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

Thanks to Martyn Stevens, Barclays Bank, ENGLAND.
Change 07.085 VM/XA IUCV/VMCF fields had two labels reversed; PLSVSEVM

VMACVMXA should be "GOOD FROM" and PLSVSTVM should be "GOOD TO".

May 24, 1989 The following table may make IUCV/VMCF analysis easier.
--------TO--------- --FROM--
Service Destination GOOD BAD GOOD
IUCV: BLKIO PLSISTBL PLSISUBL PLSISEBL

CCS PLSISTCC PLSISUCC PLSISECC

MSG PLSISTM PLSISUM PLSISEM

MSGALL PLSISTMA PLSISUMA PLSISEMA

MONITOR PLSISTMO PLSISUMO PLSISEMO

RPI PLSISTRA PLSISUTA PLSISETA

SIGNAL PLSISTSI PLSISUSI PLSISESI

CP*SYSTEM SERVICES PLSISTVM PLSISUVM PLSISEVM

VMCF: A Virtual Machine PLSVSTVM PLSVSUVM PLSVSEVM
Thanks to Rob Owens, SAS Institute Europe, GERMANY.
Change 07.084 The 1984 MXG Guide, page 213, shows the IBM CICS tables

MXG Guide that need to be assembled to create type 110 CMF SMF data

May 24, 1989 records. The syntax CLASS=(PERFORM,ACCOUNT,EXCEPTION) is

not acceptable, apparently, for CICS 1.7. A site reported

IBM support said a separate DFHTYPE=RECORD statement is

needed in the MCT for each Class of data monitored. Also,

the example on page 213 should have suggested CONV=YES be

specified, so that each CICS interaction/conversation's

response time is measured.

Thanks to Bill Fox, National Futures Association, USA.

Change 07.083 RMFINTRV variables SIO74CNT,DEVACTTM,DEVCONTM,DEVDISTM, &

RMFINTRV DEVPNDTM were documented in the MXG Supplement to contain

May 24, 1989 only DASD device data, but RMFINTRV was never updated to

select only DASD type 74 records. (Note that even after

this change, that total SIO counts are still carried in

the SIO78CNT variable). Insert new line 072010:

IF SUBSTR(PUT(UCBTYPE,HEX8.),5,2)='20';

Thanks to Cathy Librachi, Aurora Health, USA.

Change 07.082 CICSTRAN variables MAXTASK and SHRTSTOR were not correct,

VMAC110 because once either condition was found in a transaction,

May 24, 1989 the rest of the transactions in that physical SMF record

were also flagged "Y", because the two variables were not

reset/initialized.

Insert two new lines 058410 and 069310:

ELSE MAXTASK=' ';

Insert two new lines 058510 and 069410:

ELSE SHRTSTOR=' ';

Thanks to Dennis Mahbubani, Aetna Casualty, USA.

Change 07.081 TYPE72MN (MVS/ESA-Only RMF Monitor III records) are all

VMAC7072 trashed because MONIDLS, MONPGIN and MOSLOT should have

May 9, 1989 all been PIB4. (this was a finger fault; the SKIP logic

expected 4-byte counters!)

Thanks to Don Friesen, B.C. Systems, CANADA.
Change 07.080 These sample CICS SAS/GRAPH routines had not been updated

GRAFCICS for SAS 5.18. All GOUT= were changed to PDB.GRAFCICS and

May 9, 1989 the final data step deleted. After execution, the graphs

can be replayed with PROC GREPLAY IGOUT=PDB.GRAFCICS;

Thanks to Bill Salassi, BHP Systems, AUSTRALIA.
Change 07.079 DB2ACCT and DB2STAT0 variable QTXAPREC was decoded into a

VMACDB2 set of eight new variables: QTXABLLM QTXAIOLM QTXAINLM

May 9, 1989 QTXASALM QTXASELM and QTXASPLM for ease in reporting DB2.
Change 07.078 IBM added four undocumented bytes to the end of RMF 77

VMAC77 record in MVS 3.1.0e (and, yes, it is a lower case "e" in

May 4, 1989 the MVSLEVEL variable!). MXG did not use "skip" logic in

type 77 processing, and the TYPE77 data set is trashed.

