* copyright (C) 1984-2019 merrill consultants dallas texas usa
Yüklə 28,67 Mb.
|
|
was added to the ID= statement. -New CICLGG Logstream Global Statistics added. Thanks to Bruce Widlund, Merrill Consultants, USA.
ANALRMFR (Percent Online) always missing in PDB.TYPE70, which then VMAC7072 caused ANALRMFR CPU Activity Report to have blank values. Jan 19, 2007 Fortunately, the other variables in PDB.TYPE70 were okay; this is NOW the final "SPLIT70" correction, and is needed for z/OS 1.7 and later, if you have Dedicated z/OS CPUs. -Unrelated, accidentally observed and corrected, the CPU Number printed by ANALRMFR on the CPU Activity Report was often incorrect. Thanks to Bob Keller, Safeway, USA. Change 24.275 ERROR.MORE THAN 255 STRUCTURES was due to an archaic test VMAC74 for 255; the arrays had been increased to 1024, but the Jan 18, 2007 test value and error message text were not revised. Thanks to Bill McDonald, KCC, USA. Change 24.274 Utility for ASCII execution to read SMF data via ftp and ASCISMFC create a local disk file of only selected SMF records. Jan 17, 2007 Thanks to Bill McDonald, KCC, USA. Change 24.273 Extraneous '60'x character in column 1 of line 140 caused WEEKBLDD a syntax error, now removed. Jan 17, 2007 Thanks to Lisa Lawver, Land's End, USA. Change 24.272 -Enhancement to ASMRMFV RMF III adds ENC Extension feature ASMRMFV and ASM symbolics to let users tailor the ASMRMFV default VMACRMFV parameters were added. Jan 18, 2007 -Enhancement to VMACRMFV to decode the ENC Extension, which adds these new variables to the ZRBENC dataset: ENCCNM ='SERVICE*CLASS*NAME' ENCCDE ='SERVICE*CLASS*DESCRIPTION' ENCCWN ='ASSOCIATED*WORKLOAD*NAME' ENCCRN ='ASSOCIATED*RESOURCE*GROUP' ENCCPO ='OFFSET TO*SERVICE*CLASS*PERIOD ENTRY' ENCCPN ='NUMBER OF*SERVICE*CLASS*PERIODS' ENCCGI ='RESOURCE*GROUP*INDEX*IN ENCRG' ENCCWI ='WORKLOAD*INDEX*IN ENCWD' ENCCRC ='PERIODS*WITH*RESPONSE*TIME GOAL' ENCRNM ='REPORT*CLASS*NAME' ENCRDE ='REPORT*CLASS*DESCRIPTION' Thanks to Jerry Urbaniak, Acxiom CDC, USA.
VMACRMFV RMF III dataset ZRBDVT to match IBM response metrics, and Jan 17, 2007 variable SWPODLTM is now kept. Thanks to Jerry Urbaniak, Acxiom CDC, USA. Change 24.270 Variable GDESDP2 was incorrectly input @11 instead of @9, VMACQACS causing a value of 16,448 for available processors. Jan 16, 2007 Thanks to Robert Gilbert, Fortis Bank, BELGIUM. Change 24.269 This is the schematic of zIIP CPU time variables in the ADOC30 TYPE30xx, JOBS, STEPS datasets. zIIP CPU time is always Jan 16, 2007 in separate variables that are never included in the old "CPU" variables that, then and now, contain ONLY the time spent on "CP Engines". CPUZIPTM /*SMF30_TIME_ON_ZIIP*/ CPUDZITM /*SMF30_DEP_ENCLAVE_TIME_ON_ZIIP*/ CPUEZITM /*SMF30_IND_ENCLAVE_TIME_ON_ZIIP*/ CPUZIETM /*SMF30_ELIGIBLE*TIME_ZIIP_ON_CP*/ CPUDZETM /*SMF30_DEP_ENCLAVE_TIME_ZIIP_ON_CP*/ CPUEZETM /*SMF30_IND_ENCLAVE_TIME_ZIIP_ON_CP*/
CPUDZQTM /*SMF30_DEP_ENCLAVE_TIME_ZIIP_QUAL*/ CPU TIME ON ZIIP ENGINES CPU TIME ON CP ENGINES "Actual" "Eligible" |--------CPUZIPTM---------| |--------CPUZIETM---------| |--CPUDZITM--|--CPUEZITM--| |--CPUDZETM--|--CPUEZETM--| (DEP) (IND) (DEP) (IND)
(Sum of DEP Actual and Eligible) |-------CPUDZQTM----------| |--CPUDZITM--|--CPUDZETM--| "Qualified - Independent Enclave" (Sum of IND Actual and Eligible) |-------CPUEZQTM----------| |--CPUEZITM--|--CPUEZETM--| See also MXG Technical Note 31 in Newsletter FORTY-NINE, "zIIP CPU Time Comparisons between TYPE72GO and TYPE30_V". Thanks to Bob Keller, Safeway, USA.
