* copyright (C) 1984-2019 merrill consultants dallas texas usa
Yüklə 28,67 Mb.
|
|
SMF data is processed, the TYPE70PR data values are used to create TYPE70, and the new TYPE70EN dataset must be created from the (temporary) TYPE70EN/EL data if you have more than 64 engines, since only the first 64 engine's per-engine-detail is kept in TYPE70. But the below tailoring example shows how these old-style substitution macros are used to create only the datasets PDB.TYPE70 and PDB.TYPE70EN, by reinstating the _Wdddddd dataset tokens to be created when SMF is read, and the two _Ldddddd output dataset tokens for clarity. %LET MACKEEP=%QUOTE( _N7072 MACRO _WTY70 TYPE70 % MACRO _WTY70PR TYPE70PR % MACRO _WTY70SP TYPE70SP % MACRO _WTY70EC TYPE70EC % MACRO _WTY70EL TYPE70EL % MACRO _LTY70 PDB.TYPE70 % MACRO _LTY70EN PDB.TYPE70EN% ); %INCLUDE SOURCLIB(VMACSMF,VMAC7072,IMACKEEP); DATA _VAR7072
_SMF _CDE7072
_STY70 Or, you could use the %UTILBLDP macro to generate and execute the code to create only those two datasets: %LET PTY70PR =WORK; %LET PTY7002 =WORK; %LET PTY70X2 =WORK; %LET PTY70Y2 =WORK; %LET PTY72 =WORK; %LET PTY72DL =WORK; %LET PTY72GO =WORK; %LET PTY7204 =WORK; %LET PTY72MN =WORK; %LET PTY72SC =WORK; %UTILBLDP(BUILDPDB=NO,USERADD=7072,OUTFILE=INSTREAM); The %UTILBLDP macro is STRONGLY recommended for any SMF processing that needs tailoring. Thanks to Chuck Hopf, Independent Consultant, USA.
VMACSMF the SAS log was wrong if BUILDPDB was used; Change 28.089 Oct 7, 2010 modified VMACID to output SMFBYTES, but that variable was used internally in VMACSMF to total the byte count for that log message. The byte counter in VMACSMF is changed to SMFBYTOT to correct the message and not impact ID. When wrong, xxx had the record length of the LAST record. Thanks to Rudol Sauer, T-Systems International GmbH, GERMANY. Change 28.237 Variable R792PRFX was labeled FRAMES (MB) but was not VMAC79 converted from frames. While the (MB) in the LABEL should Oct 7, 2010 document when frame counts are converted to bytes, those variables must also have MGBYTES. format to confirm that the contents of the variable is bytes and "prints pretty" with K/M/G/T suffix with that format. Thanks to Jim Horne, Lowe's Companies, Inc, USA.
ANALINIT run all jobs in the same job class. That makes ANALINIT Oct 5, 2010 by jobclass mostly meaningless since TPM will prioritize work based on service class and other criteria given it via parameter. This change allows you to choose the variable to use for breaking the report down. The default value for SORTBY is JOBCLASS but you might find in these cases that SRVCLASS or RPTCLASS have more meaning. -HTMLDEST is also added as a parameter to send the output to HTML and the MEAN/MAX input queue delays are added to the final report. Thanks to Chuck Hopf, Independent Consultant, USA. Change 28.235 Using %VGETDDS(GOOVOO=gdgbase) caused ERROR: LIBRARY DOES VGETDDS NOT EXIST. The wrong macro variable &GDGBASE was used Oct 4, 2010 where &GOOVOO was required. Thanks to Brian A. Harvey, HCL America, USA. Thanks to Rhonda Martin, USAA, USA.
