Question & Answer
Question
Many z/OS "debugging aids" intercept system dump processing. How can they be disabled so a complete, unformatted, SYSMDUMP can be captured for analysis via IPCS? What other measures are needed to ensure that a usable dump will be obtained?
Answer
Here are the basics.
1. Make certain you have a SYSMDUMP DD statement in the JCL of the job or started task.
2. SYSMDUMP datasets must be pre-allocated with RECFM=F, LRECL=4160.
3. z/OS dumps can be very large. A typical zWS CONTROLLER dump is likely to be 300 cylinders or more. Make certain the dump dataset is large enough.
4. Abends often occur several at a time, and the very first one is normally the one that is most useful in debugging the problem. Therefore, SYSMDUMP datasets should be allocated DISP=MOD so that subsequent dumps do not overlay the first.
5. When a dump dataset is allocated MOD, it may fill up over time, so that unless the user deletes and reallocates the dataset after each series of dumps, it may not be able to accept more data when needed. This problem can be avoided by a directing the SYSMDUMP to a GDG dataset. That way, a new generation will be created each time the owning job is run, or the started task is restarted.
6. To prevent debugging applications from intercepting the dump, so that only an "analysis" is captured, add one of the following DUMMY DD statements to the JCL/PROC as needed.
//IDIOFF DD DUMMY IBM FAULT ANALYZER OFF
//ABNLIGNR DD DUMMY ABEND-AID OFF
//ESPYIBM DD DUMMY EYE-SPY OFF
//CAOESTOP DD DUMMY CA-OPT II & CA-SYMDUMP OFF
//DMBENAN DD DUMMY TURN OFF DUMPMASTER
//PSPOFF DD DUMMY TURN OFF SOFTWORKS PERFORMANCE ESSENTIAL
2. SYSMDUMP datasets must be pre-allocated with RECFM=F, LRECL=4160.
3. z/OS dumps can be very large. A typical zWS CONTROLLER dump is likely to be 300 cylinders or more. Make certain the dump dataset is large enough.
4. Abends often occur several at a time, and the very first one is normally the one that is most useful in debugging the problem. Therefore, SYSMDUMP datasets should be allocated DISP=MOD so that subsequent dumps do not overlay the first.
5. When a dump dataset is allocated MOD, it may fill up over time, so that unless the user deletes and reallocates the dataset after each series of dumps, it may not be able to accept more data when needed. This problem can be avoided by a directing the SYSMDUMP to a GDG dataset. That way, a new generation will be created each time the owning job is run, or the started task is restarted.
6. To prevent debugging applications from intercepting the dump, so that only an "analysis" is captured, add one of the following DUMMY DD statements to the JCL/PROC as needed.
//IDIOFF DD DUMMY IBM FAULT ANALYZER OFF
//ABNLIGNR DD DUMMY ABEND-AID OFF
//ESPYIBM DD DUMMY EYE-SPY OFF
//CAOESTOP DD DUMMY CA-OPT II & CA-SYMDUMP OFF
//DMBENAN DD DUMMY TURN OFF DUMPMASTER
//PSPOFF DD DUMMY TURN OFF SOFTWORKS PERFORMANCE ESSENTIAL
[{"Product":{"code":"SSRULV","label":"IBM Workload Scheduler for z\/OS"},"Business Unit":{"code":"BU058","label":"IBM Infrastructure w\/TPS"},"Component":"--","Platform":[{"code":"PF035","label":"z\/OS"}],"Version":"All Versions;Version Independent","Edition":"","Line of Business":{"code":"LOB35","label":"Mainframe SW"}}]
Was this topic helpful?
Document Information
Modified date:
28 November 2023
UID
swg21389957