VMXG70PR in ASUMCEC and ASUM70LP datasets; now, using the new Jan 15, 2007 MINLONG= argument added to VMXGSUM, they are.
VMXGSUM min/max output that are 8-bytes long, like the existing Jan 15, 2007 SUMLONG= argument.
TYPEIMS7 dataset IMS07 NOT FOUND, because _IMSWORK was not used in Jan 14, 2007 TYPEIMS7. Now, the syntax is consistent. Thanks to Erling Andersen, SMT Data, DENMARK. Change 24.265 MXG 24.04-24.09. PDB.ASUMTAPE lost many observations, ASUMTAPE only for TMNTEXIT='IBM'. Change 24.109 added incorrect Jan 14, 2007 logic to propagate READTIME into 2nd-vol (HAVEMNT=501A) events that are always missed by TMNT. Propagation is now corrected, but there can ALWAYS be missing values in many of the variables in PDB.ASUMTAPE, depending on which events were found for this mount (which combines TMNT Mount, Syslog Mount or Keep, SMF21 dismount events.) An output from PROC MEANS N DATA=PDB.ASUMTAPE shows: Variable N Implication ZDATE 4030 Total mount event obs created READTIME 1675 Mounts for jobs with a TMNT record TAPMNTTM 1328 Mounts with a TMNT record BYTES 3866 Mounts with matching TYPE21 TAPMTDTM 3615 Mounts with BEGTMNT/ENDTMNT
the TOTMNDTM duration was negative, SYLMTIME was wrong, and there were other defects. This happens when SYSLOG mount event was in yesterday's event for a long running task that's writing lots of datasets (TMS9 for each one). -When SYLMTIME is missing, EVENTIME=SYLKTIME-5 is now set with a 5 second adjustment; SYLKTIME can be fractions of a second later than TY21TIME, and this ensures TMNT will be seen before SYSL in the merge. -When TMS9 message was first, retained times were not clearedl Thanks to Doug Medland, IBM Canada, CANADA. Change 24.264 Support for CMRDETL (T6E) records for Mainview for CICS VMACMVCI Version 5.9.00 adds (COMPATIBLY) 132 new variables to the Jan 12, 2007 CMRDETL dataset.
UTILGETM select SMF data, you need to invoke %VMXGGETM (... ) ; Jan 11, 2007 as UTILGETM will not accept arguments. Thanks to Flavio Lima, Bank of America, USA. Change 24.262 If you want to know how many bytes of SMF data is written Example for each of your CICS regions, by subtype, this example: Jan 11, 2007 //SMF DD //SYSIN DD * %LET MACKEEP= MACRO _KCICTRN ) CICSHDR (KEEP=APPLID MEGABYTE SUBTYPE % _N110 ; %LET MAC110H= %QUOTE(MEGABYTE=LENGTH/1048576; OUTPUT CICSHDR;) ; %INCLUDE SOURCLIB(TYPE110); PROC FREQ DATA=CICSHDR; TABLES APPLID*SUBTYPE; WEIGHT MEGABYTE; TITLE SMF 110 MEGABYTES BY SUBTYPE FOR EACH APPLID; Thanks to Bruce Sloss, PNC, USA.
