* copyright (C) 1984-2019 merrill consultants dallas texas usa
Yüklə 28,67 Mb.
|
|
This caused MXG to send this "real" purge record to the PDB.NJEPURGE, and the job's other records sat in the SPIN file waiting on the nonexistent purge record until SPINCNT was exceeded, when the job was finally output to PDB.JOBS. This change modifies the test for DEVNAME=:'OFF' and is (DEVNAME=:'OFF' AND JPURTIME LT JSTRTIME) so that only the unwanted SPOOL transfer purge record is excluded from BUILDPDB processing, by comparing the purge record time with the job start time. Thanks to David Ehresman, Univerity of Louisville, USA. Change 09.242 PR/SM variable NRPRCS is the number of partitions, butthe VMAC7072 pseudo-partition named PHYSICAL added by APAR was also Feb 18, 1992 being counted. Now it is subtracted out from NRPRCS. Thanks to Tim Follen, Blue Cross Blue Shield of Ohio, USA. Change 09.241 A DB2ACCT dataset with 250,000 obs requires 5,000 tracks. ANALDB2R A DB2 Accounting Summary report needed 10,000 tracks of Feb 17, 1992 work space (twice the 5,000 tracks because at the end of the sort phase both the input and output datasets exist), because ANALDB2R unwisely kept all 221 DB2ACCT variables! Now, only the 43 variables needed for the Summary report are kept, and only 1800 tracks are now required! A change in 9.7 that required yet another copy of DB2ACCT in the work file was also eliminated in this re-design. Thanks to Neil Ervin, Huntington Bank Service Company, USA. Change 09.240 The IRG (inter-record gap) calculation used to estimate feet of tape should have been IF IRG LT 38000 THEN .... VMACTMS5 instead of NE (VMACTMS) or LE (VMACTMS5). The 3480 tape Feb 17, 1992 length was correct, but 3490s calculated too long! Thanks to Sam Sheinfeld, Kraft General Foods, USA.
BUILDPDB required these changes. Feb 17, 1992 1. CICS processing must be moved from BUILDPDB to a second step. In member BUILDPDB, you will have to comment out the _VAR110 and _CDE110 lines, the lines which PROC SORT datasets starting with CIC....., and %INCLUDEs for members starting with CIC..... 2. Set MEMSIZE=6M, a Region size of 6M, and insert an OPTIONS BLKSIZE=2048; preceding the %INCLUDE of BUILDPDB. 3. If you need to process the type 110 CICS SMF records, you should use IMACEXIT to select ID=110 and write them to SMFTEMP the BUILDPDB step, then use a second step to execute %INCLUDE SOURCLIB(TYPE110), with the SMF DDname in this second step pointing to the temporary dataset name you used on the SMFTEMP DDname in the BUILDPDB step. 4. Again, all of this flailing around is necessary ONLY if you are trying to execute SAS 6.06/6.07 under MVS/370. If you have MVS/XA or MVS/ESA, you need none of the above! Thanks to Darrell Allen, Special Metals, USA.
VMXGHSM plus SMS-related flag, VSAM organization and last update Feb 18, 1992 datetimestamp were added to the three MCD datasets created from the Migration Control Data Set. Thanks to John E. Schultz, First National Bank of Maryland, USA.
EXPDQAMJ datasets: EXPDQAMO PDQAMON - Auxilliary Storage by performance group EXPDQEMJ PDQAMONJ - Auxilliary Storage by job (ASID) EXPDQEMO PDQEMON - Expanded Storage by performance group IMACPDQ PDQEMONJ - Expanded Storage by job (ASID). TYPEPDQ This vendor contribution has only been syntax checked. VMACPDQ
Feb 14, 1992 Thanks to Bill Ek, BEI Systems, USA. ---Changes thru 9.235 were printed in MXG Newsletter TWENTY-ONE----- Change 09.235 Support for Boole and Babbage IMF 2.7 is provided in MXG TYPECIMS (it was easy - there was no change in the IMF log record Feb 12, 1992 format since IMF 2.6!)
