* copyright (C) 1984-2019 merrill consultants dallas texas usa
Yüklə 28,67 Mb.
|
|
nor the PDB.SMFINTRV data set):
APPCCN ='TOTAL*CONVERSATIONS*(ACTIVE+DEALLOC)' APPCCNA ='NUMBER OF ALL*CONVERSATIONS*ALLOCATED' APPCDAR ='BYTES OF DATA*RECEIVED*BY THE TP' APPCDAT ='BYTES OF DATA*SENT*BY THE TP' APPCREC ='RECEIVE CALLS*ISSUED*BY TP' APPCSEN ='SEND CALLS*ISSUED*BY TP' APPCTAC ='NUMBER OF*ACTIVE*CONVERSATIONS' d. The JCTJOBID field from which TYPETASK and JESNR are extracted has changed with APPC. Instead of three characters JOB/STC/TSU follwed by the five digit JESNR the JCTJOBID will contain an "A" and a seven digit number, which MXG stores in TYPETASK and JESNR, but the maximum JESNR is 999,999 because the first of the seven digits is always a zero. You should check if any reporting programs use TYPETASK for selection, and to ensure they are coded robustly so they will not fail if TYPETASK='A' is encountered. The JCTJOBID change was not compatibly implemented and changed VMAC6,VMAC30,VMAC32,VMAC26J2, and VMAC26J3. SMF records use the fieldname SMFnnJNM for the eight byte JCTJOBID field. This IBM change in that field could also affect "banner page" printing code in your SMF exits.
"TP" resources, for each TP work scheduled by ASCH. Each type 33 is an APPC event, with the same APPC counts that are totaled in the type 30, above. In addition, times of receipt, scheduling, and execution of the TP are recorded as is the CPU times, I/O connect time and EXCP counts for APPC TPs. This new data is the resource record for APPC, and will be used instead of the type30s for APPC resource accounting, performance measurement, etc. The variables in this data set from the type 33 subtype 1 record are:
APPCCNAS='CONVERSATIONS*ALLOCATED*BY THIS TP' APPCCONS='NUMBER OF*CONVERSATIONS*FOR THIS TP' APPCDARS='BYTES*RECEIVED*BY THIS TP ' APPCDATS='BYTES*SENT BY*THIS TP ' APPCRECV='RECEIVES*ISSUED*BY THIS TP' APPCSEND='SENDS*ISSUED*BY THIS TP ' CPUSRBTM='CPU*SRB*DURATION' CPUTCBTM='CPU*TCB*DURATION' ELAPSTM ='ELAPSED*DURATION' EXCPTOTL='EXCPS*FOR*THIS TP' INPREQTM='INPUT*REQUEST*DURATION' IOTMTOTL='DEVICE*CONNECT*DURATION ' LENACCTn='LENGTH OF*ACCOUNT FIELD* n ' LOCLLUNM='LOCAL*LU NAME*FOR THE TP' NRACCTFL='NUMBER*ACCOUNT*FIELDS' PARTLUNM='NODENAME*LUNAME*OF PARTNER' QUEUETM ='SCHEDULER*QUEUE*DURATION' RACFGRUP='RACF*GROUP*IDENTIFICATION' RACFUSER='RACF*USER*IDENTIFICATION' REQDTIME='REQUEST*RECOGNIZED*BY FMH-5' SKEDTIME='REQUEST*PLACED ON*SCHEDULER QUEUE' SMFTIME ='TIMESTAMP*WHEN RECORD*WAS WRITTEN' SYSTEM ='SMF*SYSTEM*ID' TPCLASS ='TP*CLASS' TPNAME ='TP*NAME' TPNDTIME='TP*EXECUTION*ENDED' TPROFILE='TP*PROFILE*NAME' TPSKEDTY='TP*SCHEDULER*TYPE' TPSTTIME='TP*EXECUTION*STARTED' TPUSECNT='USES OF*THIS TP*BY THIS USER' TPUSRTYP='TP*USER*TYPE' ZDATE ='ZEE DATE*ZEE OBS*WAS CREATED'
to describe the events for a standard TP scheduled event: INPREQTM QUEUETM ELAPSTM Waiting on Waiting on Executing Scheduler TP queue |_____________|_______________|____________| | | | | REQDTIME SKEDTIME TPSTTIME TPENTIME Request Placed on Execution Execution Received TP's queue Started Ended 4.New variables were added to TYPE4345 for JES2 start options:
LOGOPTN ='LOG*OPTIONS*SPECIFIED?' QUIKSTRT='JES2 DID*QUICK*START?? RECNFGOP='RECONFIG*OPTION*SPECIFIED?'
max,avg number of APPC address spaces active, and twelve distribution buckets APPC00 thru APPC11 count the percent of time when 0, 1-3, ... over 36 APPC address spaces were active, paralleling the similar set of fifteen variables for the existing other three types of address spaces for BATCH, TSO, and STC addresses.
