Problem with messaging after installing patches

[Wayne D Lepkin]

Running 10.20, we have an application with two processes that put messages on a UNIX message queue, and a third process that takes the messages from both processes off the same queue. After recently installing patches that brought us up to approximately February '00, one of the message-sending processes no longer works. When attempting to do a 'msgsnd' to place a message on the queue, we're receiving unusual errno return codes (for example, 82), and the message does not make it on the queue. The first process (puts messages on) and the third process (takes them off) continue to work fine. Since we installed relatively old patches, we presume that not very many others have experienced this. Our "guess" is that perhaps one of the patches "tightened up" something, so that if the process was doing something "against the rules" before but was getting away with it, it's not any more. That's fine, but we need to understand what it's doing wrong, and which patch is causing the "problem". Anyone have any thoughts about this? Thanks very much.

[Albert E. Whale, CISSP]

Wayne,

It's not clear if your problem is with customized software (written in-house) or with third party software.

If it is with customized software written in house, then I would recommend that you recompile it. Perhaps one of the patches was the Libc or shared Library Patches?

[Andy Monks]

error 82 isn't an hp-ux error number. 72 is (eoverflow). I would check to see if your application is returning error 82 as a response to a specific hp-ux error.

Assuming it's error 72, it's possible, that patching changed the system file and maybe some of the messaging kernel parameters have been changed. Worth checking anyway.

Andy

[Wayne D Lepkin]

Albert, it's a combination of inhouse and 3rd-party software, and the 3rd-party stuff comes in pre-linked without source code, so we can't recompile or relink it. I believe we have recompiled the inhouse code. One of the patches does indeed hit libc, and that patch (PHCO_20441) is one of the candidates for backing out.

Andy, the 82 was just an example, in fact it's my understanding that we get different invalid codes at different times. I know that I as the SA haven't changed any kernel parameters -- is it possible for a patch to change the system file?

Thank you both for your replies.

[Albert E. Whale, CISSP]

Wayne,

At lthis point I would first check to see if you have the Latest Shared Libraries Patch. That one is first.

Secondly, the Kernel Patches could change a configuration file, but it may not be very likely. Instead I would assume that you may need to increase a few parameters dure to a Kernel Patch. When in dount, check it out! Look at the modification dates in the /stand dir and the /stand/build dirs.

Lastly if none of these work, check for a newer patch which supercededs one of the ones you've installed. Even though you've installed the latest Group, doesn't mean that they've been tested together WIHT all oyour software installed (though it's pretty likely that it's fairly stable together).

Lastly, Have you run the Custom Patch Manager? Perhaps this is the simplest method to update the server.

http://us-support2.external.hp.com/wps/bin/doc.pl/sid=4bda9d010dacbb4efa

Hope that helps.