ANALCPUV investigation of IRD impact. Reads PDB.TYPE70 to create Jan 9, 2007 a temporary format $MGNRCPU with NRCPUS for each STARTIME interval, uses that format to add the variable NRCPUS to to each PDB.SMFINTRV observation, also finds each JOB's MINCPUTM and MAXCPUTM, used to calculate each interval's PCTOVRMN (Percent CPU TCB was above minimum recorded) and PCTBLOMX (Percent CPU TCB was below maximum recorded). The intent is to analyze benchmark data in which the same job is run multiple times on a system with wide range of IRD-controlled NRCPUS, to see if there is a measurable impact on recorded CPU TCB time due to IRD. It is theorized that the recorded CPU seconds should be smaller when NRCPUS is small, and larger when the NRCPUS is large, because of the "MP effect", partly because the SU_SEC used to calculate service units is NOT adjusted when IRD changes NRCPUS. This program is ready to test that theory, and will quantify the observed variability in CPU times, if any is observed. Let's discuss, before you run your benchmark.
DOCVER in the DOCVER documentation, because JOBS/STEPS/PRINT in QASAS DOCVER were all from the JES3 BUILDPD3, but SUBMUSER is a Jan 8, 2007 JES2-only variable. While I figure out how to change the descriptions in DOCVER, where same-named datasets with different variables are created by MXG, moving the JES3 BUILDPD3 ahead of the JES2 BUILDPDB in the QASAS QA job will cause DOCVER to contain the descriptions of the datasets built by the JES2 BUILDPDB.
VMACIPAC error, because the final field, IPPACCES was only 4 bytes Jan 5, 2007 while MXG expected 8. Both lengths are now protected. This change also added support for R6.3. Thanks to Jolene Halibry, Nationwide, USA.
VMACPROS of them!) were wrong because MXG's INPUT statement was Jan 5, 2007 off by one byte. Some character variables with hex data were not formatted but now are. Thanks to John Kim, ATCO I-Tek, USA.
VMACBETA variables were added compatibly for tested subtypes: Jan 4, 2007 Subtype 0: BETA0 dataset: BETACTYP='CLEANUP*TYPE' BETARCME='ARCHIVE*MEDIA*TYPE' BETARETA='ARCHIVE*RETENTION*PERIOD' Subtype 1: BETA1 dataset: no changes. New subtypes 10, 51, and 52 are documented but will only be supported when they are available for validation. Thanks to Engelbert Smets, Provinzial, GERMANY. Change 24.256 New variables added to IFCID 226 and 227 are supported. VMAC102 Variables QW0226PN/QW0227PN are now character zeros, as FORMATS new QW0226PG/QW0227PG variables now contain page number, Jan 3, 2007 new QW0226FG/QW0227FG contain table space type, which is decoded by new MGD226S format. Thanks to Bill Schray, IBM Global Services, USA. Thanks to Ted Blank, IBM Global Services, USA.
