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

command string to eliminate the ambiguity. I also noted

there are 8 undefined bytes at the end of the TELNET

record that I am investigating.

Thanks to Barry Pieper, Norwest, USA.

Change 12.040 The %VMXGSUM invocation requires the addition of

ASUMPRTR KEEPIN=STDUPLEX TMBUPLEX,

May 24, 1994 to eliminate the UNITIALIZED VARIABLE condition.

Thanks to Diane Eppestine, Southwestern Bell, USA.

Change 12.039 Member IEBUPDTE now contains both the IEBUPDTE.BAS BASIC

IEBUPDTE program and the IEBUPDTE.SAS program that will create a

May 24, 1994 separate file (named "member.SAS") for each PDS member

of the mainframe source library that was downloaded, as

described in MXG Newsletter TWENTY-FIVE. As any believer

knows, the SAS program was faster (26 minutes versus 66

minutes for MXG 11.11) than the BASIC program!

Thanks to Freddie Arie, Lone Star Gas, TEXAS.

Change 12.038 Support for BatchPipes/MVS APAR PN45746, which adds local

VMAC91 timestamps for interval begin/end, and for first/last

May 24, 1994 access for the input and output pipe connections.
Change 12.037 Support for AS/400 Transaction Summary Trace File QTRTSUM

ADOCQTRT is added by this user contribution, which has been tested

EXQTRTSU with OS/400 Version 2 Release 2. Instructions for this

IMACQTRT data source are in member ADOCQTRT.

TYPEQTRT

VMACQTRT

May 24, 1994

Thanks to Colin Adams, ENGLAND.

Change 12.036 APAR OW00484 adds Date of Open to TYPE1415 (finally!), so

ANALDSET now OPENTIME will be correct (prior to this APAR, I could

VMAC1415 only use the record' SMFDATE, or SMFDATE-1 if open time

May 24, 1994 was later than SMF time). Variable OPENTM is now created

with the correct duration of open (but only if the APAR

is installed - OPENTM will be missing if OPENTIME is not

based on a true Date of Open). Also, OPENTM was added

to the variables kept in ANALDSET analysis.

Change 12.035 The CICS Shutdown reports delivered in MXG 11.11 are now

ANALCISH revised so that report selection matches IBM's DFHSTUP

May 23, 1994 program (i.e., so that only EOD reports can be selected -

the original report printed everything, EOD, INT, etc, if

those records were found), and individual datasets are

now created for each report (instead of a single dataset

that could require massive WORK space), and the errors in

Change 12.005 are also corrected by this revision. The

comments in the new member describe the new parameters.

Thanks to Kusol Voratanitkitkul, Health Care Service Corp, USA.

Change 12.034 Support for MVS/ESA 5.1.0:

