* copyright (C) 1984-2019 merrill consultants dallas texas usa

VMACSMF storing the IFCID in SUBTYPE. If you used %READDB2 and

Jun 19, 2013 listed any numerics in IFCIDS= argument, NO OBSERVATIONS

were created in those T102Snnn datasets, because READDB2

incorrectly tested SUBTYPE instead of QWHSIID.

Note: If you use the EXITCICS SAS INFILE EXIT on z/OS

to decompress DB2 records as they are read, or if you

use IBM PGM=DSNTSMFD on z/OS to decompress before you

read them, the SUBTYPE value is correctly set to the

IFCID, which is why the SUBTYPE test usually worked!

Now, QWHSIID is tested in READDB2 to correct this error.

-Because IBM DB2 does not populate the IFCID value into

the SMF Header SUBTYPE Field, MXG logic in VMACSMF gets

IFCID from the DB2 Product Segment to put into SUBTYPE,

when the data is uncompressed or the CICSIFUE is used,

but with the internal MXGDECOM, the record has not yet

been decompressed when the SMF header is processed, so

the offset to the Product Segment is wrong and the IFCID

cannot be INPUT, and SUBTYPE is zero (except for IFCIDs

105,106,196,258 and maybe others that aren't compressed).

and so ANALID reports 102.000 for those IFCIDs that can't

be known due to comressed records with MXGDECOM.

-After Change 32.175, the DB2 Subsystem is ALWAYS decoded

in the SMF Header Processing; the SMF SUBSYS field is the

DB2 Subsystem, but that was not realized in 31.115.

This Change Text was revised AFTER CHANGE 32.175.

-However, with compressed records and the internal decomp,

the DB2 version will be blank, since that, like IFCID, is

only in the product segment.

-The UTILGETM/VMXGGETM utility to extracts records of each

ID and Subtype to create the SMFSMALL dataset or report

on SMF record counts will also lump all those ID=102

IFCIDs as SUBTYPE=0 with compressed/internal.

-EXITCICS and DSNTSMFD are members in MXG Source Library.

-VMACSMF was revised to bypass reading the Product Segment

with compressed data and the internal code is used.

Thanks to Paul Walters, Navy Federal Credit Union, USA.

====== Changes thru 31.114 were in MXG 31.03 dated Jun 17, 2013=========
Change 31.114 SMF70NCA, PERCENT*WHEN*CAPPING*DELAYED*WORK is added to

VMXG70PR ASUM70PR ASUM70LP ASUMCEC ASUMCELP datasets, and the

Jun 17, 2013 maximum Group MAX70NCA is added to ASUM70GL.

Thanks to Helene BARDON, Silica Production Information, FRANCE.

Change 31.113 Discovered during QA. TRNDing RMFINTRV created workload

VMXGRMFI variables in TRNDRMFI/MNTHRMFI datasets with suffixes TRN

Jun 17, 2013 RSP SWP for TSO1/TSO2/TSO3 that should not have been

created nor kept, as the actual TSOn TRAN RESP SWAP were

created. Because of the original 8-byte variable name

restriction, and because you can specify Workload prefix

with five characters, the OTHER multi-period workload

variable names ARE suffixed with TRN RSP and SWP, which

is a regrettable inconsistency in MXG names that we'll

have to live with. Also, workload variables that are

used only for normalizing (suffix CNT) are not kept.
Change 31.112 MXG support for ODS graphics executes PROC TEMPLATE when

FORMATS FORMATS is run (during install of a new MXG Version) and

GRAFWRKX the MXGSTYL1 style is stored in the //LIBRARY DD/LIBNAME

MXGSTYL1 and VMXGINIT adds LIBRARY to the ODS Path so that style,

VMXGINIT used initially in GRAFWRKX, can be retrieved and used.

Jun 16, 2013 GRAFWRKX comments show how to use the ODS graphics code.