RMFINTRV Instead of the total memory (bytes) for the workload, Feb 12, 1992 they contained the per-user bytes for the workload, and variable TOTMEMR contained frames instead of bytes! The MEMR variables should have been divided by DURATM and not the RESD variables, and TOTMEMR should have been multiplied by 4096 to convert frames to bytes. Thanks to Shirley Rementer, Atlantic Electric, USA. Change 09.233 NETSPY variable LRSPNET normally contains the response VMACNSPY time of the last transaction, but a new variable exists, Feb 12, 1992 LRSPNETV='Y', to flag that the LRSPNET value contains the calculated average response for the interval. This is particularly important for LU 6.2 and APPC sessions, where only the calculated LRSPNET is available. Labels for LRSPNET and LRSPHOST should have indicated "LAST". Thanks to Marty Quinlan, Virginia Power, USA. Change 09.232 Sample tape error report enhanced with a report by device ANALESV type added, allowing comparisons of errors by different Feb 10, 1992 types of tapes and drives in your tape pool.
DOCDB2PM ANALDBTR, and READDB2 members adds new reports and DB2 Feb 10, 1992 2.3 considerations.
GRAFTRND add an additional set of graphs. If you had more than Feb 10, 1992 18 months of data in your TRNDRMFI dataset, some of the graphs (Avg vs Max CPU Busy and All Workload Bar charts) could not be produced because the x axis could not be fit on most graphics devices. The best circumvention would have been to use PROC GPLOT and AREA fill but this proved impractical because if AREAS=16 is specified and only 5 workloads are present, the entire area of the graph is filled with the pattern for workload 5. This is clearly not what you (or we) wanted. Until this is fixed (if ever) the workload bar charts will be limited to the past 65 weeks. Why 65? It appears to be the least common denominator for most graphic devices. 65 bars fit on all of the devices we have tried. The data to be retained is constructed based on the data in your TRNDRMFI data. An MXGNOTE is produced detailing the range of dates in your TRNDRMFI dataset and, if too much data is present, MXGWARNINGs are produced outlining the error condition and what will be done to circumvent the problem. There will be notes on the SAS LOG that some observations were missing or out of range. This is normal. When data is being projected, the actual CPU is missing. The use of PROC GLM to project using linear regression has been extended to the workload level (only if you have the SAS/STAT product.) These graphs are identical to the current workload graphs except that they carry workloads forward into the future the number of weeks specified by WEEKSOUT parameter. As with the other workload charts, only 65 weeks are charted. If a workload is shrinking, at the point in time at which it reaches zero or goes negative, it will be set to missing. Change 09.229 Support for Memorex Telex Library Management Software EXLMSADO Performance Monitor (LMS/PM), a significant contribution EXLMSAIN written by Dan Kaberon of Hewitt Associates, was provided EXLMSALD to MXG users by Memorex Telex. Not only did Dan write EXLMSATL this code in MXG style, he also provided Chapter 40 style EXLMSCHG MXG documentation, which you will find in ADOCLMS. EXLMSCIN Seventeen datasets are created from the twenty subtypes EXLMSEJE of the LMS SMF record. See ADOCLMS for more information. EXLMSENT LMSXX LMS SUBTYPE INVENTORY EXLMSEVE LMSINIT LMS SUBTYPE 0: LMS INITIALIZATION EXLMSINI LMSSDOWN LMS SUBTYPE 1: LMS SHUTDOWN EXLMSISC LMSCHG LMS SUBTYPE 2/12: SETTING CHANGE EXLMSLIN LMSALDOM LMS SUBTYPE 4: ALLOCATE/DOMAIN STMTS EXLMSMOU LMSEVENT LMS SUBTYPES 5,13-15: ATL EVENT EXLMSSDO LMSATLI LMS SUBTYPES 10: ATL INIT EXLMSUNL LMSADOWN LMS SUBTYPE 11: ATL SHUTDOWN/INACTIVE EXLMSVER LMSISCR LMS SUBTYPE 19: ATL INTVL SCRATCH STATS EXLMSXX LMSENTRY LMS SUBTYPE 20: CARTRIDGE ENTERED IMACLMS LMSVER LMS SUBTYPE 22: CARTRIDGE VERIFY TYPELMS LMSMOUNT LMS SUBTYPE 21: ATL MOUNT VMACLMS LMSEJECT LMS SUBTYPE 23: CARTRIDGE EJECTED Feb 10, 1992 LMSUNLD LMS SUBTYPE 24: UNLOAD LMSCINIT LMS SUBTYPE 25: CART INITIALIZED LMSLINT LMS SUBTYPE 30: INTERVAL LMSAINT LMS SUBTYPE 31: ATL INTERVAL
VMACDCOL in dataset DCOLDSET) is wrong; the total allocated space Feb 10, 1992 in the dataset records can exceed the allocated space in the volume record, because IBM uses the wrong track size in computing the size of the dataset. The unformatted track capacity is used, instead of the formatted track size. Changes 9.180 and 9.144 discuss APAR OY36364 which corrected the track size used in DCOLLECT volume records, but the correction was not applied to the dataset record. APAR OY48920 discusses DCDALLSP for 3390s, but it is not clear if that APAR recognizes the other fields that are wrong, nor if the fix (when a PTF exists) will apply to both 3380s and 3390s. Check the status of that APAR. Thanks to Terry Duchow, Minnesota Mutual Life, USA. Change 09.227 CICS Exception variables MSWAITCN and TSWAITCN should be VMAC110 missing, as these values existed only under CICS 1.6. The Feb 10, 1992 label for MSREQWT now correctly indicates it is the bytes of main storage acquired only after waiting. Thanks to Paul Masters, Blue Cross and Blue Shield of Texas, USA.