BUILDPDB -TYPE0 new variables

BUILDPD3 PRODNAME - SMF Product Name

BUILD003 SYSNAME - System Name from IEASYSxx member

EXTY44 SYSPLEX - Sysplex Name from COUPLExx member

EXTY72GO -TYPE6 new variable

EXTY7204 CUTSHEET - flag if Cut Sheet Printer

EXTY74CF -TYPE23 new variables

EXTY74ST SYSNAME - System Name from IEASYSxx member

FORMATS SYSPLEX - Sysplex Name from COUPLExx member

IMAC44 -TYPE30_4,TYPE30_V new variables

IMAC7072 IEFUSICH - Flag if IEFUSI changed Region Size

IMAC74 WLMNAME - Workload Name

IMACPDB RESGROUP- Resource Group Name

IMACWORK RPTCLASS- Reporting Class Name

MONTHBLD SRVCLASS- Service Class Name

RMFINTRV SYSNAME - System Name from IEASYSxx member

VMAC0 SYSPLEX - Sysplex Name from COUPLExx member

VMAC6 -TYPE30OM new variables

VMAC23 OMVSEXNP - OMVS Executed Program Name

VMAC30 OMVSOPP - OMVS Parent Process ID Number

VMAC44 OMVSOKR - OMVS I/O Blocks read for remote Socket

VMAC7072 OMVSOKW - OMVS I/O Blocks written for remote Socket

VMAC71 -TYPE30MU new dataset for Measured Usage - See 12.028

VMAC73 -TYPE32 new variables

VMAC74 SYSNAME - System Name from IEASYSxx member

VMAC79 SYSPLEX - Sysplex Name from COUPLExx member

VMAC89 -New type 44 Subsystem Modify Record creates TYPE44

WEEKBLD SMF44PRC - SS06 Proc Name

Feb 14, 1994 SMF44OPT - Initialization Options

May 23, 1994 -TYPE71 new variables

PGBKLPA - LPA Block Page Page-ins

PGBKSYSA - System Pageable Areas Block Page-ins

-TYPE72 new datasets:

TYPE72GO - Goal Mode (Equivalent of TYPE72)

This dataset is of major importance when Goal Mode

is used, as it replaces the TYPE72 data for resources

by Service Class. TYPE72GO is merged in with RMFINTRV

so that correct CPU utilization is constructed whether

you are in Goal Mode or Compatibility Mode.

There are five kinds of observations in TYPE72GO:

I. Address Space Service Class

(Criteria: R723CRCA='Y' AND NOT RPRTCLAS='Y').

=1. Response Time Goal Sub-Class

(Criteria: Has RTS Section)

RESOURCES + RESPONSE-DIST

=2. No Response Time Goal Sub-Class

(Criteria: Has no RTS Section)

RESOURCES

II. Transaction Service Class

=3. (Criteria: Has WRS Section).

RESPONSE-DIST + WM-STATES

III. Report Class

(Criteria: RPRTCLAS='Y').

=4. Address Space Sub-Class

(Criteria: R723CRCA='Y').

RESOURCES

=5. Transaction Sub-Class

(Criteria: R723CRCA NE 'Y').

Only TRAN count from resources.

Variable R723TYPE has value 1 thru 5 to identify what

kind of TYPE72GO observation you are dealing with.

Note that just like we had Report Performance Groups,

Goal Mode has Report Classes in addition to Service

class, and resources in a Report Class record have

already been reported in the task's Service Class.

One bright spot - work can only be reported in one

Report Class. The dim spot - there is no fall-thru

report class created if you define a report class,

so you can't sum Report Class data and be sure you

have all work in a report class, so MXG still must

use only Service Class (not Report Class) to build

the workload variables in RMFINTRV. Member IMACWORK

has been revised, and you must specify what Service

Class is what workload for Goal Mode, just like you

specify what Control Performance Group is what for

compatibility mode. MXG defaults do try to recognize

workload by your Workload Name. The observation is

a Report Class if RPRTCLAS='Y', or is a Service Class

if RPRTCLAS=' '.

TYPE7204 - Subtype 4

This is the RMF Monitor III data for Goal Mode, much

like TYPE72MN was for Compatibility Mode. Sampled

workload delay statistics and swap counts by all 17

swap reasons are captured for each period of each

service class monitored.

-TYPE73 new variable

SMF73CPD - Channel Path Description

-TYPE74 new datasets from new subtype 4:

TYPE74CF - Coupling Facility Utilization

This dataset is a mini-RMF for the Coupling Facility,

with:

-per-engine CF CPU utilization;

-Storage defined and free for Control, CF, and Dump

Space Storage, Dump Space defined and used,

-Path and subchannel use and contention counts, busy,

duration of delay for unsuccessful requests

-List of each structure and its size that is allocated

to this Coupling Facility. See TYPE74ST dataset.

TYPE74ST - Coupling Facility Structure Request Activity

For each Connected Structure to this CF, the request

activity is provided. Multiple observations in

TYPE74ST can be matched to their parent TYPE74CF

observation BY SYSTEM STARTIME R744FNAM. Contains:

-Requests, both SYNC and ASYNC, and separate durations

for service for SYNC and ASYNC, and for delays due to

queue delay, dump delay, high or low priority queue,

and min/max counts for hi and low priority queues and

dump serialization, plus lock contention counts!

