Conference tuxedo::dce-products

2221.0. "Rpc not ready yet 382312451 0x16c9a003" by RDGENG::CHAMBERLIN (Danger! Do not Reverse Polarity) Thu Apr 17 1997 07:56

A partner has an application which uses an NT 3.51 client conneccting to
multiple Alpha Unix Server processes (V3.2D) using DCE-RPC on predefined sockets.

When the server is shutdown, and then restarted (client stays running), there
seems to be a time delay (~ 1-2 min) before the server can reconnect to the
previuosly used sockets. Connection attempts return the above error message.


Is there some timeout before sockets can be reused, or are we seeing some other
problem? 

If a timeout, can it be adjusted, or any other way round it?

Is there any documentation re the error message numbers?

Thanks,
		Ian Chamberlin.

	Software Partner Engineering (UK).

    Yes, there's a naturally occuring timeout period before the client
    detects that the server has gone away.  Any open connections will have
    to be shutdown before communications can occur again.  The timeout
    period depends upon the protocol being used.  For UDP it should be
    fairly quick (say, less than a minute).  For TCP it can be rather long
    (say, a couple of hours).
    
    You can check the current state of the sockets with 'netstat -a'.  The
    sockets that are being closed will be in a 'TIME_WAIT' state
    
    The error you're seeing is:
    % grep 382312451 /usr/include/dce/rpcsts.h
    #define rpc_s_cant_bind_socket (382312451)
    
    ..which makes sense under the circumstances.
    
    -- brandon