VMACSMF the log at the end of SMF input are enhanced with elapsed Dec 29, 2006 duration and the read rate of the input data:
*** MXG 24.09 SUCCESSFULLY COMPLETED READING SMF.*** 26750 LOGICAL SMF RECORDS WERE READ. THE SMF FILE CONTAINED 325,284,835 BYTES, WHICH IS 310MB. MINIMUM SMF RECORD TIMESTAMP WAS 02SEP2004:09:00:00.04. MAXIMUM SMF RECORD TIMESTAMP WAS 02SEP2004:09:34:57.14. MXG FINISHED READING SMF FILE AT 29DEC2006:15:54:40.06. ELAPSED TIME TO READ SMF FILE 0:00:08.69. SMF READ RATE PER ELAPSED TIME 35MB/SEC. ******************************************************* Thanks to Chuck Hopf, Bank of America, USA. Change 24.254 -Support for the rest of the Candle/IBM optional CICS data IMACICC5 segments are added to UTILEXCL: IMACICC6 IMACICC5 CANSUPRN SUPRA*TOTAL*REQUESTS IMACICC7 IMACICC6 CANSUPRT SUPRA*TOTAL*DURATION IMACICC8 IMACICC7 CANDCOMN DATACOM*TOTAL*REQUESTS IMACICC9 IMACICC8 CANDCOMT DATACOM*TOTAL*REQUESTS IMACICE3 IMACICC9 CANRES01 VSAM*TOTAL*REQUESTS IMACICE4 IMACICE3 CANIDMSN IDMS*TOTAL*REQUESTS UTILEXCL IMACICE4 CANIDMSN IDMS*TOTAL*REQUESTS VMAC110 -Validation discovered that the CANGMTOF GMT offset was Dec 24, 2006 never correct, but IMACICC1 is now revised to correctly IMACICC1 decode the partial TOD stamp into the offset duration. Dec 28, 2006 Technical Note ons tailoring CICS IMACICxx members: When you use UTILEXCL to create IMACEXCL (RECOMMENDED!!), you can remove Comment Blocks in ALL of your IMACICxx members, because IMACEXCL's code only %INCLUDEs the IMACs needed for each "DO GROUP" found in your PDB.CICSDICT. Each execution of _BLDDICT appends new CICS dictionary records found in SMF to the old PDB.CICSDICT dataset, which is then read by _BLDEXCL to create the IMACEXCL code that will read your SMF 110 CICSTRAN records. The KEEP= list in IMACEXCL has only those variables that exist in your PDB.CICSDICT records. Or, the KEEP= list could have ALL of the hundreds of optional variables and now defunct variables, if you have a old CICS dictionary record from a test system long ago in your PDB.CICSDICT.
IMACEXCL based only on today's CICS dictionary records. And/or you can delete unwanted APPLIDs from PDB.CICSDICT before you run the _BLDEXCL. Thanks to Richard Schwartz, State Street Bank, USA.
READDB2 due to a typo. Dec 23, 2006 Thanks to R. Narruli, DST Systems, USA. Change 24.252 The eight-byte EXCP count in SMF33EXX that replaces the VMAC33 four-byte SMF33EXP count is now INPUT when it exists. Dec 23, 2006 Thanks to Andreas von Imhof, Rabobank, THE NETHERLANDS. ====== Changes thru 24.251 were in MXG 24.09 dated Dec 20, 2006========= Change 24.251 Protection for Mainview MQ records RTIN='26'x that didn't VMACBBMQ contain the ISHD header segment that MXG thought would Dec 20, 2006 always be there. Variables DURATM ENTC and GMTOFF might be missing when there is no ISHD header segment. Thanks to Stuart Wildey, Morgan Stanley, ENGLAND.
VMACMWNT expected MMDDYY requires you to set new &DATEFMT macro VMXGINIT %LET DATEFMT=DDMMYY; Dec 19, 2006 %INCLUDE SOURCLIB(TYPSMWNT); if the "MWA EXPORT DD/MM/YY" text in the first record has that format. -The INPUT of SOFTWARE and RELEASE was revised to protect for a Logfile name that contains blanks. Thanks to Dominik Covens, KBC, BELGIUM.