-TYPE89 new record for Measured Usage Pricing was added to

BUILDPDB/BUILDPD3/etc. See Change 12.028.
Status:

Almost all records are now supported, and compatibility

mode data verified. Goal mode data has now been validated

and omissions in the first release were corrected. The

RMFINTRV member now processes either GOAL or COMPATIBLITY

mode type 72 data. Note that IBM thinks there may still

be some incorrect values in some of the data fields.

It's really nice to have new records to test with!!!!!

b. MVS/ESA 5.1 changes were compatibly made, so this

prerelease is required only to support new facilities,

i.e., CPU measurement of Goal Mode requires prerelease.

c. These new records have not been added to this PreRelease.

TYPE90 Subtype 10 coded, not yet described herein

TYPE90 Subtype 23 coded, not yet described herein

TYPE90 Subtype 24 WLM - not yet coded -

TYPE90 Subtype 25 coded, not yet described herein

TYPE92 - OMVS File System activity 11 subtypes not coded.

TYPE99 - WLM data - subtype 2 decoded in MXG 12.03.

Final notes: I spent a lot more time on the code than on this

documentation. Please feel free to fax/call if you see problems

with the data, but I feel real confident about the core of the

change.
I am extremely impressed with the design of the new Workload Manager,

and while migrating to Goal Mode may take a little planning, I am

convinced it is the righteous thing to do as soon as its ready;

it's architecture is clearly what we wanted from MVS all along!

See my note on this subject on page 2 of MXG Newsletter TWENTY-FIVE.

INCOMPATIBILITY NOTE: These IMACs were changed to add new

datasets and to support the Workload Manager:

IMAC7072 IMAC74 IMACWORK IMACPDB

If these members exist in your MXG tailoring library, you

MUST retrofit your changes to the new IMAC member, or you

will get "variable GOALMODE not found" errors in RMFINTRV

or "dataset not found" errors in the other members.
Change 12.033 DB2 Version 3.1 support was not completely correct in MXG

DIFFDB2 11.11. The Buffer Pool variables QBnCaaa in DB2ACCT are

EXDB2ACB all wrong (misaligned, because a new reserved field was

EXDB2STB not skipped), and in both DB2ACCT and DB2STATS, the new

IMACDB2 buffer pools (BPID =3 thru 49 for 4K pool, BPID=81-89 for

VMACDB2 32K pools) were not recognized. In addition, DB2STATS

May 23, 1994 did not correctly de-accumulate most of the variables new

in DB2 Version 3.1. If you have not exploited DB2 3.1,

and still have only BP0, BP1, BP2 and BP32 defined, you

can correct the DB2ACCT variables QBnCaaa by changing the

sixteen lines reading INPUT (QBnCGET QBnCSWS ...

to read INPUT (QBnCGET QBnRSVD QBnCSWS ..

However, if you have exploited DB2 3.1 and have defined

the new buffer pools, both DB2ACCT and DB2STATS will now

include the new buffer pools in the existing MXG variable

groups, as described below. Additionally, the interval

statistics for each buffer pool are now provided in the

new DB2STATB dataset. Furthermore, the detail activity

counts for each buffer pool used by each plan execution

are available in the new DB2ACCTB accounting detail,

Nov 25, 1996 revision: Original text stated:

but as DB2ACCTB could be very large, MXG, by default,

does not output any observations in DB2ACCTB (because

the OUTPUT statement is commented out in member

EXDB2ACB). If you actually need the details in

DB2ACCTB, all you need to do is to remove the comment

block, as described in the comments, in member

EXDB2ACB. In that member, you can also conditionally

execute the OUTPUT statement so that only a specific

buffer pool and/or a specific plan name are output.

(This paragraph was revised 22Sep94).