Change 31.111 -RMF III Enhancements and Notes

ASMRMFV -RMF III SVP (Service Policy) tables over 32K in length

SENDVSAM are now supported. Prior versions of ASMRMFV would skip

VMACRMFV the entire SVP table if it exceeded 32K because that was

Jun 15, 2013 beyond the maximum output LRECL possible. Now when

needed this table is segmented and output to RMFBSAM as

several records. This was a long standing documented

ASMRMFV program restriction that is now lifted. This is

the last RMF Monitor III table that had a size support

restriction. Users with Service Policies less than or

equal to 32K in size will have a single SVP record

output to RMFBSAM as before.

-VMACRMFV has been upgraded to accept segmented SVP table

records when present as input in the RMFBSAM file; these

variables are now RETAINed from the header and are kept

in ZRBSVPC ZRBSVPG ZRBSVPR ZRBSVPW ZRBSVPZ datasets:

SVPDSP SVPIDD SVPIDN SVPIDU SVPIPU

SVPNSP SVPSNA SVPTDI SVPTIB SVPTPA

-Note: Users most likely to be affected by this change

are those with large number of Service Classes and/or

Report Classes in their WLM Service Policy. Their

ZRBSVPP, ZRBSVPW, ZRBSVPC, ZRBSVPZ, ZRBSVPR, and ZRBSVPG

data sets will now be populated in the RMF III PDB if

the SVP table is selected in ASMRMFV processing.

-When the POLICY parameter is in effect (NOPOLICY is the

default), a second RMFV024I message will now appear in

ASMRMFV SYSPRINT showing the Name and Description of the

Service Definition. This is in addition to the Name and

Description of the Service Policy already displayed.

-Also when the POLICY parameter is in effect, the

RMFV027I message in ASMRMFV SYSPRINT will now display

the total length of the SVP table. When this value

exceeds 32760 then the RMFBSAM output will be segmented

into several records instead of a single record to avoid

skipping the table. The segmentation behavior occurs

regardless of whether POLICY or NOPOLICY is in effect.

-The Service Policy itself is now counted as an entry in

the ENTRIES COUNT column for SVP tables. This is added

to the total number of Workloads, Service Classes,

Service Class Periods, Report Classes, and Resource

Groups defined in the Service Policy already included in

that value.

-ASMRMFV prologue source documentation has been updated

to remove documentation on the skipping of the SVP table

if greater than 32K in size.

-Minor corrections made to SENDVSAM sample JCL member

used to send RMF III VSAM file to MXG support.

-REQUIREMENT: In order to receive these improvements the

current ASMRMFV utility program from this MXG change

must be installed. See MXG SOURCLIB member JCLASM3 for

sample JCL for the assembly and link-edit install steps.

Thanks to Warren Cravey, FMR Corporation, USA

Change 31.110 CICS/TS 5.1 only, MNSEGCL=5 (RESOURCE class FILE segment)

VMAC110 ID=110 Subtype 1 INPUT STATEMENT EXCEEDED RECORD LENGTH

Jun 13, 2013 because 16 bytes were inserted but not protected in MXG.

New variables

FCFIXCTM='FILE*EXCLUSIVE WAIT*DURATION'

FCFIXCCN='FILE*EXCLUSIVE WAIT*COUNT'

FCFIVSTM='FILE*VSAM STRING*WAIT*DURATION'

FCFIVSCN='FILE*VSAM STRING*WAIT*COUNT'

are now created in CICSRDFI. Fortunately (for ME!), only

leading edge CICS sites have enabled the RESOURCE class

segments, so the exposure is (hopefully) small for most.

Thanks to Tom Buie, Southern California Edison, USA.

Change 31.109 Documentation only, to create an uncompressed record for

VMAC110 technical support. The comment block is revised to show

Jun 11, 2013 how to print a hex dump of a single uncompressed record.
Change 31.108 DB2STAT1 variables QXSTCWLP/QXSTCW?R/QXSTCWLM/QXSTCWLD

VMACDB2 were not deaccumulated, were incorrectly kept in DB2STAT0

Jun 7, 2013 and QISTWMXU should not have been deaccumulated. Real

data with non-zero variables is required to validate if

fields are accumulated or not; the DSECTS are frequently

insufficient to clearly identify field contents.