PWSS (Primary working set) pages migrated, and the number of ESTORE frames that were freed without migration: ESPGFRNO='ESTORE FRAMES FREED W/O MIGRATION' PGBKtoin='BLOCKED*PAGES*PAGED IN' PGBKtoou='BLOCKED*PAGES*PAGED OUT' PWSSMIIN='PWSS PAGES*MIGRATED*FROM ESTORE'
Eleven new variables, including the long needed new ESA CPU times, and the ESTORE residency time (which can be compared with the existing variable ACTFRMTM, previously discussed in NEWSLTRS member) to estimate usage of ESTORE frames. ACTESFTM='EXPANDED*STORAGE*RESIDENCY' ACTFRMTM='ACTIVE*FRAME*TIME' BLKSAUIN='BLOCKS*PAGED IN*FROM AUX' BLKSESIN='BLOCKS*PAGED IN*FROM ESTORE' CPUHPTTM='HIPERSPACE*CPU TIME' CPUIIPTM='I/O*PROCESSING*CPU TIME' CPURCTTM='REGION*CONTROL*CPU TIME' PGBKESIN='BLOCKED PAGES*PAGED IN*FROM ES' PGBKTOIN='BLOCKED PAGES*PAGED IN' PGTOESIN='PAGES*PAGED IN*FROM ESTORE' SYSTRNTM='SYSTEM*TRANSACTION*TIME'
CFGCHGFL='SMF73CFL CONFIGURATION CHANGE FLAGS' CHPADDED='CHANNEL*PATH*ADDED?' CHPDELET='CHANNEL*PATH*DELETED?' CHPMODIF='CHANNEL*PATH*MODIFIED?' IODFNAME='IODF*NAME' IODFSUFX='IODF*NAME*SUFFIX' IODFTIME='IODF*CREATION*DATETIME' 9.Type 74 record contains the same fields added to the type 73 listed above, but the variables are NOT kept in TYPE74 because I don't think the 44 bytes of IODF name needs to be kept in each device segment of MXG's TYPE74 dataset. If they really turn out to be needed, I will create a new data set, one per type 74 record, instead of one for each device. The code is in place, so the variables do exist and are available in EXTY74, if needed. _______ _______ ________ |Offset |Offset |Offset | SMF Type 33 APPC/MVS TP Accounting |to POF |to IOF |to UOF | |_______|_______|________| Product Section: ____________________________ | | POF: | "ASCH", MVSLEVEL | |____________________________| TP Identification Section: _________________________________________________________ |Offset | | IOF: |to TPO |TP Class, Standard/Multi-Trans, TP Profile Name | |_______|_________________________________________________| \ \
\ TP Program Name Section: \ _______ _______________________________ | | | TPO: | Len |TP Name (up to 64 characters) | |_______|_______________________________| ______________________ / \
TP Usage Section: / \ _______________ _______ _________ _______ \ | |Offset | |Offset | / UOF: |RACFUSER,GROUP |to AOF |Usecount |to TDO | / |_______________|_______|_________|_______| / / /
/ / / /
/ / _________________/ / / / / TP Usage Accounting Section: / ______ __________________ / | |Accounting fields | / AOF: | Len |(up to 175 char) | / |______|__________________| / /
TP Usage Detail Section _______ ___________________________________________ |Offset | | TDO: |to TSO | Sends,received,data,conversations,TCB,SRB | |_______|___________________________________________| \ \ \ \ TP Usage Scheduler Section \ _____________________________ | |
TDO: | LLU, PLU, Four Time Stamps | |_____________________________| 10.Type 78 record contains the same fields added to the type 73 listed above, but the variables are NOT kept in TYPE78 because I don't think the 44 bytes of IODF name needs to be kept in each subtype 3 segment. See above note. Change 08.223 Deaccumulation of EXCP count from type 14/15 records by ANALDSET this MXG analysis routine did not handle concatenated PDS Jan 25, 1991 libraries, noticibly causing JES2 job's PROCnn DDNAMES to have incorrectly high EXCP counts in this analysis. The BY list in lines 017300 and 017600 should have had the variable DSNAME inserted between DDNAME and UNITADR: BY SYSTEM READTIME JOB STEP DDNAME DSNAME UNITADR SMFTIME; Thanks to Lee Hollis, Lowe's Companies, Inc, USA. Thanks to Dee Ramon, Mutual of America, USA.