New text: Now DB2ACCTB by default DOES have observations

since it turned out it's not really that large, so you do

not have to do anything

To compatibly support DB2 3.1's up-to-60 buffer pools in

MXG's existing DB2ACCT and DB2STATS datasets and ANALDB2R

reports, I simply redefined the meaning of the four sets

of buffer pool variables that already existed (in DB2ACCT

the variables are QBnCaaa, in DB2STATS they are QBnTaaa,

where n=1,2,3,4). The QB1 and QB2 sets of variables

still are the counts for the (4K) BP0 and BP1 buffer

pools respectively, but the QB3 set of variables (instead

of just containing counts for the (4K) BP2 pool), now

contain the SUM of the activity for all of the OTHER 4K

Buffer Pools (BPID 2->49), and the QB4 set of variables

(instead of just containing counts for the (32K) BP32),

now contain the sum of activity for ALL of the 32K Buffer

Pools (BPID=80->89).

INCOMPATIBILITY NOTE: IMACDB2 was changed to add the new

datasets. If you have modified IMACDB2 into your MXG

tailoring library, you must retrofit your changes using

the IMACDB2 from this MXG library.

Thanks to Tuomo Rahko, KOP Kansallistieto, FINLAND.

Thanks to Waldemar Schneider, SAS Europe, GERMANY.

Change 12.032 Dataset T102S196 for DB2 type 102 IFCID 196 Lock Timeout

VMAC102 Details record, which identifies holder and wanter(s) of

May 18, 1994 lock requests that resulted in the timeout of a DB2 task,

is now populated by this change. T102S196 contains all

details on the lock request that timed out, and identity

details of the first three holders of an incompatible

lock (the first three agents that caused the timeout.)

Thanks to Majid Abai, Southern California Edison, USA.

Change 12.031 STK's ICEBERG dataset ICEBRGCH is trashed, because there

VMACICE must be a "+1" before "INTENCUR $CHAR1." so that MXG

May 18, 1994 is properly aligned with each data segment. Let me point

out that STK, just like IBM, measures disk capacity in

"Million Bytes" and not "Mega-Bytes"; MXG always converts

sizes in bytes to Mega-, or Giga- bytes, so a 3380 with

a DASD capacity of 630 "Million Bytes" can really only

store 601 true "Mega Bytes" of data!

Thanks to Diane Eppestine, Southwestern Bell, USA.
Change 12.030 All instances of INPUT XXXXXXXX MSEC8. were replaced by

DOC INPUT XXXXXXXX &PIB.8.6 ... @; XXXXXXXX=XXXXXXXX/4096;

May 18, 1994 because the MSEC8. format sets XXXXXXXX to missing if the

value is greater than 24 hours! While SAS is aware of

this defect, it will not be corrected until SAS Version 7

and I couldn't wait that long! The specific case that

precipitated the error was a CICS Shutdown record which

had total wait time of 25 hours, but all occurrences were

changed (311 times in 29 members) just to be sure. (Note

that IBM's DFHSTUP report did not handle the 25 hour time

either - it reported only 1 hour!).

NOTE: MSEC8. PIB8.6/4096 CPU TIMER UNITS

Change 12.029 Labels for variables DS4VTOCE and DS4VTOCI were reversed.

VMXGVTOF

May 18, 1994

Thanks to John Taylor, Newport Management Corporation, USA.

Change 12.028 Support for Measured Usage License Charges new SMF record

EXTY30MU type 89 (MVS/ESA versions 3.1.3, 4.2, 4.3, and 5.1) and

EXTY89 enhancements to type 30 (MVS/ESA version 5.1 and later).

FORMATS The new architecture is well described by IBM in manual

IMAC30 GC28-1098-00. The new TYPE89 dataset has one observation

IMAC89 for each interval (default, and maximum, one hour) for

TYPE89 each "registered product" that had usage during that

VMAC89 interval. TYPE89 data will be summarized by IBM's IFAURP

May 18, 1994 reporting program to calculate total hourly product usage

Jun 24, 1994 daily, and then the fourth-largest daily-hour during the

month is used as the basis for Measured Usage Charges.

When IBM releases IFAURP, MXG will replicate the report.

For MVS/ESA 5.1 and later, new dataset TYPE30MU will

report on each address space that uses any "registered

product". This preliminary change is based only on the

documentation - no records are yet available to test!