VMACHSM sort macros for HSMWWFSR and HSMWWVOL datasets, probably Dec 18, 2006 because the _Sxxxx Product Sort macro had replaced the early "DIFF" nomenclature, and _DIFFHSM was overlooked. Now, those two datasets are included when _DIFFHSM is invoked, but the preferred name to use in your EXPDBOUT member is to have a _Sxxxx statement for each VMACxxxx that you %INCLUDed in your EXPDBINC tailoring member. Then, MXG is responsible for any deaccumulation as well as adding any new datasets into your PDB library as part of your BUILDPDB job. Thanks to Dwain Majak, B, B, and T, USA. Change 24.248 Enhancement to the "BUILD PDB EXAMPLE", BLDSMPDB, adds BLDSMPDB optional processing of DCOLLECT and TMS/CA-1 records into Dec 18, 2006 the daily PDB library, so they can also then be created Dec 20, 2006 in your weekly and monthly PDB libraries. New argments: DCOLLECT=DCOLLECT - read INFILE DCOLLECT output PDB =dsname - alloc FILENAME DCOLLECT to dsname all output to PDB TMC =TMC - read INFILE TMC output PDB dsname - allocates FILENAME TMC to dsname Additionally, SMF data weekly/monthly processing can be weekly/wtd, and monthly/mtd. Weekly/Monthly will copy all, or only selected, datasets. Exits were added for flexibility during weekly/monthly/trend processing. Thanks to Chuck Hopf, Bank of America, USA.
VMACCMA -VMACCMA User SMF record now have the standard, expected VMACQACS macro token names and definitions for the syntax for VMACTNG "Single Infile, Multiple Datasets Per Product" data: VMACTMO2 For each output dataset: VMACCIMS MACRO _Vdddddd Dec 19, 2006 KEEP= list of variables % MACRO _Wdddddd &Wdddddd..DATASET % MACRO _Kdddddd % Output all datasets for the product: MACRO _VARXXXX _Wdddddd
(LABEL='dddddd: description' _Vdddddd _Kdddddd ) ....
_Wdddddd repeat for each product dataset. %
macro token names, as above for "Single Infile, Multiple Datasets Per Product" data. -VMACCIMS User log record updated with standard, expected macro token names, as above for "Single Infile, Multiple Datasets Per Product" data.
"MACRO" missing from each statement; it would have failed if it had been used!
now, the VMACQACS AS/400 macro names _VQAPxxx _CQAPxxx cannot be changed without serious exposure to existing, fine running jobs. For the record, for these datasets from "MULTIPLE INFILES, ONE OUTPUT PER INFILE" data: For each output dataset: MACRO _TQAPxxx DATA
_VQAPxxx _CQAPxxx
% MACRO _WQAPddd &Wdddddd..DATASET % MACRO _Kdddddd % Where the _VQAPxxx was already defined as: MACRO _VQAPxxx _Wdddddd
(LABEL='QAPxxx: description' KEEP= list of variables ) %
that's the lesser of causing production job failures! Thanks to Erling Andersen, SMT Data A/S, DENMARK. Change 24.246 REGION=0M added to MXGSASV9/MXGSASV8 JCL PROC examples to MXGSASV8 protect sites that did not specify a REGION on their JOB MXGSASV9 card. (The MXG JCL examples do show REGION=0M on JOB.) Dec 15, 2006 -When REGION=0M is specified on the JOB JCL statement, the job gets your installation default REGION=0M size, often 100MB-300MB, which is quite sufficient for most MXG jobs, for every step in the job. -With REGION=xxxM value specified on the JOB card, all of the steps get that xxxM REGION size, even if there is a larger or smaller REGION= value on a STEP card. -If the JOB does not have a REGION= parameter, the job and each step gets a different default region, of only about 40MB (9MB Private Area + 32MB Above the Line). While much of MXG 24.08 does run in a 40MB REGION, (including the JCLINSTL job that successfully created the MXG Format Library), the BUILDPDB job failed when run in only 40MB, with SAS FORMAT NOT FOUND errors (but each individual formats was there and usable by itself). The 40MB wasn't enough region for the "BUILDPDB big DATA step", which allocates virtual storage for all of the output buffers for all of the datasets to be created, and then loads all referenced formats into virtual storage. -The default BUILDPDB needs a 64MB REGION, generally, but may need 100MB+, if you have tailored your BUILDPDB to process additional SMF records. Thanks to Donald Likens, Combined Insurance, USA.