Insert two new lines after line 018800:

018810 SKIP=LENEDSS-156;

018820 IF SKIP GT 0 THEN INPUT +SKIP @;

Thanks to Carol Rogers, K-Mart Apparel, USA.

Change 07.077 If you used %VMXGSUM with MAX= but without MIN=, wrong

VMXGSUM values were calculated. This did not affect any MXG code,

May 4, 1989 as all MXG uses of %VMXGSUM specified both MAX= and MIN=.

Delete line 031500 %LET LASTX = &NUMMIN;

Change 07.076 Landmark Monitor system record daccumulation (MONISYST)

TYPEMONI is wrong. The revised de-accumulation algorithm added in

May 4, 1989 MXG 6.6 to process concatenated files always fails. This

ONLY affects the MONISYST data, and not the MONITASK data

set. (That it took until now to be noticed suggests that

few users take advantage of the MONISYST dataset!). The

correction is to change lines 084600 thru 088900. Each

line is now of the form Hnn = Variable;

Each line must be changed to read Hnn=Variable+Hnn;

For example, after the change:

084600 H1 =CPUTCTBM+H1 ;

...

089000 H45 =WTTSIOTM+H45;

Thanks to Joseph J. Faska, Depository Trust, USA.

Change 07.075 Processing IMS 1.3 Log data is NOT possible with the MXG

TYPEIMS1 algorithms for IMS 2.1 and IMS 2.2. These power of these

VMACIMS1 algorithms (which recognize each IMS transaction instead

May 4, 1989 of each program schedule) depend on several fields that

simply do not exist in the archaic 1.3 release:

-LINENR is not in type 35 log record

-LTERM is not in type 31O log record

-ARRVTIME is always nulls in 03P record

-DLRTOKEN only exists in the 07/08 records

The ability to recognize multiple transactions within a

program schedule is absolutely dependent on these data.

Thus MXG cannot claim to support transaction measurement

for IMS 1.3. We have restored MXG Version 5.5 IMS logic

in new members TYPEIMS1/VMACIMS1 for IMS 1.3 sites. That

code made no attempt to recognize multiple transactions

per program schedule, but instead matched IBM's DFSILTA0

results to produce one observation in IMSTRAN for each

program schedule. Since only 7 sites have requested our

IMS 1.3 processing, we can only offer this restoration.
Change 07.074 Unused change number.
Change 07.073 In TYPE26J2, TYPETASK may contain JOB0, TSU0, or STC0 as

VMAC26J2 IBM prepared for JESNR greater than 9999 in the JCTJOBID

May 4, 1989 field. This change reads only the first three bytes of

JCTJOBID into TYPETASK, but keeps its length at four to

avoid conflict with other data sets. Insert new line

013311 after 013310: TYPETASK=' '; /*4 blanks*/

In line 013400 Change $CHAR4. to $CHAR3.

In line 013500 Change +4 to +5

Thanks to Carol Harper, E G & G Idaho, USA.
Change 07.072 In TYPE90, variable DETAIL was not reset for each class

VMAC90 os SMF recording (JOB, TSU, STC). Now it is by insertion

May 4, 1989 of ELSE DETAIL=' '; after DETAIL='Y';

Thanks to Diane Eppestine, Southwestern Bell Telephone, USA.

Change 07.071 In ROSCOE 5.6 RTM the number of complexity levels is set

TYPEROSJ by ROSCOE sysgen parameter RTMCMARK and the number of

VMACROSC time levels is sysgened by RTMTMARK. The actual number

VMACSMF of buckets in the RTM record is one less than the number

May 3, 1989 of levels. MXG expects three complexity level buckets and

five time level buckets (assumed RTMCMARK=4,RTMMARK=6).

You will encounter a STOPOVER abend if you sysgen less.

This change now uses information in the record to read in

only as many fields as exist. Many lines were changed.
An unrelated change to ROSCOE processing was also made:

member TYPEROSJ has been deleted, and the code in VMACSMF

that set OFFROSC was removed, because ROSCOE journal and

ROSCOE SMF records can be processed with member TYPEROSC

using a DDNAME of SMF for either dataset. TYPEROSJ was

only for ROSCOE 5.4 and earlier, and caused confusion.