Change 12.027 Candle's OMEGAMON for CICS 550/551 user SMF record DSECT

VMACOMCI was wrong, affecting the contents of OMCISYST dataset.

May 17, 1994 New variable SMRECNT (count of transactions) is created,

and fields are now correctly input. Change these lines:

SMOASID &PIB.2. /*ADDRESS SPACE ID */

SMRES3 $EBCDIC6. /*RESERVED AREA */

SMOFLGS &PIB.4. /*SYSTEM FLAGS -- */

SMRES4 $EBCDIC3. /*RESERVED AREA */

SMCVTTZ &PIB.4. /*GREENWICH MEAN TIME OFFSET */

to read:

SMOASID &PIB.2. /*ADDRESS SPACE ID */

SMRES3 $EBCDIC2. /*RESERVED AREA */

SMOFLGS &PIB.4. /*SYSTEM FLAGS -- */

SMRECNT &PIB.4. /*output transaction rec count*/

SMCVTTZ &PIB.4. /*GREENWICH MEAN TIME OFFSET */

and add SMRECNT to KEEP= list for OMCISYST.

Thanks to Linda Berkley, Amdahl, USA.
Change 12.026 Jobs with JCL errors before execution (i.e., ABEND='JCL')

BUILDPDB were not output in PDB.JOBS; they are in PDB.NJEPURGE due

BUILD005 to Change 11.226. Find

May 17, 1994 ((INREASON=' ' OR INREASON='JR' OR INREASON='JT')

and change it to read

((INREASON='JR' OR INREASON='JT')

Thanks to Jack Hwang, Great American Reserve Insurance Co., USA.

Thanks to Jerry Byers, Great American Reserve Insurance Co., USA.

Change 12.025 TYPE26J2 INPUT STATEMENT EXCEEDED error due to truncated

VMAC26J2 type 26 SMF record; field SMF26NRA does not exist in the

May 17, 1994 record in at least one MVS/ESA 4.3 site (although it does

exist at other sites). To circumvent, replace two lines:

SMF26NRA &PIB.1.

@;

with these two lines:

@; IF SMF26LN8 GT 2 THEN INPUT SMF26NRA &PIB.1. @;

ELSE SMF26NRA=0;

The actual MXG change was more extensive.

Thanks to Diane Eppestine, Southwestern Bell, USA.

Change 12.024 MXG Tape Allocation and Mount monitor ABENDs were caused

ASMTAPES by a typo, wrong field was moved into LOCLINFO (which

May 17, 1994 caused INPUT STATEMENT EXCEEDED in TYPETMNT), and STIMER

loop was relocated to reduce CPU cost (ASMTAPES now takes

about twice the CPU time to monitor both tape mounts and

tape allocations as ASMTMNT to monitor just mounts).

-Find the two lines:

L R5,QBDFELMP-QDB(,R1) GET 1ST ELEMENT IN QUEUE

LTR R1,R1

and change the second line to read

LTR R5,R5

Be careful to change the correct LTR instruction, as

there is a valid LTR R1,R1 just three lines earlier.

This is the typo causing the SRB abends in the monitor.

-Find LOCKLOOP EQU *

Delete that STIMER, and then four lines later, find and

replace the line BNZ LOCKLOOP with these four lines:

BZ UNLOCKED

STIMER WAIT,BINTVL=THETIME

B LOCKLOOP

UNLOCKED EQU *

This change eliminated the high CPU utilization

-Find and delete the line:

MVC SRUSER,JCTUSER

Find the two lines:

SPACE 1

MVC SRRSTM,JMRENTRY READER START TIME

and insert a third line as shown below:

SPACE 1

MVC SRUSER,JMRUSEID LOCLINFO FIELD

MVC SRRSTM,JMRENTRY READER START TIME

Change SMFUSER DS CL7 to SMFUSER DS CL8

Change SRUSER DS CL7 to SRUSER DS CL8

This correction solved the TYPETMNT record error.

