I had been working on a bug and getting out a System Dump (SDUMP). Then, after a fix, I stopped getting dumps, and just got a message with a return code indicating an abend had occurred. So where was my dump?
z/OS has Dump Elimination and Analysis DAE. This keeps a note of the dumps being taken, and can be configured to say “If you get the same abend many times, don’t bother taking a dump”. This stops you getting many identical dumps, and filling your disk storage.
In IPCS option 3 UTILITY – Perform utility functions, 5 DAE – Process DAE data, you can display the contents of the DAE information. On my system the data set is SYS1.DAE.
Command ===> Scroll ===> PAGE
Enter an Action Code next to an entry.
Enter / next to an entry to choose from a list of Action Codes.
Dataset: 'SYS1.DAE'
Dumps since last DAE Display: 0 Total Dumps suppressed: 214
Events since last DAE Display: 0 Suppression rate: 81%
A Last Last Total Date of Symptom String information:
C Date System Events Dump Abend Reason Module CSECT
_ 05/01/26 S0W1 7 04/28/26 S0EC6 055B0718 BPXINPVT BPXFSMNT
_ 05/01/26 S0W1 66 04/28/26 S00C4 00000004 BPXINPVT BPXVOTHD
_ 05/01/26 S0W1 10 04/28/26 S00C4 00000004 BPXINPVT BPXVOTHD
s 05/01/26 S0W1 53 04/28/26 S00C4 00000004 BPXINPVT BPXVOTHD
Selecting a record with S gave
Date Time System Name
Last (most recent) Event: 05/01/26 19:02:16 S0W1
Dump Taken: 04/28/26 18:44:59 S0W1
Symptoms used for Dump Suppression:
MVS RETAIN
Key Key Symptom Data Explanation
MOD/ RIDS/ BPXINPVT LOAD MODULE NAME
CSECT/ RIDS/ BPXVOTHD ASSEMBLY MODULE CSECT NAME
PIDS/ PIDS/ 5752SCPX1 PRODUCT/COMPONENT IDENTIFIER
AB/S AB/S 00C4 ABEND CODE-SYSTEM
REXN/ RIDS/ BPXMIPCE RECOVERY ROUTINE CSECT NAME
FI/ VALU/H 7542A0704742A0708BB917 FAILING INSTRUCTION AREA
REGS/ REGS/ C07DC REG/PSW DIFFERENCE
HRC1/ PRCS/ 00000004 ABEND REASON CODE
SUB1/ VALU/C OPENMVS COMPONENT SUBFUNCTION
This information is called a Symptom String. It provides a very short summary of the problem. You can search the internet with this information, to see if the problem has been found before.
For me, this symptom string, there has been 66 occurrences of it in my system (see the first display).
DAE can be configured to say if you get the same symptom string, do not take a dump, as is is most probably the same problem. This means your disks are not fully of identical system dumps.
Similar dumps
In
Command ===> Scroll ===> PAGE
Enter an Action Code next to an entry.
Enter / next to an entry to choose from a list of Action Codes.
Dataset: 'SYS1.DAE'
Dumps since last DAE Display: 0 Total Dumps suppressed: 214
Events since last DAE Display: 0 Suppression rate: 81%
A Last Last Total Date of Symptom String information:
C Date System Events Dump Abend Reason Module CSECT
_ 05/01/26 S0W1 7 04/28/26 S0EC6 055B0718 BPXINPVT BPXFSMNT
_ 05/01/26 S0W1 66 04/28/26 S00C4 00000004 BPXINPVT BPXVOTHD
_ 05/01/26 S0W1 10 04/28/26 S00C4 00000004 BPXINPVT BPXVOTHD
_ 05/01/26 S0W1 53 04/28/26 S00C4 00000004 BPXINPVT BPXVOTHD
There is a record for S0EC6 055B0718, and several records for S00C4 00000004. The detailed symptom string for these S00C4 abends are similar but different. The abend occurred at a different place, so the symptom string is slightly different.
Configuring DAE
You configure DAE through parmlib members ADYSETXX
On my system member ADYSET01 stops DAE, so I get a dump for every abend.
In member ADYSET00 I have
SVCDUMP(MATCH,SUPPRESSALL,UPDATE,NOTIFY(3,30)),
Where
- MATCH: Specifies that DAE is to compare the symptoms from the current memory dump to those that have already been recorded in the DAE data set. (Coding MATCH does not indicate that DAE suppresses duplicate memory dumps or update the DAE data set.)
- SUPPRESSALL: Specifies that duplicate memory dumps are to be suppressed when all criteria for matching and suppressing memory dumps are met except for the VRADAE key.
- Update: Specifies that the DAE data set is to be updated with the results of matching.
- Notify: This sends an internal signal to system applications listening, that a dump has occurred. The default is three memory dumps in 30 minutes for a particular symptom string.
How am I meant to know what’s happened?
As well as DAE processing, a record is written to EREP/LOGREC. EREP is a repository for hardware errors, and software errors.
For example, formatting the dataset, gave me
TYPE: SOFTWARE RECORD REPORT: SOFTWARE EDIT REPORT DAY.YEAR
(PROGRAM INTERRUPT) REPORT DATE: 121.26
...
SEARCH ARGUMENT ABSTRACT
PIDS/5752SCPX1 RIDS/BPXINPVT#L RIDS/BPXVOTHD AB/S00C4 PRCS/00000004 REGS/C082C
RIDS/BPXMIPCE#R
SYMPTOM DESCRIPTION
------- -----------
PIDS/5752SCPX1 PROGRAM ID: 5752SCPX1
RIDS/BPXINPVT#L LOAD MODULE NAME: BPXINPVT
RIDS/BPXVOTHD CSECT NAME: BPXVOTHD
AB/S00C4 SYSTEM ABEND CODE: 00C4
PRCS/00000004 ABEND REASON CODE: 00000004
REGS/C082C REGISTER/PSW DIFFERENCE FOR R0C:-082C
RIDS/BPXMIPCE#R RECOVERY ROUTINE CSECT NAME: BPXMIPCE
...
TIME OF ERROR INFORMATION
PSW: 07047001 80000000 00000000 20E1F2F4
INSTRUCTION LENGTH: 04 INTERRUPT CODE: 0004
FAILING INSTRUCTION TEXT: 00175045 00005049 01B0E368
TRANSLATION EXCEPTION ADDRESS: 00000000_01B6A404
BREAKING EVENT ADDRESS: 00000000_20E1F280
AR/GR 0-1 FFF00001/00000000_00000097 00000000/00000051_01B6AC90
AR/GR 2-3 00000001/00000000_0239A000 00000000/00000051_02FCFA00
...
HOME ASID: 0010 PRIMARY ASID: 0010 SECONDARY ASID: 0056
PKM: 8040 AX: 0001 EAX: 0000
RECOVERY ROUTINE ACTION
THE RECOVERY ROUTINE RETRIED TO ADDRESS 20E1F7DA.
THE REQUESTED SVC DUMP WAS NOT TAKEN. THE DUMP WAS SUPPRESSED BY DAE.
NO LOCKS WERE REQUESTED TO BE FREED.
THE SDWA WAS REQUESTED TO BE FREED BEFORE RETRY.
The information is a superset of the information in DAE. You get the registers at the point of failure, and what recovery action was taken. (Programs can ignore some abends, or pass the decision to a program higher up the stack).
In this case I can see there was no dump taken because of DAE.
The information in the logrec record was enough for me to debug the program. I did not need an SDUMP.
ColinPaice 