TCP/IP Bind Failed

[ConvertFromOldNGs]

by torrie >> Wed, 31 Oct 2001 11:41:18 GMT

When opening a TCP/IP connection immediately after closing a connection on the same port I often get a TCP/IP Bind Failed exception (31005) It appears that the TCP/IP stack is still holding onto that port even though I have finished with it. I can get around it by closing the jade node that was using the port, waiting and then restarting. Does anyone know if I can just wait and then try to connect again after waiting for a period of time?

[ConvertFromOldNGs]

by cnwrab1 >> Wed, 19 Jun 2002 5:32:03 GMT

Torrie,

Some code I was developing had this annoying problem. It happened at the server end of the connection after the tcp object had been passed to my callback method following callup to a listenAsynch or listenContiniuousAsynch.

I cured it by running the server (called up) end as a Fat Client. The only time I get 31005s now is following a catastrophic software failure. Presumably the problem I had was related to testing it from a Dev Env running as a Thin client.

Hope this helps.

Adrien

[ConvertFromOldNGs]

by allistar >> Thu, 20 Jun 2002 0:48:32 GMT

Hi there,
I have the same issue: A server application method that binds to a particular port. When I "kill" the server application process, the
port is still bound to, which prevents me from re-binding to it. The
only way I have found around this is to close and restart the database server again.

Allistar. ------------------------------------------------------------------
Allistar Melville
Software Developer
Auckland, NEW ZEALAND

Greentree International,
Developer of Greentree Financial Software. ------------------------------------------------------------------

[ConvertFromOldNGs]

by cnwjhp1 >> Thu, 20 Jun 2002 0:57:12 GMT

Could this be because the dll attaches to the Windows process rather than the thread? When I was calling a dll for a tcp interface (using thin client for testing), I had to drop the jadapp to replace the dll. There can be multiple threads in the jadapp.exe (ie, multiple thin client sessions), but only one Windows process, and the dll is attached to the process, not the threads. So using a fat client means you only have to bounce the client to reset.

[ConvertFromOldNGs]

by encos_dev >> Mon, 14 Oct 2002 6:29:43 GMT

We unfortunately have the same problem with 31005 tcpip bind failed. We are stress testing our software and when we have a lot of connections pumping a lot of data we get this problem. Our only way out is by closing down the Node.

What can lead to tcpip bind failed (31005) apart from not closing a tcpip connection instance & deleting it?

We are not using thin-client. We intend the connection monitor app I am talking about to run as a server app - but we have been testing it out of this mode (i.e., not on the same node as the RAP).

It seems to us that Jade looses it's handle to the tcpip connection which means we have no way of closing or releasing the port (apart from closing the whole sha-bang down)

Any help or comments much appreciated,


Glen
Encos

[ConvertFromOldNGs]

by allistar >> Mon, 14 Oct 2002 22:15:21 GMT

I can't comment on the 31005 exception except to say that with our
tcp/ip listener application (which runs inside the RAP) when we "end user" it using the monitor we can no longer bind to that port. It
seems that Jade isn't releasing that port when the process is ended in this manner. The only fix, as you have suggested, is to close and
restart the entire Jade database server.

Allistar. ------------------------------------------------------------------
Allistar Melville
Software Developer
Auckland, NEW ZEALAND

Greentree International,
Developer of Greentree Financial Software. ------------------------------------------------------------------