Thanks to Giuseppe Giacomodonato, EPV Technologies, ITALY.
Change 31.107 These TYPE120 variables are now correctly labeled:

VMAC120 SM1209DD='ENCLAVE*ZIIP*ELIGIBLE*ON CP*CPU TIME'

Jun 7, 2013 SM1209DE='ENCLAVE*ZIIP*ELIGIBLE*CPU TIME'

SM1209DF='ENCLAVE*ZIIP*TIME*ON ZIIP'

Thanks to Mark Wittie, Fidelity Investments, USA.,
Change 31.106 -Variable TYPETASK in dataset TYPE6 was inconsistent for

VGETJESN PSF and Printway records, and did not identify Printway

VMAC6 Basic from Extended mode. Now, TYPETASK and SUBSYS6 will

BUILD005 contain 'PSF ' for PSF, and 'TCP ' for Printway Basic or

BUIL3005 'TCPE' for Printway Extended format records.

Jun 4, 2013 -BUILDPDB logic revised so these TYPE6 TYPETASK values are

May 14, 2014 NOT accidentally used for PDB.JOBS. HOWEVER: this can

cause TYPETASK to be blank in PDB.JOBS for any job for

which ONLY the PRINT record was found. And that will not

likely happen in full BUILDPDB production with the SPIN

library populated. It is when you have not populated the

SPIN library, and read "today's" SMF, it will have those

type 6 record for jobs that executed prior to today, and

that is the primary source of jobs with only TYPE6 obs.

Once you've populated the SPIN library, there should be

very few cases of only a type 6 record so TYPETASK should

normally be populated in production BUILDPDBs

This paragraph revised May 14, 2014.

-IP PrintWay fills in the page count (variable PAGECNT,

SMF6PGE) when it uses the direct sockets protocol to send

data to the printer and ONLY IF "Record pages printed for

accounting" option is selected in the Printer Inventory.

Thanks to Jennifer D. Ayers, West Virginia State Government, USA.

Thanks to Jeff Fracas, WIPRO, USA.

Change 31.105 ASMTAPEE ML-51 update corrects MXGC010E error message.

ASMTAPEE Note that if the HSC exit is not installed, STKX=NO must

Jun 1, 2013 be specified as a PARM when ASMTAPEE is executed.

//MXGTMNT EXEC PGM=MXGTMNT,PARM='STKX=NO'

Thanks to Scott Barry, SBBworks, Inc., USA.
Change 31.104 Change 30.250 was incorrectly made in ANALCAPD. The code

ANALCAPD should have had PDB=&PDBMXG, in the Macro Definition, but

ANAL72GO this discovery caused examination and correction of these

ANALAVAI other members. Many cases of lower case variable names

ANALBLSR were upper cased for consistency.

ANALCISH

ANALDB2R

ANALDBTR

ANALDMON

ANALNPMR

ANALRAID

ANALWHO

May 31, 2013

Thanks to Randy Shumate, Reed-Elsevier, USA.

Change 31.103 Variable QACCN is now kept in TMMQQAA dataset.

VMACTMMQ

May 30, 2013

Thanks to Homayoun Riazi, UHC, USA.

Change 31.102 IMS 13 inserted new DLRAZAAP field in the 07 Log Record

VMACIMS nearly at the front, that was overlooked, and caused all

VMACIMSA subsequent variables to be invalid, but error messages of

May 28, 2013 invalid data were only rarely printed.

Thanks to Daniel Erikols, Handelsbanken, SWEDEN.
Change 31.101 PDB.JOBS new variables HICPUPGM and HICPUPCT were wrong,

BUILD005 as the job-level values were taken incorrectly from the

BUIL3005 last step. They are now populated from the TYPE30_5 job

May 23, 2013 record, (using SPIN30_5 names HICPUPG5/HICPUPC5 because

they come from multiple records). With this change,

HICPUPCT is the maximum value of any step or interval

record, and HICPUPGM is the corresponding program that

recorded that HICPUPCT.

Note: These variables were one of the last of the late

Bernie Pierce's implementations, and they exist so that

IBM might be able identify single threaded programs.