Thanks to Art A. Brax, Jefferson Co. Colorado School District, USA.

Thanks to Sandy Recchio, Chase Lincoln First Bank, USA.

Change 07.070 This enhancement to dataset activity analysis added VSAM

ANALDSET TYPE64 records, new fields (eg. OPENTIME) from TYPE1415,

Apr 30, 1989 and enhanced the description of the job. If you want to

know which program in which job opened which file for

how long and for how much I/O, this is the place to go.
Change 07.069 This enhancement to DB2PM-like reporting adds many small

ANALDB2R changes as a result of more detailed validation with DB2

Apr 30, 1989 Version 2.1 data and report formats, and adds additional

reports. This revision now invokes %VMXGSUM to summarize

which reduced the size from 12,000 lines to only 7,150!
Change 07.068 The VTOC processing code only expected 16 extents, since

VMXGVTOC the code was written before there was VSAM. With this

Apr 30, 1989 enhancement, MXG now expects up to 128 extents to fully

support VSAM data spaces in a VTOC.

Change 07.067 The long promised MXG Tape Mount Monitor, MXGTMNT, lives!

ASMTMNT Member ASMTMNT contains the assembly source for the load

EXTMNTMN module MXGTMNT and the JCL to create the proclib member

EXTMNTSW MXGTMNT. MXGTMNT executes in its own address space and it

IMACTMNT creates an SMF record for every tape mount with the start

TYPETMNT and end timestamps, duration of the mount, and the job

VMACTMNT causing the mount. MXGTMNT also creates a record if a

Apr 30, 1989 "tape swap" occurs as the result of a permanent error.

The installation instructions for the tape mount monitor

are contained in comments at the beginning of ASMTMNT.

Please read completely before executing. Member IMACTMNT

contains the SMF record ID that you chose in ASMTMNT.

Member TYPETMNT/VMACTMNT creates two data sets from the

SMF record: TYPETMNT contains mount information and

TYPETSWP contains swap events. The monitor has been

tested under MVS/370 and MVS/XA, and should execute under

MVS/ESA as well. Members ASMTPMON and ASMTP370 have been

replaced by ASMTMNT.

Change 07.066 Exits EXPDBWEK and EXPDBMON have been added to the PDB

EXPDBMON WEEKBLD and MONTHBLD members so that users who have added

EXPDBWEK additional SMF records to BUILDPDB/3 (using the MXG Exit

MONTHBLD Facility, page 139 in the MXG Supplement) can copy those

WEEKBLD additional data sets into their MONTH and WEEK PDBs.

Apr 29, 1989

Change 07.065 DB2 Trace SMF type 102 record support has been expanded

FORMATS and further validated and revised with Release 2.1 data.

IMAC102 The subtype 106 (System Parameter) record (in T102S106)

VMAC102 was completely re-formatted by IBM in 2.1, requiring a

Apr 28, 1989 rewrite. Subtypes 67, 98, 99, 101, 112, 122, and 123

May 9, 1989 have now been tested with 2.1 data. An entire new class,

Audit (subtypes/IFCIDs 140-145) is now supported, though

only subtypes 140 and 141 were available for testing.

Logic now tests QWHSNSDA (in 2.1) or record length (1.3)

to determine the actual number of triplets, which avoided

a STOPOVER conditions. Subtype 53 and 58 were rewritten

to decode the SQLCA and to avoid STOPOVER conditions. New

macros are now defined in IMAC102 for the Audit classes.

Some subtypes that were not in their correct trace class

groups were moved to where they belong, and the subtype

to trace class documentation (in IMAC102) revised. This

was a significant change to type 102 processing for 2.1.
Note: All 120+ T102Snnn datasets are always created by

MXG, but only those "enabled" by IMAC102 will actually

have observations. However, SAS requires virtual storage

for each data set at compile time. With only a 4MB region

it was necessary to specify SAS Option BLKSIZE=8192 for

the TYPE102 program to execute.

