IBM Mainframe Discussion List <[email protected]> wrote on 09/03/2007 02:15:40 AM:
> >So you almost always want exactly one and only Store Status for SAD > >to capture. > > >In real life, many shops take advantage of HMC wizardry to define SAD load > >profile(s). These profiles must be different from regular OS load profiles > >because the IPL device for SAD is never a sysres volume and may differ > >from one LPAR to another. You have the option in an SAD load profile-- > >still, I think?--to set the Store Status option. > > Yes, the option is still there. > > >The temptation is to set it because this is by definition a SAD profile. > >I used to do that until I was advised otherwise: if the first shot > at SAD >goes awry for some transitory and immediately correctable > reason, you DO >NOT want a second Store Status on retry. Only the > first Store Status, the >one from the OS, is of any use UNLESS you > trying to debug SAD itself. BTW >I have never once done that in my > whole career. > > Well, this thread comes in handy. I have just generated (and tested) > sadump for 1.8. And during testing, I clicked on the option for > store status (despite having read NOT to do it, as I was really > unsure). The sadump I then took was definitely readable and was NOT > a dump of an sadump. It was showing everything I expected it to > show, including all system address spaces that were up at the time. > > I had been regenerating sadump from tape to DASD (dsntype LARGE), > and my first attempt went wrong due to operator error. I re-IPL'd > sadump (without store status, I think), and *that* was definitely a > dump of a dump (sadump even tells me so). > > So I guess now I am really confused with the whole issue of store > status or not. When SADMP gets through its initialization and then goes awry and you need to start over, the first thing to try is a PSW Restart. If that works, it starts the dump over without losing any data. If you do have to ReIPL SADMP, you will lose the 4 megabytes of data for the frames which SADMP used for its code and buffers. SADMP tries to use MVS read-only nucleus frames for this purpose, and these are rarely required for debugging MVS problems. Since they are read-only, we know what should be there, unless the problem was an overlay via a real address or channel program. If you have to ReIPL SADMP on z/OS 1.8 or an older release, SADMP will only dump real storage, and is unable to locate MVS control blocks to allow it dump paged-out virtual storage. I improved SADMP in z/OS 1.9 so that it can dump paged-out storage if SADMP is reIPLed. Jim Mulder z/OS System Test IBM Corp. Poughkeepsie, NY ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