With all processor vendors acknowledging that processor

speed is now limited by physics, and unlikely to be

increased significantly, applications that are single

threaded need to be identified for possible revision so

they don't become bottlenecks in the future, and these

variables provide a POSSIBLE candidate identification.

addition. Many IBM customers provided data, and I was

contacted because the MXG program, ASMRMFV, was flagged

as likely single-threaded, and it is! That program

decompresses the RMF III monitor data, and we had

previously examined with STROBE, which reported that

96% of the CPU time was inside the IBM-provided code

ERB3RDEC that does the decompression, which kicked this

ball back in to IBM field of play!

Thanks to Scott Barry, SBBWorks, Inc., USA.

Change 31.100 Member ASUMUOW executed twice because lines 214 following

ASUMUOW were duplicates of the first 214 lines, in MXG 30.06 thru

May 23, 2013 MXG 31.02.

Thanks to Mike George, Northern Territory Government, AUSTRALIA.

Change 31.099 Unused Change Number
Change 31.098 Support for local CICS variables TUXTRAP and UMBPROG.

IMACICV3

IMACICV4

UTILEXCL

VMAC110

May 22, 2013

Thanks to Jim Polleti, Edward Jones, USA.
Change 31.097 -CICS Statistics SMF ID=110 subtypes 2-5 contain multiple

VMAC110 segments per record, each identified by its "STID" value,

May 14, 2013 which controls which dataset is output, but MXG didn't

identify which SUBTYPEs had which STIDS to create which

datasets. These comments in VMAC110 are updated to list

CICS Statistics Subtypes 3-5 that contain STIDs 121-129:

CICS RECORD SUBTYPES AND CORRESPONDING MXG DATASETS CREATED:

0=JOURNAL SEGMENT

CICSJOUR (DEFAULT IF UNKNOWN JOURNAL).

CICSSAP (IF IMACICSA ENABLED FOR SAP).

CICSSMED (IF IMACICSM ENABLED FOR SHAREDMED).

1=MONITOR/TRANSACTION

MNSEGCL=1 DICTIONARY RECORD: CICSDICT ONLY IN UTILEXCL

MNSEGCL=2 CICSACCT (ZERO OBS ALWAYS, WAS PRE ESA)

MNSEGCL=3 PERFORMANCE CLASS: CICSBAD, CICSTRAN

MNSEGCL=4 EXCEPTION: CICSEXCE

MNSEGCL=5 RESOURCE: CICSRDS,CICSRDFI,CICSRDQU

MNSEGCL=6 IDENTITY: CICSIDNT,CICSIDND

2=STATISTICS: ALL OTHER STIDS, ALL OTHER CICXXXXX DATASETS

3=TS DATA SHARING STATS: CICXQ1,CICXQ2,CICXQ3

STID: 121 122 123

4=CF DATA TABLE STATS: CICFS6D,CICSF7D,CICFS8D,CICFS9D

STID: 126 127 128 129

5=NAMED COUNTER STATS: CICNS4D,CICNS5D

STID: 124 125

-TYPE110 processing always examines subtypes 2-5 for all

possible STIDs, which control which dataset is output,

but the example in comments in CICINTRV to build from raw

SMF selected only subtype 2 in its IMACFILE tailoring, so

that example would not populate those nine datasets. That

example now selects SUBTYPE GE 2.

Thanks to Richard Schwartz, IBM Global Services, USA.

Change 31.096 Example summarizes ENTIREX data to 15 minute intervals.

ASUMENTX

May 11, 2013
Change 31.095 Variable CA_XIIS /* GET CACHE DATA ACCESS XI'S */ now

ANALRMFR includes R744CXFW /* XI WRITE */ to match IBM RMF report.

May 11, 2013

Thanks to Marvin Silverman, ???, USA.

Change 31.094 BUILDPDB/BUILDPD3 enhancement: create your own accounting

BUILD005 variables for PDB.JOBS/PDB.STEPS/PDB.SMFINTRV, especially

BUIL3005 useful when converting from MICS to MXG. While IMACACCT

EXPDBACC allows you to create account variables, and the _KTY30Ux

VMXGINIT tailoring macros, placed in member IMACKEEP, will add

May 11, 2013 them to the TYPE30xx datasets, this new EXPDBACC exit

lets you create additional account variables for BUILDPDB

and, similar to other PDB "ADDxxxx" macro variables, use

%LET ADDACCT= ACCTBITS ACCTNo1-ACCTNo5;

and your new variables will be kept in the PDB datasets.

