BIT LISTSERV IBM-MAIN 29 RE ABEND822 Z OS
From: JEREMIAH.EDEN@no-spam (Jeremiah Eden)
Subject: Re: abend822 z/OS
Date: 7 Jul 2003 08:02:55 -0700


Martin,

Our systems group tries to be as pro-active as possible and tune systems so that there is the least impact when resources are unavailable. Especially. since our real time cash lanes are tied indirectly to our DB2 system for Credit Cards sales,
etc. So, no DB2, no Sale.

So being, pro-active, the only problem I have is the word 'unlimited' which is related to specifying 0K|0M. This allows extended private to grow to any size without a check until in infringes on the Buffer pools, work pools, edm pool etc.

If everything were perfect and every piece of code in DB2 and MVS were perfect,
I wouldn't be concerned. So if there is .0001% chance that this could happen, and Murphy's law runs rampant in 24/7 shops, then we have to try and determine ahead of time what the least amount of impact would be.

Given that this is a possibility, then what would be the least impact on DB2,
and take the least amount of time to resolve? Running out of Extended Private,
or running out of Work Space? If I have a cap on extended private and the DB2
modules are already loaded, then the big problem would be VSAM processing which would show up as data set open and close failures. This could be alleviated very quickly by lowering DSMAX on the fly, assuming I had 8K to load IDCAMS close module below the line. Or, we could stop a database that is known to have a lot of objects.

Now if I don't have a cap on extended private and it does grow to the point that it uses DB2 work space, then I have seen various performance problems and even DB2 crash. Usually the first sign is the SOS reason code you get back in the applications, but if this is a critical SYSOPR unit of work that is in a must complete status or is holding locks then it gets more complicated.
At that point, if the system is up, we are involved in a more timely analysis to determine what unit or units of work are causing the problem. And cycling DB2 is usually the last resort after console dumps and storage analysis.

I hope you see my point on this and you can probably add 1000 more since this is your specialty. But it seems the damage and repair would take less time on a private storage failure than a shortage in a work pool. So, having a hard cap on private storage appears to be the best route to go.

-----Original Message-----
From: Martin Packer [mailto:martin_packer@no-spam
Sent: Friday, July 04, 2003 3:42 AM To: IBM-MAIN@no-spam
Subject: Re: abend822 z/OS

Jeremiah scribbled... :-)

> Hey Martin, believe it or not I put a Cap on DBM1 of 64M, so if something is > filling up low extended private besides DB2 load modules and what he needs > for the table space allocations, I will know about it.

I think I heard you mention that before.

So do you expect DB2 to Abend or just for the load/GETMAIN to fail? In the former case I'm not sure your users'd be happy about it. :-)

Martin Packer, MBCS Martin Packer/UK/IBM 020-8832-5167 in the UK (+44) (MOBX 273643, Internal 7-325167, Mobile 07802-245584)

----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to listserv@no-spam with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html