Thanks, James - I guess I will try to find some TEC support then. Just wondering if anybody had seen this problem.
There's nothing wrong with the perl script though, given that I can reproduce the same problem from the command line (i.e. not involving my script at all). nvaction.alog is good - no problems there.
Thanks,
John.
>From: James Shanks <jshanks@us.ibm.com>
>Reply-To: nv-l@lists.us.ibm.com
>To: nv-l@lists.us.ibm.com
>Subject: Re: [nv-l] wpostzmsg hanging or core dumping
>Date: Wed, 15 Sep 2004 08:27:31 -0400
>
>wpostzmsg is a part of TEC, not NetView. I suggest you open a problem to
>TEC and get them to look at your perl script and your problems with using
>wpostzmsg. The ruleset is just a mechanism to pass the proper parms to
>your script so that it will work. You can see that if you look in the
>nvaction.alog/blog.
>
>James Shanks
>Level 3 Support for Tivoli NetView for UNIX and Windows
>Tivoli Software / IBM Software Group
>
>
>
>"John H" <herber_j@hotmail.com>
>Sent by: owner-nv-l@lists.us.ibm.com
>09/15/2004 07:08 AM
>Please respond to
>nv-l
>
>
>To
>nv-l@lists.us.ibm.com
>cc
>
>Subject
>[nv-l] wpostzmsg hanging or core dumping
>
>
>
>
>
>
>Environment: AIX4-r1, NetView 7.1.3, set up as TEC Endpoint
>
>I'm setting up TEC forwarding in Netview via a ruleset calling a perl
>script. The perl script does various things, sets up the event class /
>slots as required, and calls wpostzmsg. It used to use system() to call it
>so I could track the result, but now I fork() the program, and use exec()
>in the child so I can watch the child from the parent program.
>So, to the questions / issues:
>1) The documentation I read said that I should use wpostzmsg as it was
>effectively the replacement for wpostemsg - any comments on that? Am I
>right to do this?
>2) Most event forwarding is just fine. Some of the wpostzmsg calls
>however, end up as hung processes - and hang the parent perl process too,
>despite it having a timed SIGALRM that will attempt to kill the child
>process and itself after 2 minutes.
>3) Some calls result in a core dump. Calling it repeatedly from the
>command line sometimes gives this:
>Illegal instruction(coredump) wpostzmsg -m "Test" TEST_CLASS TEST_SOURCE
>Has anybody else seen this? I can reproduce the hung processes and
>coredumps (mixture thereof) as well as some successful sends at the
>command line by issuing "wpostmsg -m "Test" TEST_CLASS TEST_SOURCE &"
>repeatedly so that there is more than one process running concurrently.
>It would seem then that there is an issue with the volume of wpostzmsg
>calls - but surely this is something that wpostzmsg was designed for? I
>can't imagine I have to attempt some kind of lock process to ensure that
>my scripts only ever call wpostzmsg one at a time?
>Comments and advice gratefully received, as I'm getting bored of running
>my "reap zombie processes" shell script!
>Many thanks,
>John.
>
>What type of MSN user are you? Take our quick quiz and find out!