May 25, 1994 status:

The monitor now only occasionally fails, and we are

examining dumps of two allocation events that are the

cause - TMS initialization of tape volumes, and when an

ATL doesn't have a tape volume in the library. The

earlier concern for MIM and SMS see unwarranted, as the

monitor is now functioning in those environments.

Thanks to Bob Kinney, Kaiser Permanente, USA.

Change 12.023 CICS Statistics dataset CICLSRR index variables will be

VMAC110 wrong if all LSR pools do not have indexes; an LSR pool

May 15, 1994 without indexes will have index variables from the prior

LSR pool that did have indexes, because MXG did not

set missing the index variables when A08FLAGS NE 'Y'.

To correct, all index variables that were INPUT in the DO

group IF A08FLAGS='Y' THEN DO; are now set missing in

the immediately following ELSE DO; group.

Thanks to Anita A. Bradley, Virginia Power, USA.
Change 12.022 This utility to convert character variables containing

UTILCVRT Hex (Binary) data from ASCII back to EBCDIC was listed in

May 10, 1994 Newsletter TWENTY-FIVE, but did not exist in MXG 11.11;

it now exists and is self-documenting.

Thanks to Damian Stevenson, Policy Management Systems, AUSTRALIA.
Change 12.021 This utility to count CICS type 110 records by APPLID

UCICSCNT produces unclear counts, because there must be an

May 9, 1994 OUTPUT; statement just before the last END;

Thanks to Chris Powell, Vancouver Stock Exchange, CANADA.

Change 12.020 CICS Statistics dataset CICDBUSS variables STACTIME and

VMAC110 STADTIME were missing, because their INPUT should have

May 5, 1994 been TODSTAMP8. instead of MSEC8.
Change 12.019 Type 42 subtype 14 record (ADSM) offsets have now been

VMAC42 corrected by IBM, causing INVALID ADSM SECTION TRIPLET

May 4, 1994 message, and MXG typing errors caused incorrect values.

After first ELSE IF SUBTYPE=14 THEN DO; delete the line

OFFPROD=OFFPROD+4;

After the second ELSE IF SUBTYPE=14 THEN DO; change line

OFFADSM=OFFADSM+1+OFFSMF; to now read

OFFADSM=OFFADSM-3+OFFSMF;

In the INPUT statement for SUBTYPE=14, the final period

is missing from the input format for YYYY (&NUM.4 must be

&NUM.4.) and for MO,DD,HH,MM,SS (&NUM.2 must be &NUM.2.).

Thanks to Harry Price, Florida Power and Light, USA.

Change 12.018 Type 30 Interval records with MULTIDD='Y' do not have a

VMAC30 CPU section, which is where the GMT offset is stored, and

Apr 22, 1994 thus INTBTIME and INTETIME were not corrected from GMT to

local time. Since GMTOFF30 is a retained variable, the

times can be corrected (as long as the MULTIDD record is

found after another type 30 which has a CPU section was

found) by inserting the correction logic:

INTBTIME=INTBTIME+GMTOFF30;

INTETIME=INTETIME+GMTOFF30;

inside the DO group which sets MULTIDD='Y';

Thanks to Don Friesen, B.C. Systems, CANADA.
Change 12.017 Variable BUFNO is now documented by IBM as always zero;

ADOC1415 the number of buffers is valid in JFCBUFNO. MXG now sets

VMAC1415 IF BUFNO LE 0 THEN BUFNO=JFCBUFNO

Apr 22, 1994 so you won't have to read this note in the future!

Thanks to Tom Elbert, John Alden Insurance, USA.
Change 12.016 CA-DISPATCH 5.1 corrupted READTIME (See Change 11.342) is

VMAC6 now corrected by CA PTF T97E056.

Apr 22, 1994

Thanks to Giovanni Dossena, Einchem Elastomeri S.R.L., ITALY.

Change 12.015 FMXGSID and FMXGUCBL functions ABEND 0C4 with SAS 6.08 at

FMXGSID TS407, but work on earlier versions. This is SAS error

FMXGUCBL related to pre-V6 style user written functions where the

Apr 21, 1994 SSI field has a min and max arguments of 0 and function