VMACRMFV now kept: Dec 18, 2006 CMRTM Command*Response*TIME DVTCUQTP Control Unit Queueing Time Previous DVTCUQTN Accum CU Wait for non-FICON devices DVTCUQTF Accum CU Wait for FICON devices. Thanks to Jerry Urbaniak, Acxiom CDC, USA.
ASUMDB2G is created from PDB.DB2ACCTG dataset by ASUMDB2G member. VMXGINIT Dec 13, 2006 Thanks to Hugh Lapham, Royal Canadian Mounted Police, CANADA
ASUMDB2A NORMAL was revised to include QWACRINV=4 thru 16 and 40 Dec 12, 2006 as NORMAL and all other QWACRINV values as ABNORMAL, to be consistent with the formatted values of QWACRINV in the MGDB2RC format. Thanks to Nigel D. Greenwood, EDS, ENGLAND. Change 24.242 Revisions to force TEMPxx macro variable explicitly to a VMXGINIT value of WORK, and revised setting of SASSWORK, etc., for VMXGSUM anticipated SAS/ITRM changes to support SAS V9 BI. Dec 12, 2006 Change 24.241 Keyword parameter WORK73 was accidentally typo/deleted in VMXGRMFI the macro definition, causing an error only if there were Dec 11, 2006 73 or more workload's defined. Thanks to Clayton Buck, UniGroup, USA. Change 24.240 -All durations were 1000 times too large; I assumed the VMACSYSI times were in 256*milliseconds, like most prior IMS data, Dec 7, 2006 but data and documentation show they are 256*microsecs, so all &PIB.4.3 informats were changed to &PIB.4.6. Thanks to Betra Reeves, Infocrossing, USA. Thanks to Joel Medberry, Infocrossing, USA. Change 24.239 MACRO _ROSCDDN has not been used since the &Pdddddd and VMACROSC &Wdddddd macro variables were defined, but comments in Dec 6, 2006 IMACROSC and VMACROSC were still present/confusing. To send all of the ROSCOE datasets to the //PDB DD, use %INCLUDE SOURCLIB(TYPSROSC) which will sort, remove any duplicates, and output them to //PDB. Thanks to Lori Martin, Lockheed Martin, USA.
VMACRMFV in the RMF III record, but are not documented as such. Dec 5, 2006 They are now corrected in their INPUT, and I have also made the assumption that these IFA time variables in the same segment are also in millisecs in the record, and are also corrected in their INPUT informat. ENCTIFAT ENCTIFCT ENCIFAT ENCIFCT Thanks to Brenda Rabinowitz, Merrill Lynch, USA. Change 24.237 Label for NRZIPCPU and NRIFAS in PDB.RMFINTRV had text of VMXGRMFI "IN THE BOX", but as RMFINTRV is a PER-SYSTEM dataset, Dec 1, 2006 the label is changed to "FOR THIS SYSTEM'. Thanks to Douglas C. Walker, Citigroup, USA. Change 24.236 Reserved Change. EXITCICS
VMACSMF Nov 30, 2006 Change 24.235 EJBCRECT was input twice, the second time where EJBREMCT UTILEXCL was located, so EJBCRECT was wrong and EJBREMCT did not Nov 30, 2006 exist when UTILEXCL was used to process CICS data. Thanks to Harald Seifert, HUK-COBURG, GERMANY. Change 24.234 If you specified %LET MACKEEP ahead of UTILBLDP, it may UTILBLDP be ignored if you are also adding records in a BUILDPDB Nov 21, 2006 process. This change puts your MACKEEP values inside of the MACKEEP being generated by UTILBLDP. Note however the error message that your use of MACKEEP here may defeat something UTILBLDP is trying to do so use it with caution. Thanks to Stan Dylnicki, Royal Bank of Canada, CANADA. Change 24.233 MXGERROR: More than 70 200 Byte Strings is circumvented VMXGSUM by increasing the MXG default to 99 200 byte strings for Yüklə 28,67 Mb. Dostları ilə paylaş:
|