FORMATS needed to be input as HH PK1. MM PK1. SS PD2.1. Variable VMACWSF AUDACT was incorrectly input as $CHAR1. when it should be Feb 10, 1992 input PIB2. It is removed from the LENGTH statement, and it is now formatted with MGWSFAC (a new numeric format). Thanks to Douglas Clarke, General Accident, ENGLAND. Change 09.225 Revised JCL for MONTHBLD requires only ONE tape drive, JCLMNTH eliminates many tape mounts (especially if your weekly Feb 6, 1992 PDBs are multi-volume), and significantly reduces the run time of the monthly PDB build job. The revision copies each of the weekly PDBs from tape to temporary DASD (and selects only the datasets you intend to build monthly), and then uses those temporary DASD files as input instead of spinning input tapes back to the beginning of the tape for each input dataset to be read. The output monthly PDB is written to tape, on the same drive used to copy the weekly datasets. As long as you have enough temp DASD space for the job, this is a much better and faster method of building your monthly PDB. The revised JCL is in member JCLMNTH, but the existing member JCLMONTH was left in the library as well.
EXOPC03 Tracking and Audit Trail Log Records. Twelve datasets EXOPC20 are created from the standard IBM log records: EXOPC23 OPCA20 JT STARTED EVENT EXOPC24a OPCA23 OPERATION EVENT EXOPC24n OPCA24 MCP EVENT EXOPC25 OPCA25 SUBMIT EVENT EXOPC26 OPCA26 AUTO RECOVERY EXOPC27 OPCA27 MISSED FEEDBACK RECORD EXOPC28 OPCA28 FEEDBACK RECORD EXOPC29 OPCA29 AUTO-TRACKED EVENT EXOPC30 OPCA30 SPECIAL RESOURCE EVENT EXOPC31 OPCA31 ETT TAB FILE MAINTENANCE EVENT EXOPC32 OPCA32 AUDIT TRAIL LOG RECORD EXOPC33 OPCA33 MESSAGE LOG RECORD EXOPC34 In addition, four user datasets for user log records are EXOPC35 defined, OPCA00-OPCA03, for sites which write additional EXOPC36
VMACOPC data records to the log. TYPEOPC
Feb 5, 1992 Thanks to Wolfgang Vierling, Vereinte Versicherungen, GERMANY. Change 09.223 Support for Infomation Builder's FOCUS Multi-Session EXFOCMSO Option accounting SMF record. One dataset is created: IMACFOCU FOCUSMSO FOCUS MSO Accounting TYPEFOCU This record provides CPU and EXCP resources and duration VMACFOCU of session for accounting, in Release 6.5 Level 6 at PUT Feb 5, 1992 9109. Technical Memo 7857.1 describes the contents. Thanks to Mike Boyce, Anderson Consulting, USA.
VMAC39 causing INPUT STATEMENT EXCEEDED RECORD LENGTH error. MXG Feb 5, 1992 now checks for complete header before INPUTing record, and prints message and dumps record on SAS log instead of failing with a STOPOVER abend. Thanks to Connie Wilson, AMR Travel, USA. Change 09.221 An archaic and confusing comment at the end of this IMACFILE member was removed. Feb 5, 1992 Thanks to Connie Wilson, AMR Travel, USA. Change 09.220 DCOLLECT Migration and Backup datasets have new variables VMACDCOL UMFLAG1/UBFLAG1 and UMDSORG/UBDSORG. Variables added by Feb 4, 1992 APAR OY37378 to the Migration dataset were also added to the Backup record, but were overlooked by MXG until this change. (Since on one noticed their absence, they must not be very important, but they are now decoded!) Thanks to Joseph J. Faska, Depository Trust, USA.