Thanks to Randall Schlueter, First Data, USA.
Change 31.093 -The DB2 "INVALID" IFCID=369 ID=100 SUBTYPE=5 DB2STAT5 SMF

CLEARDB2 record, cited and purportedly protected by Change 31.088,

READDB2 is more accurately described (in fairness to IBM DB2) as

VMACDB2 "UNDOCUMENTED" or "UNEXPECTED", but not INVALID, although

May 14, 2013 it caused MXG to take a floating point exception or INPUT

STATEMENT EXCEEDED error that ABENDED the JOB. (Older MXG

Versions print the 'CALL HOME" message 1440 times per DB2

Subsystem, but fortunately, with no actual impact on the

other existing DB2 datasets).
The UNEXPECTED subtype 5 record is now diagnosed as a

startup record, written when IFCID 0369 is enabled, and

it contains only two segments (PROD and OFF3691), with

only QW0369ST (enabled datetimestamp) populated, so MXG

sets QW0369CN='ENABLED' (or, for a record with QW0369SP

populated, MXG sets QW0369CN='DISABLED'). Since DB2STAT5

is a per-connection-type statistics, I presumed all five

segments would be present, but now test for QWHSNSDA=2 to

identify this record.
There was one record written with impossible DB2TCBTM of

152 minutes for a 1 minute interval for BATCH connection,

not possible, but the peak sum of DB2TCBTM recorded 38

minutes on a machine with XX CP engines.

Support for Subtype 5 was not in the original 31.02; it

was at least 5 minutes after I posted the GA announcement

before Joe Babcock sent me the CALL HOME message about

the new DB2 subtype, and Change 31.081 updated VMACDB2 to

create DB2STAT5, since it tested fine with a day's data

(other than the separate internal invalid data issue)

and because, without that updated VMACDB2, there would be

1440 "CALL HOME" log messages per DB2 subsystem printed

each day While in no way impacting the actual execution,

those messages can ONLY waste user's time in asking about

them, so the updated VMACDB2 at 31.081 was moved into the

redated MXG 31.02 on May 5. Unfortunately, by then, the

short record had been found, Change 31.088 detected and

protected, but the VMACDB2 in 31.02 of May 5 was still at

Change 31.081. This change, now does recognize and read

and output these records to DB2STAT5 without error.

-READDB2 is updated to create DB2STAT5 when STATISTICS or

IFCID=369 is requested. DB2STAT5 is NOT combined into

DB2STATS because STAT5 is per connection-type metrics.

-CLEARDB2 is updated to add DB2ST5 tokens to be redefined.

Change 31.092 Reserved Change Number.

May 8, 2013

Change 31.091 Instructions for updating IMACSMFF to add the description

FORMATS for user SMF records for the ANALID report and SMFRECNT

May 7, 2013 dataset didn't state that the syntax that is required is

'III.SSS' (III=ID with SSS=SUBTYPE) if the record has the

a subtype in the SMF header, or 'III.000' if not.
Change 31.090 Default LENUSRCH=32 was left behind from debugging that

IMACICUS should be LENUSRCH=0; this could cause EXCLUDED field

May 7, 2013 error messages even though you had tailored IMACEXCL.

Thanks to Scott Barry, SBBWorks, Inc., USA.

Change 31.089 Installation JCL example had inconsistent DSNAMEs for the

JCLINSTT TERSED dataset.

May 7, 2013

Thanks to Donald Williams, UNC Health Care, USA.

====== Changes thru 31.088 were in MXG 31.02 dated May 5, 2013=========
Change 31.088 New DB2 V10 ID=100 SUBTYPE=5 DB2STAT5 INPUT EXCEEDED

VMACDB2 error because of an UNDOCUMENTED short record with only

May 4, 2013 two segments. See Change 31.093 for the correction.

