:52:37:381 INFO  [9240] [..\src\servermain.cpp:58]   [2418] accepted new secure channel from client : Address=<127.0.0.1>, Port=<50436>
2019-02-06 15:52:37:388 INFO  [9240] [..\src\servermain.cpp:58]   [2418] accepted new secure channel from client : Address=<127.0.0.1>, Port=<50437>
2019-02-06 15:52:37:394 ERROR [9240] [..\src\servermain.cpp:64]   [2418] receive invalid channel id in OpenSecureChannelRequest : Local-Address=<0.0.0.0>, Local-Port=<48012>, Partner-Address=<127.0.0.1>, Partner-Port=<50436>, ChannelId=<10>
2019-02-06 15:52:37:397 ERROR [9240] [..\src\servermain.cpp:64]   [2418] receive invalid channel id in OpenSecureChannelRequest : Local-Address=<0.0.0.0>, Local-Port=<48012>, Partner-Address=<127.0.0.1>, Partner-Port=<50437>, ChannelId=<12>

Restart sequence:

    opc_server->stop();
    opc_server->shutdown();  

    LOG_INFO("Init Server IO");

    ConfigXml xmlParser;
    xmlParser.parse("OpcUaServer.xml", &conf);

    opc_server->config(conf);

    opc_server->init();
    opc_server->start();

where opc_server is std::shared_ptr<OpcUaStackServer::Server>

OS: Windows 7
Version: 3.6.2

[huebl]

Error Description:

After restarting the application, the channel should be reopened by the client. The client wants to reuse the previous channel and sends a reconnect open secure channel request. The channel does not exist anymore. Therefore, the server closes the connection. Thereafter, the server accesses the closed socket again.

Error Fixing:
After closing the socket it must not be used anymore,

[huebl]

The error description described above has not been confirmed. Testing with multiple OPC UA clients (Prosys OPC UA Client and UAExpert) did not cause any crashes.

@FLIPBack - Can you provide a CORE dump for this bug?

[huebl]

It could be a problem specific to Windows. The cancel command

  __declspec(deprecated("By default, this function always fails with "
        "operation_not_supported when used on Windows XP, Windows Server 2003, "
        "or earlier. Consult documentation for details."))
#endif
  void cancel()
  {
    boost::system::error_code ec;
    this->get_service().cancel(this->get_implementation(), ec);
    boost::asio::detail::throw_error(ec, "cancel");
  }

The cancel command should be replaced by the close command. Maybe that fixes the bug.

[atimin]

The error description described above has not been confirmed. Testing with multiple OPC UA clients (Prosys OPC UA Client and UAExpert) did not cause any crashes.

@FLIPBack - Can you provide a CORE dump for this bug?

I'll try. Unfortunately the bug's always appearing then I'm focusing on other problem, but when I try it repeat on purpose I've never had success 😕

[atimin]

The cancel command should be replaced by the close command. Maybe that fixes the bug.

Sorry, can't understand where and what I should replace.

[huebl]

You do not have to do anything. I have replaced the cancel command in the Secure Channel with the Close command.

[atimin]

@huebl , I see. If you add Closes #109 into related PR, I'll see in this issue that there is a solution. 😉

[huebl]

I hope so

[atimin]

I hope so

I mean PR's description on GitHub not the commit.