VMACSPMS change the three @113's to @113, @115, and @116. Feb 4, 1992 Thanks to John Chapman, British Gas, ENGLAND. =======Changes thru 9.218 were included in MXG PreRelease 9.8 ========== Change 09.218 Preliminary sample code for processing Amdahl's APAF data IMACAPAF from MVS or VM data was provided by Amdahl. This code is TYPEAPAF not in the style of MXG, and may be revised to conform to Feb 2, 1992 MXG architecture, but it arrived in time to save you the effort of writing it yourself when you install APAF.
TYPEIMS have accurate SERVICTM and RESPNSTM, whenever the same Feb 2, 1992 MSGDRRN was reused at a later time for the same ODEST. First, the INPUT and OUTPUT datasets contained blank values for DTOKEN (because, from IBM, not all log records contain a token), and second, the 31I record was not used to create pseudo 31O and 35O records containing DTOKEN to correct the INPUT/OUTPUT datasets. Now, 31I records create both 31O and 35O records so DTOKEN exists, and just to be sure, INPUT/OUTPUT observations without a DTOKEN are now deleted before the final merge. Like all scotch tape and chewing gum fixes to IMS log processing, I hope this is the last one, and pray that IBM will soon replace IMS with DB2, since IBM refuses to enhance the IMS log with discrete transaction resource data. Thanks to Lynn Delacourt, Volvo Heavy Truck Corporation, USA. Change 09.216 IMS 3.1 records contain duplicate values of DLRTOKEN, the TYPEIMS key that is used to match log records together. The new Feb 2, 1992 "Quick Reschedule" event creates multiple type 07/08 log records to be created with the same value of DTOKEN, one pair for each Reschedule. Because MXG expected only one pair for each DTOKEN value, MXG's IMSTRAN dataset will be incorrect (generally overreporting IMSCPUTM, DL/I calls, etc.) without this change. Now, MXG OUTPUTs the first 08 record (the initial program schedule event), totals the resources in the subsequent multiple 07 records, and then OUTPUTs the final 07 record of the set with same DTOKEN. A "Quick Reschedule" occurs when a message region has run its maximum number of transactions per program schedule, finds there are more transactions to execute, AND finds there is no other program waiting on this message region. Each 08/07 pair contains the resources consumed by all of the transactions executed by that schedule, and it has never been possible to measure the resources used by an individual "multi-sked" transaction; the best we could ever do was to divide the total resources by the number of transactions executed during each program schedule. Now, that "average" resource per transaction is even more meaningless, since these multiple schedule records now have to be summed and divided by a larger total number of transactions. It would have made much more sense, and we would not have lost ground in resource measurement, if IBM had been wise enough to create a new DTOKEN value for each Quick Reschedule. Added 1999: Variable SCHEDULE in dataset IMSTRAN flags if the obs was the scheduled or the rescheduled event. Thanks to J. Cary Jones, Philip Morris, USA.
BUIL3518 occurrences of 26J2 which should have been 26J3 in both BUIL3606 of these JES3 members. Jan 31, 1992 Thanks to Paul Oliver, BHP Information Technology, AUSTRALIA Thanks to Steve Cavill, SAS Institute, AUSTRALIA Change 09.214 Support for NPM Release 1.5 added seven new datasets: EX028CM6 NPMCMEXE - Execute commands EX028LA1 NPMLANAT - LAN Bridge Attach/Detach EX028LA2 NPMLANCO - LAN Manager Connect/Disconnect EX028LA3 NPMLANEX - LAN Bridge Monitor Exception/Resolution EX028LA4 NPMLANIN - LAN Bridge Interval Resources EX028LA5 NPMLANST - LAN Start/Stop/Alter Collection EX028NCP NPMNCPCH - NCP Add/Replace/Delete FORMATS and some new fields and bit values were added to existing VMAC28 datasets. IBM publication, "Netview Performance Monitor Jan 31, 1992 Reports and Record Formats Release 5", SC31-6147-0 is the new documentation for all of the data records created by this new release. Thanks again to IBM's excellent vendor support, documentation was made available in sufficient time to include support for this significant release in MXG's production version. Unfortunately, no data records from the new release were provided for testing, so caveat user!