The original May 4 change to VMACDB2 detected and deleted

records with large OFF3692 offset values. The DB2STAT5

data is the DB2 IFCID=0369 and must be enabled to create

the exposure. See Change 31.093.
Change 31.087 MXG 31.02. Debugging PROC PRINT of DATA=MGTMSVL was left

TYPETMS5 after it helped diagnose Change 31.078. Now removed, but

May 4, 2013 it could create a massive print output dataset.

Thanks to Scott Barry, SBBWorks, Inc., USA.

Change 31.086 -SMF Audit Report Enhanced for DB2 SMF to report if the

ANALID ACCUMAC ZPARM option was enabled (NON-ZERO VALUE combines

VMACID multiple 101-1 records into one Rollup record with all

VMACSMF detail summarized). Variable DB2PARTY='R' in DB2ACCT

May 4, 2013 flags the summary records, which can be confusing if you

Jun 26, 2013 do not know what to look for, and ACCUMAC=10 is default.

A column is added and new FOOTNOTE2 explains. Note that

if ACCUMAC is on and off, statistics for both conditions

are reported.

-CICS 110.2 records are never compressed but MXG didn't

select only subtype 1 records, so the report could have

incorrectly indicated compressed for subtype 2 records.

-Jun 26, 2013: Comments revised in IMACSMFF.
Change 31.085 Format $MGCICDS for SMF/SMT/SMS CICS Statistics values

FORMATS 11/12/13x are now replaced in CICS/TS 5.1:

May 3, 2013 '11'X='11X:GCDSA' /* WAS ESDSA PRE 5.1.0 */

'12'X='12X:GUDSA' /* WAS ERDSA PRE 5.1.0 */

'13'X='13X:GSDSA' /* WAS ETDSA PRE 5.1.0 */

Thanks to Don Deese, (CPExpert), Computer Management Sciences, USA.

Change 31.084 MXG 31.02, RMF III INPUT STATEMENT EXCEEDED if an LPAR

VMACRMFV had no engines; the SKIP logic to skip the nonexistent

May 3, 2013 segments was incorrect with unexpected zero engines.

Thanks to James Sterling, DST Systems, USA.

Change 31.083 Reserved Change Number.

May 10, 2013

Change 31.082 -ASCII only. Character variable SMFDSAIN with $HEX format

VMAC110 must be input as $CHAR instead of $EBCDIC when MXG runs

Apr 30, 2013 on ASCII. Two of three incorrect instances corrected.

Thanks to Don Deese, (CPExpert), Computer Management Sciences, USA.

Thanks to Glen Bowman, Wakefern Food Corp.
Change 31.081 Support for DB2 V10 ID=100 SUBTYPE 5 Statistics Record,

EXDB2ST5 IFCID=369, created when you enable that trace record, now

IMACDB2 creates new DB2STAT5 dataset with LOTS of CPU and WAIT

VMACDB2 times (ALL QBAC and QWAX variables) by connection type,

VMXGINIT with these connection types defined:

Apr 30, 2013 MASS - IMS attach

SASS - CICS attach

RRSAF - RRSAF attach

UTILITY - Utility

BATCH - Batch

DIST - DDF connection

From IBM:

1.Statistics collection will become enabled when both

IFCID 369 and IFCID 3 is enabled on the system.

2.All counters will be reset to zeroes when DB2 is

restarted

3.Statistics are aggregated by connection type. If no

agents for that connection type have executed since

the 369 collection in enabled, no data will be

externalized for that connection type.

4.QWACPCNT indicates the number of transactions

aggregated for a given connection type.

The MXG - ERROR: UNEXPECTED DB2 SUBTYPE. CALL HOME.

shows how long ago this code to detect new subtypes was

added to MXG: BEFORE THERE WAS AN INTERNET! The message

now references [e-mail].

There is no direct impact on ITRM by the creation of the

new DB2STAT5 dataset. When DB2STAT4/DB2ST225 datasets

were created MXG did impact ITRM, with NOT SORTED errors

(that were immediately corrected with a one line insert),

but that was because those datasets were merged into the

PDB.DB2STATS where sorts had to match; this new DB2STAT5

dataset is NOT combined into the interval PDB.DB2STATS,

at least not yet and not until it's errors are fixed,