Thanks to Stan Stoots, Virginia Power, USA.

Thanks to Stan Majka, Virginia Power, USA.

Thanks to Peter Gallinari, Reader's Digest Association, USA.
Change 07.064 CICSEXCE variables TSWAITTM,CN,MSWAITTM,CN,VSWAITTM,CN

VMAC110 were not initialized and could be carried forward into

Apr 27, 1989 a subsequent exception observation. All six variables

are now initialized to missing after line 101600.

Thanks to Dee Ramon, Mutual of America, USA.
Change 07.063 Trending RMFINTRV from PDBs built before MXG 6.6 causes

TRNDRMFI Variable Not Found condition (which cannot be solved by

Apr 27, 1989 the NOVNFERR SAS option), because of the new OTH6-OTH0

variables added in MXG 6.6 to RMFINTRV that are not in

RMFINTRV built by prior versions of MXG. If you want to

use previously built PDB librarys as input for trending

you must create 70 new lines after 014500 (one line for

each of the 14 variables in each of the 5 OTHn groups).

The 70 variable names are in lines 003900-004100 and in

lines 005700-006800. Each new line will be:

IF variable=. THEN variable=.;

For example, the first line will be:

IF OTH6ACTV=. then OTH6ACTV=.;

Thanks to Georg Simon, Federal Reserve Bank of Philadelphia, USA.

Change 07.062 NETSPY records caused zerodivide condition because the

VMACNSPY denominator in lines 041200 thru 041500 should have been

Apr 26, 1989 TOTRSPNO instead of TRANSNO.

Thanks to ???, KF Data, SWEDEN.

Change 07.061 TYPE74 variables AVGPNCHA, AVGPNCUB, AVGPNDEV are labeled

VMAC74 as containing milliseconds of pending time for channel,

Apr 25, 1989 control unit, and device, but erroneously are in seconds.

This change corrects the calculation so that the values

are in milliseconds and agree with the labels. In lines

039700 thru 039900, change the lines which now read:

AVGPNxxx=DURATM*PCTPNxxx/(100*SSCHSAMP);

to read:

AVGPNxxx=10*DURATM*PCTPNxxx/SSCHSAMP;

Thanks to Ron Higgins, American Honda, USA.

Change 07.060 NetView Release 2 writes "Hardware Monitor External Log

VMAC37 Record" as SMF type 37 Records and their contents are

Apr 27, 1989 described in NetView Administration Reference SC30-3361-2

(Appendix A pages 97ff). This change adds support for

three new segments (LPDA-2, LAN, and Generic Event) into

the TYPE37 data set.

Change 07.059 NetView Release 2 writes "Session Monitor External Log

VMAC39 Record" as SMF type 39 Records and their contents are

Apr 27, 1989 described in NetView Administration Reference SC30-3361-2

(Appendix A pages 163ff). This change adds support for

new subtypes 7 and 8. Subtype 7 in TYPE39_7, Init Failure

is identical to pre-existing TYPE39_6 dataset. Subtype 8

in TYPE39_8, Storage and Event contains 44 new variables.

(VM NetView also writes SMF format data to its external

file which should be processable with TYPE39, but this

has not yet verified).

Thanks to Russell Brooks, James River Corporation, USA.
Change 07.058 JES3 BUILDPD3 prints TYPE30_5 (twice!) due to debugging

BUILDPD3 code (lines 013900 and 014300) which are now deleted.

Apr 20, 1989 Two PROC CONTENTS of the PDB and WORK ddnames are deleted

(lines 059100 thru 059400) because ANALCONT provides the

same report. (Some sites did not want the contents, but

I find them worthwhile in tracing problems.)

Thanks to George Scott, Rockwell International Corporation, USA.
Change 07.057 VM/XA Monitor records have been altered by PTF VM35971

VMACVMXA and VM35321 which are now supported by MXG. VM35971 adds

Apr 20, 1989 VMDCTMIG (Pages Migrated from ESTORE to DASD), VMDCTXWT

(Pages written from main to ESTORE) and VMDCTXRD (pages

read into main from ESTORE) to the VXUSEACT and VXUSEATE

detail data sets. MXG adds the new variables to the MXG

created VXBYUSR, VXSUMUSR and VMXAINTV datasets, storing

the values as rates per second as are other paging data.

