TCP Issues in HP-UX 11i

[Mani Subramanian]

Are there any known issues with the following description ? If so are there any solutions or patches available ?

We have two applications running on the same system. one is a server and opens TCP socket for listening (bound to IP address 0). the other is a client and attempts to connect to server's socket. depending on client configuration connect() call will either be issued to 127.0.0.1 or to the actual IP address of the box. after reboot both methods function just fine. however, after a while the first method (connecting to loopback interface) starts failing (we don't have exact error out of connect()), while the second method continues to succeed. restarting
client and/or server does not resolve the issue -- only rebooting entire machine does.

[Jeff Schussele · Accepted solution]

Hi Mani,

I've seen this happen when either "localhost" or the 127.0.0.1 entries get into DNS somehow.
Also can happen if the TCP stack for the network card gets corrupted. Make sure you're up to date on patches & verify that neither the name nor IP is getting into DNS.

Rgds,
Jeff

[Mani Subramanian]

Thanks Jeff,

Is there any Min patch level on this regard ?

Also is there a way to check for the TCP Stack corruption ?

Regards
Mani

[Tom Dineen_2]

do you have any lingering resources pending ? like Fin-wait status ?

netstat -an

[Mani Subramanian]

No, I do not see many of them just 3 or 4 connections in TIME_WAIT status. Also all the tests regarding DNS issues seems to work fine.

Still not sure on the exact problem. The strange thing is the machine in one particular subnet gives this problem but not on the other subnet (exact same hardware / software / OS / Patchlist )