VMAC102 data are corrected to $CHAR informat, but that is only VMACCIMS impacting if MXG is executed on ASCII platforms to read VMACORAC SMF data. Detected in Freddie's final QA tests. Oct 4, 2010 -VMAC102: QW0022LM QW0022PA QBMCRLIR -VMACCIMS: RECSTAT -VMACORAC: ORALOGON Thanks to Freddie Arie, Merrill Consultants, USA. Change 28.233 Support for WebSphere MQ Version 7 Accounting Records. EXWSMQMQ Two datasets are created: EXWSMQQ dddddd Dataset Description IMACWSMQ WSMQMQ WSMQMQI Accounting MQI TYPEWSMQ WSMQQ WSMQQ Accounting Q TYPSWSMQ See Monitoring WebSphere MQ Version 7, SC34-6937, for the VMACWSMQ documentation on these data, which are written to the VMXGINIT SYSTEM.ADMIN.ACCOUNTING.QUEUE. Oct 2, 2010
ANALRMFR Change 28.155 revised token names for TYPE70EN, and some Oct 6, 2010 resets of _WTY70PR were missing. -While PDBOUT=PDB worked, PDBOUT=XXXX failed because the new TYPE70EN dataset was written to PDB instead of XXXX. Thanks to Russ Alexander, OIT Service Delivery, USA. Change 28.231 NDM-CDI NDRMTYPE='RT' record INPUT STATEMENT EXCEEDED VMACNDM error when NDMOFFPA was non-zero, a condition not seen Oct 1, 2010 before, and my guess about the order of the optional segments was wrong. Thanks to Douglas C. Walter, Citigroup, USA. Thanks to Brent Turner, Citigroup, USA. Change 28.230 FORMAT MG070CR decodes SMF70CCR to identify the Capacity FORMATS Change Reason: VMAC7072 SMF70CCR Formatted Value Oct 1, 2010 0 0:NO CHANGE 1 1:POWERSAVE 2 2:CYCLE STEERING 3 3:EXTERNAL 4 4:EXTERNAL Change 28.229 MXG Search tool did not select CHAR variables because the VMXGSRCH TYPE field was not UPCASED, and if the last variable in a Sep 30, 2010 dataset is numeric, the generated code did not wrap to close the continuation correctly; these are now fixed. Also, MXG 28.05 updates added the capability to search for text in Labels and for format names. On a positive note, Scott commented: I really do get a lot of benefit using VMXGSRCH when looking for evidence of a dataset or USERID or a CICS terminal-name. It's a great tool, particularly where I needed to find evidence of a CICS terminal in some CICS PA report, but looking in MXG for it CICS STATS. And VMXGSRCH is not only for MXG built datasets. VMXGSRCH will search ALL of the SAS datasets in ANY data library, searching ALL variables in those datasets for ANY value, printing the found observations. See Change 28.147. Thanks to Scott Barry, SBBWorks, USA.
VMXGRMFI It had the number of installed processors because parked Sep 29, 2010 time, CPUPATTM, was not kept nor subtracted from SMF70ONT when NRCPUS, the average number of ONLINE CP engines was calculated. Thanks to Richard Schwartz, State Street Bank, USA. Thanks to Brian Kruse, State Street Bank, USA.
VMACITRF because Change 28.197 didn't correctly INPUT these fields Sep 27, 2010 CORDBI CORDBP CORDBT that were added. Thanks to Lindholm Orjan, Volvo, SWEDEN. Change 28.226 -TYPE113 variable LPBUSY replaces wrong-named LPARCPU, as ASUM113 percent busy in TYPE113 is for each Logical Processor. VMAC113 -ASUM113 variable LPARBUSY replaces LPARCPU since there, Sep 23, 2010 the percent busy is for the LPAR. New variable LPARCPUS Oct 4, 2010 counts the number of LPs active in this LPAR. Normally, I'm reluctant to change variable names, but the SMF 113 is new so only a few "early adopters" will be effected, and the old names were deceptively wrong. -LPARCPUS is the true Average Number of ONLINE Engines, as it is calculated as SMF70ONT/DELTATM. -Variables from TYPE70PR that were added to the dataset PDB.TYPE113 are also now added to PDB.ASUM113. -ASUM113 corrected to not fail when PDB.TYPE70PR doesn't exist, but all variables that are merged in from 70PR will always exist in PDB.ASUM113, with missing values if there was not 70PR dataset. -SM113CPT is only populated if SM113VN2 GE 2, i.e., z196. (See Change 30.123 for revision). Thanks to Don Deese, (CPExpert), Computer Management Sciences, USA.