READDB2 value for "ALL", and to relocate the %INCLUDE until after Jan 29, 1992 requests have been parsed for efficiency. ==Changes thru 9.212 were included in MXG PreRelease 9.7==============
ANALDBTR were added for DB2 2.3 IFCIDs 170-171,173-176,178-179,and VMAC102 183, and these IFCIDs are now decoded in VMAC102. Jan 27, 1992 Change 09.211 Sites that execute BUILDPDB only six days per week will MONTHBLD need to locally modify MONTHBLD, as it expects seven day Jan 27, 1992 per week possible execution. One line in MONTHBLD, (line 003400, DAY=UPCASE(PUT(TODAY,WEEKDATE3.)); ) was moved to immediately after line 002100, as this creates the variable DAY containing the name of the day of week. With that permanent MXG change in place, the changes you need to tailor in you own sites copy of MONTHBLD are more compact and logically clearer. There are two changes to be made by you: a. Change IF DAY(TODAY) NE 1 THEN DO; to read IF DAY(TODAY) NE 1 AND (DAY(TODAY) NE 2 AND DAY NE 'MON') THEN DO; This permits MONTHBLD to execute either on the first day of the month, or on the second day of the month if the first day fell on Sunday (the day you do not execute BUILDPDB). b. Insert after LASTDAY=TODAY-1; the statement IF DAY(TODAY) EQ 2 THEN LASTDAY=LASTDAY-1; This causes LASTDAY to contain the date of the last day of the last month, which is then used to set the _BEGIN macro value for date selection. Note that as the _END macro value is TODAY, the selection values will now be correct whether you execute on the first of the month or on Monday the second day.
ANALDB2R reports, and many additional internal enhancements: Jan 26, 1992 1.All WARNING or NOTE messages printed on the log by MXG now print as MXGWARN or MXGNOTE to clarify their source. 2.Internal macros are now documented with update date. 3.All uses of VMXGSUM use the INVOKEBY parameter to message which report is being generated ("PMACC01 - ANALDB2R"). 4.Selection was consolidated into DB2RSELA macro. 5.DB2DBID macro uses T102S105 and T102S107 (IFCID 105,107) to build a temporary format which maps hex DATABASE and PAGESET values to their actual names. Under SAS 5.18, the INSTREAM DD pointing to temporary sequential space is required, but under SAS 6.06+, the CNTLIN option of PROC FORMATS is used to create formats directly from SAS datasets. 6.Reports now check for the existence of required datasets (as opposed to just zero observations) and messages now indicate why reports were not produced. 7.PDB=SMF option (produce DB2 reports directly from SMF) cannot be used under SAS 5.18 (compiler limit) in a 7MB region. Messages now tell you how to use READDB2 plus ANALDB2R instead of PDB=SMF to produce almost all reports directly from SMF if you are still using SAS 5.18.
TRNDDB2A and subsequent week's processing, old data was not even TRNDDB2S carried forward, and timestamps were not set correctly. Jan 26, 1992 Thanks to Dave Leeker, Southwestern Bell, USA.
VMAC79 added to the TYPE79A (subtype 10) dataset for the target Jan 24, 1992 IN and OUT MPLs respectively. Thanks to Diane Eppestine, Southwestern Bell, USA. Change 09.207 Variable RDRTM in TYPE30 was always missing after Change VMAC30 9.102 because the test in line 075198 should also have Jan 24, 1992 tested for SUBTYPE NE 1 to detect the MULTIDD condition. Thanks to Don Friesen, B.C. Systems, CANADA. Change 09.206 Variable EXCMNTYP in dataset CICSEXCE (CICS Exceptions) FORMATS did not describe the exception type because it was kept VMAC110 in one byte instead of two. In the LENGTH statement, it Jan 24, 1992 must be $2. instead of $1. Additionally, the $MG110EX format in member FORMATS should have been '01X:' instead of the obvious typographical error '02X:' Thanks to Bill Hess, Belk Stores, USA.
ANALRACF not in the member. $TABSYS has been replaced by $4., and Jan 14, 1992 $FORHEX (which converted hex strings to numerics) was Yüklə 28,67 Mb. Dostları ilə paylaş:
|