IMACWSF records are intended to all be 512 (an impossible value) Jan 25, 1991 but some IMACs slipped into the library with a different value; if your site had a record of that value, the test of JCLTEST could cause an "INPUT STATEMENT EXCEEDED " STOPOVER condition. All IMACs for User SMF records now have the value of 512, as they should have all along! Thanks to Glenn Hanna, Textron Defense Systems, USA. Change 08.221 SAS 6.06 Compatibility. Mixed case from PUT function. MONTHBLD The DAY=PUT(TODAY,WEEKDATE3.); may create mixed case in Jan 25, 1991 the character variable DAY under SAS 6.06, depending on several SAS 6.06 options; the first letter is upper case, while the second and third letters of the name of the day are lower case. Since the subsequent test is for caps, the test fails. Line 005200 is now replaced with DAY=UPCASE(PUT(TODAY,WEEKDATE3.)); to force DAY to be uppercase, independent of the sites options for case. Thanks to Barry Lampkin, Blue Cross Blue Shield of Mass., USA. Thanks to Debora Reinert, Nordstrom, Inc, USA. Change 08.220 DCOLLECT variable DCDBKLNG should have been multiplied VMACDCOL by 1024 (like the other adjacent values) to convert to Jan 25, 1991 bytes for printing by the MGBYTES format. Thanks to David Stern, CSX Technology, USA. Change 08.219 Hiperbatch statistics added to the TYPE64 data set were VMAC64 not kept in the correct dataset. The line containing Jan 24, 1991 the six new variables was mis-located in the keep list for TYPE64X (which is output before the variable are input) instead of the correct keep list for TYPE64. Variable JFCBDSNM was spelled incorrectly as JCFBDSNM. Thanks to Michael Enad, Dun and Bradstreet, USA.
IMACPDB 1.Line 029800 in IMACPDB should be _FREQ_ vice _FREQ IMACACCT (but no error was set as _FREQ_ only exists in 6.06 VMAC110 and the missing underscore only caused _FREQ_ to be kept. Jan 21, 1991 2.Comments in IMACACCT were enhanced to suggest that the unwanted LENACCTn variables should also be dropped, and that meaningless SAS "393" warnings print on the log. 3.CICS 3.1 processing was not protected for an unexpected Statistics subtype. It did not fail, but the messages produced were unclear. Now it is protected. Thanks to Norbert Korsche, OMV-AG, AUSTRIA.
EXEPMEP 180" data record is added by this user contribution. EXEPMIO Three EPMV.. data sets are created: EXEPMEQ EPMVEP - The fixed workload information and the wait IMACEPMV statistics for the two kinds of workloads, TYPEEPMV PG (perf groups, periods, total sys), and VMACEPMV AS (TSO, JOB, STC). Variable SM180SUB shows Jan 16, 1991 which of the 10 subtypes created the record: TOTS - all perf groups combined PGN - all periods of a perf group PGP - a single period of a perf group BAT - end of step for batch address space TSO - end of step for TSO address space STC - end of step for STC address space bat - end of interval for batch ASID tso - end of interval for TSO ASID stc - end of interval for STC ASID RSRC - EPILOG captured data (not from RMF) EPMVEPIO - I/O counts for each device with counts. EPMVEPEQ - ENQ counts for each major name with counts. Thanks to Billy Westland, Candle Corporation, USA. Thanks to John Ebner, Litton Computer Services, USA. Change 08.216 Both SIO73CNT and SIO78CNT are missing in RMFINTRV for a RMFINTRV 4381/3090/ES9000 processor, because I/O counts moved into Jan 16, 1991 different records for different processors and software. Mar 13, 1991 Insert after line 070200 DATA TYPE78; SET PDB.TYPE78; (so a PDB can be on tape), and expand the SET statement in line 070400 to read: SET TYPE73P TYPE78 (IN=IN78) PDB.TYPE78IO (IN=IN78IO); and insert new line after 070700: IF IN78IO THEN SIO73CNT=IOPACTRT*DURATM; to now populate variable SIO73CNT to contain total SSCH count to all channels from the appropriate data set. Thanks to John McDonald, Arizona State University, USA.
