Unable to delete zombie TCPIP$PWIP_ACP sockets

[Jonathan Price_1]

Dear All,

On one of our systems TCPIP$PWIP_ACP is the owner process of 100-200 what I call "dead" sockets, i.e. where Local and Remote ports = 0 and the Remote Host = *.

The number of dead sockets increases over time. I am investigating what is causing these sockets to appear. (Prime suspect is an RPC server application.)

Unfortunately when I attempt to shutdown either TCPIP (via TCPIP$CONFIG) or the PWIP driver itself (via TCPIP$PWIP_DRIVER_SHUTDOWN.COM) VMS crashes!

Looking at TCPIP$PWIPACP_.LOG I can see that something is attempting a connection every 30 seconds. The resulting log entry is:

Tue Apr 14 10:19:56 2009: connReq_ast : *** Connect Request failed *** : 556

While I continue to track down the culprit do you have any suggestions? Specifically, do you know a means of deleting or clearing a specific socket connection?

Thank you very much. The attachment has system details. In brief, these are:
VMS 8.3 unpatched, Decnet Phase V, TCPIP 5.6, on an ES45

[Hoff]

You may end up calling HP support here (it fills up with sockets, and it crashes), but the first thing they're going to ask you to get current on your ECO kits. So I'd start there.

[Jonathan Price_1]

I've downloaded the latest patch kit for VMS 8.3 but didn't see anything that addresses TCPIP$PWIP_ACP socket issues specifically.

Nevertheless, patch I must.

Thanks for the comment.

[Hoff]

There are (usually) patch kits available for OpenVMS and for TCP/IP Services and for DECnet-Plus (Phase V).

[Volker Halle · Accepted solution]

Jonathan,

did you try $ TCPIP DISC DEVICE BGxxx:

Volker.

[Jonathan Price_1]

Of course! DISCONNECT DEVICE_SOCKET does the job. I don't know why I didn't see that before.

Thanks for your help.

[Jonathan Price_1]

The solution will be to
1) disconnect zombie sockets as they appear
2) apply the latest patches
3) track down the remote system(s) that are creating the zombies in the first place.

Later.