> >>> I feel that our system trace is a bit small, especially since we > moved to z13. We currently have set them to 15M (per processor). > >>> What do others use? > > > >> No-one interested in reading system trace? Or is this company secrect? > > > > I'm surprised you got no reactions. > > > > Perhaps if you say *where* do you set that limit per processor, > this will stir IBM-MAIN into life? > > > > I'm also curious *where* do you set it... even after I have RTFM > and GYIF and STFI... > > > > System trace size is set on the TRACE operator command > (TRACE,ST,nnnK/nnnM/nG), along with master trace size, and some other > options. > > We offer some (admittedly very slim) guidance in z/OS MVS Diagnosis: > Tools and Service Aids, in the Chapter on System Trace (Chapter 8). > > After a short chat with someone who reads a lot of dumps, I find this is > a more complicated question than it appears to be. Speaking only for > myself, 15X the default size seems like more than enough.
For master trace, I prefer something like 500K or 999K. Even the maximum of 999K is a pretty small amount of storage at today's prices, and a larger MTRACE provides the convenience of sometimes not needing to figure out where the syslog is archived. And for standalone dumps, a better chance that you won't need to dig around looking for buffers of stuff that has wrapped out of MTRACE but hasn't yet been hardened in syslog or operlog. The minimum and default (since z/OS 1.10) of 1M per CPU for system trace is often adequate, and that is what is used in many of the dumps I see. (As of z/OS 2.1, IPCS SYSTRACE STATUS will display this information). Larger system traces come with some trade-offs. The trace buffers are in fixed storage, and the snapshot buffers are also fixed while a snapshot is being created. The larger the trace, the more CPU time it takes to copy the data into a snapshot. Currently, most entries to RTM2 result in a snapshot, before we know if any dump is going to be taken. We are considering changing that behavior in the next release to take a snapshot only when a dump is actually being taken. The IPCS report dialog uses a 2GB data space for the lines in the report. Using a wide 3270 screen, this works out to a a maximum of around 16 million lines. Very large formatted system traces can exceed 16 million lines. Of course, you can do IPCS SYSTRACE PRINT NOTERM to send the report to a large data set, and then browse the data set, but that is a nuisance and it does slow you down a bit. Or you can use SYSTRACE filtering to reduce the number of formatted lines. The new EXCLUDE(BEFOREALL AFTERALL) introduced for z/OS 2.2 in APAR OA49814 can be very useful for this purpose. Jim Mulder z/OS Diagnosis, Design, Development, Test IBM Corp. Poughkeepsie NY ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