ANALDB2R received. Line 076980, variable QW0054AI should have Jan 16, 1991 numeric 0's instead of alphabetic o's. Line 077020, variables MINTIME and MAXTIME need to be added to that RETAIN statement. Thanks to Elliot Weitzman, Oryx Energy Company, USA. Change 08.214 "INVALID THIRD ARGUMENT TO FUNCTION SUBSTR" error message VMAC6156 from syntax WANT6156=INPUT(SUBSTR(SYSPARM(),10,7),7.); Jan 15, 1991 when the SYSPARM value is less than 17 characters. This soft error can be eliminated by storing the value of the SYSPARM() function into a variable, because SAS sets that variable's length to 200 bytes. The variable is then used in place of the subsequent SYSPARM() references. Lines 012200 thru 012600 now read: IF WANT6156=. THEN DO; SYSPARM=SYSPARM(); IF SYSPARM EQ: 'TYPE6156-' THEN WANT6156=INPUT(SUBSTR(SYSPARM,10,7),7.); ELSE WANT6156=-1; RETAIN WANT6156; END;
only occurs in SAS 6.06 if ZAP Z6060627 was installed, and that zap is now marked REMOVED by the Institute). But I'll leave my robust circumvention in place! Thanks to Debbie Matic, Commercial Union Assurance of Canada, CANADA.
Book place of the old style _DAY macro and INSTREAM DD. Jan 15, 1991 The example in the Chapter 35 to copy the DAY PDB into the correct "Day-of-Week" PDB uses a temporary dataset (pointed to by the INSTREAM DDname), and writes out an old-style MACRO named _DAY that contains the character name of the day of the week, which is then used as the DDNAME into which the daily PDB will be copied. That example has been updated and commented below:
OPTIONS NOTITLES; DATA _NULL_ ; FILE INSTREAM NOPRINT RECFM=FB LRECL=80 BLKSIZE=8000; TODAY=TODAY()-1; /* get yesterday's date */ WEEKDAY=' '; WEEKDAY=PUT(TODAY,WEEKDATE3.); PUT 'MACRO _DAY ' WEEKDAY $CHAR3. '%' ; RUN; %INCLUDE INSTREAM; /*read macro just created*/ PROC COPY IN=DAY OUT= _DAY; Some users have been directed by SAS technical support to replace the old-style substitution macro with the %MACRO facility. One recommended implementation is:
TODAY=TODAY()-1; DAY=UPCASE(PUT(TODAY,WEEKDATE3.)); CALL SYMPUT('_DAY',DAY); PROC COPY IN=DAY OUT=&_DAY;
VMACEPIL for EPILOG CL1000 back in member VMACEPIL. Change 8.019 XMACEPIL was applied to member XMACEPIL, which was then copied in Jan 14, 1991 to VMACEPIL. It is believed to support 4.4.0, 4.5.0, and 4.5.1 versions, but test data is not at hand.
VMACSYNC record produced by SYNCSORT now identifies if this SORT Jan 14, 1991 was invoked by the SAS System. Note that SYNCSORT's enhancement PTF number EW2903 must have been installed by your SYNCSORT installer before this field will be "Y".
VMACDCOL UMTIME or UBTIME were wrong. Both occurrences of JULDATE Jan 14, 1991 must be changed to DATEJUL (lines 037800 and 040000). In addition, the value written in the record by IBM is apparently wrong itself, until you install APAR OY37378. Thanks to Derek Cespedes, Florida Power and Light, USA. Change 08.209 TSO/MON variable READTIME is missing in PreReleases 8.3 VMACTSOM thru 8.7. Change 8.104 to correct CA7 corruption of the Jan 14, 1991 first byte of Reader Date worked ONLY if for corrupted records; normal, non-CA7 records produced missing values. In line 040110, remove the DO; and delete line 040130 (containing END;) so that READTIME is always created by the INPUT function. Thanks to David B. Adams, National Life of Vermont, USA. Change 08.208 IBM SMF APARs have added new data fields. VMAC23 1.SMF now flags when the CPU timer goes bad in a new field VMAC30 in the type 30 timer section. APAR OY24857 applies. Jan 12, 1991 New variable CPUERROR (which was formerly a reserved field) will have hex 0000 (APAR not installed), hex 8000 (APAR installed, no error). The subsequent bits are on if the corresponding CPU measurement value (there are now seven CPU measures) was defective and set to zero in this record. The APAR also eliminates an OC9 ABEND of SMF. 2.SMF now flags a step cancellation due to NOT CATLG 2 or NOT CATLG 7 condition (one bit for both reasons), but only if the installation specified JOBFAIL(YES) in ALOCXX SYS1.PARMLIB member (which ABENDs the NOT CTLG JOB). APAR OY38977 applies. IBM's bit 11 of SMF30STI is used, and MXG ABEND will contain NOTCTLG if the step was cancelled due to NON CATLG condition. In addition, new variable TERMIND contains all 16 bits of the step/job indicator for detail examination in your programs, if needed. 3.SMF type 23 record (buffer statistics) is expanded to better describe the SMF writer's use of its memory. APAR OY27449 applies. New variables created in TYPE23 dataset: CURALLOC='SMF BUFFER*CURRENT*ALOCATION' HWMALLOC='SMF BUFFER*HIWATERMK*ALLOCATION' MAXBUFSZ='BUFFER*MAXIMUM*LEVEL' MVSLEVEL='MVS*SOFTWARE*LEVEL' PERALLOC='INCREMENT*PER EACH*ALLOCATION' WRNBUFSZ='BUFFER*WARNING*LEVEL'