VGETOBS in character when the SAS dataset was created, and macro Sep 23, 2010 variable VGETMTYP is the type of dataset (data/view/etc). Oct 5, 2010 VGETCRDT is only populated under SAS; it is not in WPS. Thanks to MP Welch, Aprize, Inc, USA.
VMAC7072 PCTCPUBY for Parked Engines; PCTCPUBY is now recalculated Sep 23, 2010 using SMF70PAT. Also, SMF70CIN was blank, now is not. Thanks to Don Deese, (CPExpert), Computer Management Sciences, USA. Change 28.223 Support for DB2 V10 Compressed SMF records has been added DFH$MOLS inside the existing EXITCICS member, which is the ASM EXITCICS source that creates a SAS INFILE EXIT named CICSIFUE. Sep 23, 2010 Once assembled into a load module placed in //STEPLIB, and enabled with %LET SMFEXIT=CICS; in your //SYSIN, the exit will be used for either CICS or DB2 records, processing either compressed or uncompressed SMF records. Full documentation is in the comments in EXITCICS member. While IBM CICS provides their DFH$MOLS program that will decompress CICS records (JCL example in that MXG member), IBM initially chose NOT to provide a DB2 decompression utility, but in Feb, 2011, APAR PM27872 announced that the DSNTSMFD sample program was available to decompress. Note: IF THE EXIT DOES NOT EXIST, INTERNAL SAS CODE IN MXG WILL BE USED TO DECOMPRESS EITHER CICS OR DB2 DATA, BUT THAT CODE IS EXTREMELY CPU EXPENSIVE. MXG messages warn you to instead use EXITCICS. Note: See MXG Newsletter FIFTY-SEVEN for comparisons processing compressed and uncompressed CICS data with DFH$MOLS versus EXITCICS versus SAS code. Thanks to Rich Anderson, SAS Institute, USA.
VMACDB2 The circumvention is to insert this statement in SYSIN: Sep 22, 2010 %LET EPDBOUT= _SDB2ST4 ; so that the DB2STAT4 dataset is sorted before _SDB2STS is executed. _SDB2STS creates DB2STATS from STAT0/1/4; The MXG logic in _SDB2STS was revised in 28.04, and the PROC SORT DATA=_LDB2ST4 that had been in _SDB2STS was removed, because MXG executes _SDB2ST4 before _SDB2STS, so that sort was no longer needed. But ITRM did not execute the _SDB2ST4 macro. In ITRM, the dataset tokens _WDB2ST4 and _LDB2ST4 are the same dataset when _SDB2STS is executed, so that old PROC SORT DATA=_LDB2ST4 had put DB2STAT4 in the right order, so there was no error, but without MY sort, the unsorted DB2STAT4 exposed my error. -ITRM Support will have a Hot Fix to insert the _SDB2ST4 sort prior to the _SDB2STS execution; fortunately, this circumvention AND that hot fix insertion can coexist.
VMAC116 is harmless and there is no loss of data when the records Oct 7, 2010 originally described below are found. IBM MQ Support has confirmed Qsegment records with no WQ Segments can happen in V6 and V7 if a thread has exactly 9 queues. The next overflow SMF record/segment is allocated in anticipation of the transaction a 10th queue. With exactly 9, a 10th wasn't opened, so the second segment was left empty. Oct 20, IBM APAR PM24302 confirms that statement, quoting from my original change text in the problem description.
