Receiving "Changing data structure forced command termination." from onstat!

[Bryan Osborne]

Getting a "Changing data structure forced command termination." message when
executing onstat (-g ses, -g ath, and others). oninit is running at ~98% of
CPU (via "top"), so system is busy, but can things change that fast?!?!

What makes me more nervous is that I receive a "Memory dump (core)" when
executing some onchecks (-cc, -ce, -pt), although the -cr worked fine.

Any ideas out there, folks?

[Sanjeev Sagar]

Just a thought b'coz one time it happened with one of my another DBA.

Run "fuser $INFORMIXDIR/bin/*" and check which process thread is running on
informix binaries. Check for ontape or onbar binaries. In my case it was a old
ontape process which was hanging. When I killed that process, onstat was working
fine.

Well, it was just a thought. I do not know that why it happen.

GOODLUCK!!!

[Bryan Osborne]

Thanks Sanjeev.

Unfortunately, fuser didn't show anything.

A note of clarification, onstat (-g ses, -g ath, and others) will show a
number of threads prior to giving error message: "Changing data structure
forced command termination."

Anyone other ideas? Anyone else seen this before?

TIA...

Sanjeev Sagar wrote in message <3869238B...@hotmail.com>...

[Madison Pruet]

This is actually the error handler for a SEGV for onstat.

We don't 'freeze' shared memory when run an onstat, but most onstats read from
shared memory. Therefor it is possible that the underlying datastructures can
be changing while onstat is actually processing the request. For instance, if
while running through the thread or session lists the list changes, then onstat
will often encountere a SEGV. Instead of aborting, the message "Changing data
structure..." is printed.

Unless it is consistantly occuring at the same spot, I wouldn't worry too much
about it. If it is consistantly occuring at the same spot, then I'd contact
tech support.

Bryan Osborne wrote:

--
Madison Pruet

===========================================
Enterprise Replication Product Developement
Dallas, Texas
Informix Software
===========================================

[Murray Wood]

Bryan

This could be normal activity because the SMI tables are being changed at
the time you requested information (onstat) from them. The message is
information only, telling you that you dont have all the information
available.

There was a bug in early 7 engines (e.g 7.14) whereby this would always
occur if you had a user connected to the engine from a PC with a name
greater than 8 characters. Limit the length of PC names or upgrade the
engine here.

Murray Wood

[Rudy Fernandes]

I've seen this occur under HP-UX when shared memory kernel parameters weren't
set correctly (see the Release notes directory under INFORMIXDIR for
recommendations). I vaguely recall that if "Total Resident memory" allocated to
the instance exceeded the shmmax setting at the OS level, the OS needed to
create two segments for resident memory at instance initialization. Under this
condition, similar "Changing ..." warning messages occurred, especially when
the instance was busy. You could check a couple of items - (1) Your online log
for warnings about shmmax settings (2) Your actual shmmax settings and compare
it to the total Resident size (onstat -g seg). Also, if possible, try
experimenting on a test machine.

Rudy
(Cheers, Scott).