On 2018-08-08 5:58 AM, Cieri, Anthony wrote:
The audit trail is needed to determine what job(s) instigate the S822 ABENDs.
Well, maybe that audit trail will work now (and maybe it won't)- but it definitely would not have been reliable in the past.
In the olden days - even before the iPhone 6 was invented - the JES2 INIT number was more of a logical concept or a virtual queue - and did not mean any particular address space.
For example, suppose the "INIT 2" address space A1 was doing job termination. Then "INIT 2" would be eligible to run a new job. But suppose there were other started initiators that had no work in their assigned classes, and so address spaces A2 and A3 were idle initiators.
JES2 may well start the next job under "INIT 2" in address space A2 or A3 as JES2 saw fit, perhaps before address space A1 had become truly idle.
S822 abends are probably caused by LSQA fragmentation, where LSQA is sloppy terminology for "private high" virtual storage. If you want to find out who the "culprit" is, you need to track the ASID, not the INIT number, in my view.
In a previous life as a sysprog, I wrote and implemented an MPF exit to append the ASID to the STARTED and ENDED messages - the ones produced by the MONITOR or MN command IIRC, but it might also work for the JES2 messages. It reported the ASID for started tasks and TSO users as well, without actually increasing the number of messages.
You may not be interested in messages as such, but the most convenient window into the past I had at the time was the SYSLOG. If you want to look at SMF, then you probably do not need to change much because the ASID is in the type 30 (SMF30ASI).
Even if my observations about JES2 INIT numbers are out of date or even totally incorrect, ASID is still a valid thing to track when trying to track storage fragmentation within the private area of an individual address space, so you still don't need to audit JES2 INIT numbers.
And don't forget there is always the el cheapo version of the therapeutic IPL - just drain and restart all the initiators every so often when convenient for pristine unfragmented private storage for your batch jobs.
Cheers, Greg ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