JCLTEST6 JCL statement was missing from the FORMATS step, and the Jan 12, 1991 SMFSMALL step's //SYSIN DD DSN=MXG(UTILGETM) was replaced with a //SYSIN DD * statement and a %INCLUDE. Change 08.206 SAS 6.06 Compatibility, only in MXG PreReleases thru 8.7. UTILGETM The UTILGETM program, used only during installation of Jan 12, 1991 MXG (to build a small SMF file for testing), will put SAS 6.06 (even at the October maintenance level) in a 100% CPU Busy loop, causing a System 322 ABEND. The error is in the processing of the %VMXGVBS macro which was added by MXG Change 8.174. This macro generates different DCB attributes for the FILE statement (CMS can't create VBS) so new MXG users under CMS did not die during testing. The %VMXGVBS macro added to UTILGETM was almost exactly the same as the %VMXGLRC macro for the INFILE statement in the _SMF macro. The differences were that %VMXGVBS sets only the LRECL parameter of the INFILE SMF, while %VMXGVBS sets LRECL, DCB, and BLKSIZE options, and that their references were located in different positions in their respective INFILE and FILE statements. Over 50 runs 322 ABENDed before the %MACRO compiler error was found to be circumvented when there is a "parameter=value"syntax preceding the %VMXGVBS invocation. Adding COL=COLOUT to line 01900 in UTILGETM circumvents the loop. This works: FILE SMFOUT COL=COLOUT %VMXGVBS; This didn't: FILE SMFOUT %VMXGVBS; SAS Institute now knows the loop occurs only a %MACRO on a FILE or INFILE statement (and also only if the %MACRO itself starts with %IF), and are working on the problem. SAS Zap Z606xxxx, available Month dd,yyyy corrects this error in the SAS 6.06 %MACRO compiler. Tracking 163046. Thanks to Barry Lampkin, Blue Cross Blue Shield of Mass., USA. Thanks to Grady Tart, McM Information Services, Inc, USA. Change 08.205 PreRelease 8.2 thru 8.7 only. The label for CPUTCBTM may VMAC110 say "SUM OF THREE CPU TCB TIMES", but it is still the Jan 11, 1991 CPU TCB TIME. (The CICS 3.1 support label overrode the original label). Thanks to Raymond Zieverink, Belk Stores Services, USA.
ANALDB2R correct variable QXNSMIAP was added by this change, but VMACDB2 the original variable is also kept so that any reports Dec 20, 1990 that you've written won't fail. Please correct in your report programs and eventually QXMSMIAP will be dropped. Thanks to Ron Root, Oryx Energy Company, USA. ---Changes thru 8.203 were included in MXG PreRelease 8.7 Dec 18, 1990-- Change 08.203 WSF2 record from RSD newest version 3.3.4 has a section VMACWSF mis-located two bytes to the right and truncated at the Dec 17, 1990 end. However, MXG code had not been validated with a real record from that version, and a nine-bit bit test only failed when actually executed, and the overlay of two fields in the WSF2 DSECT was missed in MXG coding. The two MXG oversights were corrected, and the logic was expanded to support both the correct and incorrect record format. Until maintenance exists from its vendor, the value of ACCNPELT will be incorrect, and usually zero. Thanks to Dennis Mahlubam, Aetna Life & Casualty, USA.
VMAC79 The equation for R791FMCT in line 027800 should be Dec 16, 1990 ELSE R791FMCT=R791RSV1 + R791RSV2; Note R791WSS is missing here, since SMF79ASL is only 60. Thanks to Miller Dixon, Continuum, USA.
VMACVMXA VXUSEINT record the HFQUCT (Hi Freq Sample Count) resets Dec 16, 1990 itself erratically. MXG tries to capture the CPU time a VM machine used between a logon and the end of the first interval, and used HFQUCT in that heuristic algorithm. However, that logic is now invalidated by this IBM reset and the user interval record is no longer output when a logon event is detected. The actual change was to comment out the two lines "IF NOT (FIRST.BEGINMTR OR Yüklə 28,67 Mb. Dostları ilə paylaş:
|