PTF VM35321 adds VMDCPUAD (CPU Address) to the VXSCLADL,

VXSCLDDL and VXSCLAEL Schedule records. This now permits

MXG to properly de-accumulate those data (prior to this

fix, the schedule data was unusable because VMDCPUAD was

needed to de-accumulate those data).

Thanks to Richard Steele, Louisville Gas and Electric, USA.

Change 07.056 TPX SMF records read directly from active VSAM data set

VMACTPX cause STOPOVER because MXG was incomplete. In line 032000

Apr 19, 1989 change TPXOFF=TPXOFF-3; to TPXOFF=TPXOFF-3+OFFSMF;

There is no problem processing TPX records from a normal

SMF dumped dataset.

Thanks to Bob Lemaster, Central Bank of the South, USA.

Change 07.055 Support for the TLMS (Tape Library Management System) CA

TYPETLMS product is added by this user contribution. The code was

Apr 19, 1989 restructured but has yet been tested with real data in

its revised form. The TLMS Catalog is read to create an

observation for each tape volume.

Thanks to Wing Louie, Chase Manhattan Bank CDC, USA.

Change 07.054 NPM Type 28 Subtype 5C (Dynamic Reconfigure Notification)

VMAC28 caused STOPOVER because DRNTNAM, Target Resource Name, is

Apr 18, 1989 not always present (this is not documented by IBM).

Insert new line 164310 after line 164300:

164310 @; IF LENTOF GE 19 THEN INPUT

to conditionally input DRNTNAM when it is actually there!

Thanks to Larry Doub, R J R, USA.
Change 07.053 DB2 Version 2.1 causes "VMACDB2 DATA. UNEXPECTED ID=100

VMACDB2 NRINSD=14 and NRIFCID=13" messages on the SAS log. This

Apr 18, 1989 is not really a problem to worry about. The NRINSD and

NRIFCID count the number of Destination Data Sections and

Instrumentation Data Sections in the record, which count

records written for the monitor, and do not relate to the

DB2 system itself. As a result, MXG kept only the first

three NRINSD destination names and only the first five

IFCIDs (see page 203 of the MXG Supplement for the list

of variables that are kept). I don't think the counters

are useful, and if they were, it should be done with an

additional data set. Until someone demonstrates a need to

use this obscure data, I have re-worded the text of the

message: "MXG KEEPS THE FIRST THREE/FIVE IFCID/INSD, BUT

YOUR DATA HAD n. THIS IS NOT USUALLY IMPORTANT". If I

knew for sure it was always 14/13 I might suppress the

note. This change also corrected some minor logic errors

in deciding when to print the message: Line 094800 should

GT 3 instead of GT 5, and -4 should be -2, and 094900 is

PIB2. instead of $CHAR4.

Thanks to Siegfried Trantes, Gothaer Versicherungsbank VVAG, GERMANY.
Change 07.052 SYNCSORT SMF records read from "active" VSAM SMF file

VMACSYNC caused STOPOVER abend (but worked fine from dumped SMF).

Apr 18, 1989 Remove +OFFSMF from line 026200.

Thanks to Dave Greene, Kwasha Lipton, USA.

Change 07.051 RMF Monitor III VSAM data sets are supported by this

EXZRB... major user contribution. (In MVS/ESA, a very small

IMACZRB part of the Monitor III data is written as a subtype

TYPEZRB of the Type 72 record and supported in TYPE72MN data

VMACZRB set introduced in MXG Version 6.6). The workload

ZRBJCL delay monitor data is far more extensive that that,

ZRBRPTn and this extensive set of code processes those VSAM

ZRBWORK RMF III data into several data set names. Although

Apr 12, 1989 the "ERB" acronym is the more appropriate name for

these data, I renamed Dick's use of ERB to ZRB as I

expect this code will be short lived as MVS/ESA takes

over in the future.

