Seems we have to reboot our HP-UX Baan application servers about every other month from reduced performance. Wondered if anybody else has noticed this. My UNIX admin seems to think it's memory fragmentation. We're seeing this on machines supporting primarily OLTP transactions as well as another server dedicated to just Baan batch jobs. Our cpu and run queue graphs have a slow rise supporting this. After about six weeks you can tell at the command line that the system is sluggish.

We're running Baan IVc4 on HP-UX 11.11 with porting set 6.1c.06.07 with Tools service pack 12 installed.

We faced some bad performance on our HP-Box due to a bad porting set version.

It also depends on lots of factors like -

1. number of users
2. Unix kernel parameters
3. BaaN-Oracle driver parameters
4. bad application programming
5. shared memory - on/off
6. number of companies & databases
7. memory & size of the server

If you suspect memory issues you can use glance to track worst process -
you find a bshell with a large memory allocation and look at its shared memory allocations for RSS and VSS.

Then break down like-
Process RSS value Shared Mem RSS VSS value Shared Mem VSS
bshell6.2 or bshell_lmssbn
Oracle8-1
Oracle8-2
Oracle8-3
sh_server
audit
...
The hard part is isolating the problem and cause.

What's frustrating is the boxes are responsive after reboots and over time seem to degrade. It's not something you notice the next day but after a month or so. Thanks for your reply, I'll keep looking.

Instead of rebooting the machine, have you tried to only stop and start the Baan environment(s)? Could be that this gives you back your performance. If it does, you've already narrowed it down to Baan being the bottleneck! Then, as a next step, if you are using the Baan job daemon to run jobs, try stopping/starting only the job daemon(s). Again, if this solves the problem, you've narrowed it down to (one of) the jobs. Worth a try...

Also, have you check to see if you have any 'defunct' or old processes (like bshells ) running that are no longer accumulating any run time. Could be the number of processes have grown over time which could make the UNIX OS (thus BaaN ) sluggish.

As an old time and long ago Unix Admin, we use to always re-boot our Unix boxes at least once a month if not once a week (when possible) to cure any ills on the Unix OS system (like old processes).

Hope this helps

Use the ps -ef command and sort on the 7th column, check after a while again. The processes which are changing rapidly in time are using most of the capacity of the system. Mostly they are zomby processes.

Chad, are you using MWA/OVPA or 'sar' to keep an eye on critical indicators like kernel table usage, swap, etc? If not you should consider using them and see how these evolve over a monthly period.

Regards,

Pat.

Hi there,

first of all a clear no no from me.

I have several customers with your principal configuration, meaning Baan IVc4 on Oracle on HP-Ux. Only one is booting out of superstition his server every day!!! All the others (more than 20) only boot their servers, when HW needs to be replaced or a Unix kernel patch is installed. We ussually see uptimes of 100 days and more.

I agree with the comments above. Your initial assumption must be, that what you see is not normal. The next must be to find the source of trouble:

- Zombie processes in Unix
- buggy versions in HP-Ux/Oracle/Baan
- defect HW
- memeory leaks?

When your performance is bad, it should be pretty easy to identify (at best using glance) the overall memory consumption, top usage processes, system ressources and so on.


Soory for the pundit: Rebooting as a mechanism to solve a computer problem is a pure MS Windows approach and the first sign of loosing control of your system.

But please do not start a flame war on this, windows has also its advantages,

Markus

Please note that 8.1.6 had some 'real' issues. It is worth a try to test your system under 8.1.7.4.

thx