Thanks to Dick Whiting, Precision Castparts, USA.
Change 07.050 Concatenated VM/Monitor data contained invalid values

VMACVMON for the first interval after each START event, because

Apr 12, 1989 MXG incorrectly outputted all intervals, when it was

intended to only output beginning with the second. In

all twenty occurrences of IF INTRVCNT GT 0 THEN ....

must be changed to IF INTRVCNT GT 1 THEN

Thanks to Rodney L. Reisch, General Electric Plastics, USA.
Change 07.049 The references to _CICOTHR for the ddname of the CICS

ANALCICS ACCT, EXCE, and YSTM data sets should have been changed

Apr 11, 1989 to _CICACCT.CICSACCT, _CICEXCE.CICSEXCE and

_CICYSTM.CICSYSTM to be consistent with IMACCICS.

Thanks to Mark Hutchinson, Atlantic States Bank, USA.
Change 07.048 IMS Log Record decoding corrections.

VMACIMS a.After 024700 (the @; after DEST $CHAR8.), insert line

Apr 9, 1989 IF _IMSVERS GE 2.2 THEN INPUT MSGIHSQN $CHAR8. @;

IMS 2.2 added this eight byte field and MXG missed it.

Without the fix, IMS 2.2 abended with STOPOVER (though

several IMS sites looked at the hex dumps, inserted an

+8 before the SIZELU6 variable to align and circumvent!)

b.Change 027400 from LOC=MSGPRFLL; to LOC=MSGPRFLL-3;

so that MSGXDLEN is correctly read in and used for the

MSGLEN, MSGSZIN, and MSGSZOUT variables.

c.Both references to unused unkept COMMFLAG were deleted.

d.Insert after line 043500 (between DLRTOKEN and PSTTSKID

@; IF _IMSVERS GE 2.1 THEN INPUT

e.Insert after line 056010 (between MSGDRRN and DLRTOKEN

@; IF _IMSVERS GE 2.1 THEN INPUT

Items d. and e. prevent STOPOVER with IMS 1.3 data.

f.Any time stamp was set missing if PTIME was exactly zero

(midnight). The PTIME/PDATE INPUT statement uses the ??

syntax (suppresses hex dump and error message and makes

value missing for invalid PD. data). Thus PTIME could be

zero, causing MXG to not calculate a datetime variable.

All statements testing PTIME GT 0 were changed to test

PTIME GT . (missing value).

g.Log type 35x records were not always output, causing the

ENQTIME to be missing which caused an out-of-sort-order

failure to merge ENQTIME with other transaction records.

Heuristic logic testing ENQFLAG and FLAG2 bits determine

if a 35x record is for input, output, message-switch, or

program-to-program, brute-force determined by finding a

combination, then looking at a detail trace to determine

the proper IMS35x data set to output. Values previously

unseen were skipped and not output. Now, unknown values

are printed on the SAS log. (If you should encounter a

new combination, call and we'll discuss how you can dump

the IMS log records to figure out which IMS35x dataset it

should be put in - I sure wish there were a better way!)

This change restructured logic in IMACIMS to remove the

RETURN statements and replaced IF's with ELSE IF's.

Thanks to Harry Olschewski, DVG Kiel, GERMANY.

Thanks to Siegfried Trantes, Gothaer Versicherungsbank VVAG, GERMANY.

Change 07.047 RACF Auditor reports from SMF type 80 record exposures.

ANALAUDT 1.In member ANALAUDT, references to _RACF must be changed

FORMATS to _AUDT (which sets the DDNAME containing your TYPE80

Apr 9, 1989 data set, defined in IMACANAL).

2.The remainder of this change corrects only the values of

variables ALLOW and INTENT in SUF/INSUF Auth. reports.

There is no other impact if you do not put this fix on.

This report decoded ALLOW and INTENT from the wrong byte

of RACFDATA, in a block of code in the wrong location,

and were not RETAINed for multi-observation RACF events!

a.ANALAUDT changes:

-Add ALLOW INTENT to RETAIN at line 009400.

-Block COPY lines 016300 thru 018100 (from /* RESOURCE