due to SMF 116 subtype 2 records that contained only the WTID (thread identification) segment and did NOT contain any WQ (queue-level accounting) segments. The ONLY purpose of subtype 2 records is for additional WQ segments that wouldn't fit in previous subtype 1 record. In addition, the QWHCXTYP field contained a zero, which is not a documented connecting system type code. But, MXG printed that message because it expected the WQ triplet in the subtype 2 and unconditionally input it. The code was restructured to input the common header and use QWHSNSDA, the number of triplets, to conditionally now input the triplets. Subtype 2 records with no WQ segment can be detected because all WQxxxxxx variables in dataset MQMACCTQ will be blank or missing values, but the "INVALID" "INVALID WQ SEGMENT" message is gone. Thanks to Victoria Lepak, Aetna, USA.
ANALDB2R revised to create consistent names, as well as adding ASUMDB2A new DB2 V9 and DB2 V10 variables and externalizing the ASUMDB2B interval value: ASUMDB2G -Default interval for both statistics and accounting is ASUMDB2P now set by macroUvariable &INTASUM (default=HOUR), so ASUMDB2R youTcanDchangeTthatBdefaultSinterval externally, without ASUMDB2S EDITing the ASUMxxxxSinto yourDUSERID.SOURCLIB, using ASUMDBAA %LET INTASUM=yourchoice;SUtoBset your chosen interval. ASUMDBSB -Statistic dataset changes: ASUMDBSS -Datasets ASUMDBBS/TRNDDBBS renamed to ASUMDBSB/TRNDDBSB. TRNDDB2A -Datasets ASUMDB2S/ASUMDB2G/ASUMDB2B/ASUMDBSB that were TRNDDB2A created by members ASUMDB2S/ASUMDB2G/ASUMDB2B/ASUMDBSB TRNDDB2B no longer exist; datasets TRNDDB2X/TRDDDBBS/TRNDDB2S TRNDDB2G no longer exist. TRNDDB2P -Members ASUMDBSS/TRNDDBSS create these sets of the three TRNDDB2R statistics summary datasets: TRNDDB2S Member Dataset Created from TRNDDBAA ASUMDBSS ASUMDBSS DB2STATS TRNDDBSS " ASUMDBSB DB2STATB Oct 3, 2010 " ASUMDBSG DB2GBPST Dec 28, 2010 TRNDDBSS TRNDDBSS ASUMDBSS " TRNDDBSB ASUMDBSB " TRNDDBSG ASUMDBSG -Accounting dataset changes: -Members ASUMDB2S/ASUMDBSB contain only comments that the member ASUMDBSS should be used in place of both members. -Members ASUMDB2G/ASUMDB2B were a mess. Previously they summed DB2GBPST/DB2STATB into ASUMDB2G/ASUMDB2B datasets but those names were also used for DB2ACCTG/DB2ACCTB summarization. Now they process only Account datasets. -WEEKBLD's BY list for these three DB2 Stat ASUMs have SHIFT removed; those datasets are summarized by HOUR so SHIFT was not used, but was inconsistently kept, so its removal prevents NOT SORTED errors. -Accounting dataset changes: -Member VMXGDBAA, previously used in ASUMDBAA/ANALDB2R no longer is used and is archaic. Member ASUMDBAA can still be used, if you want ALL of the DB2ACCTx datasets summarized, but it now is only a series of individual %INCLUDEs of these members: Member Dataset Created from ASUMDB2A ASUMDB2A DB2ACCT ASUMDB2B ASUMDB2B DB2ACCTB ASUMDB2G ASUMDB2G DB2GBPST ASUMDB2P ASUMDB2P DB2ACCTP ASUMDB2R ASUMDB2R DB2ACCTR -Trending of individual ASUMDB2x Account datasets: Member Dataset Created from or TRNDDB2A TRNDDB2A ASUMDB2A DB2ACCT TRNDDB2B TRNDDB2B ASUMDB2B DB2ACCTB TRNDDB2G TRNDDB2G ASUMDB2G DB2GBPST TRNDDB2P TRNDDB2P ASUMDB2P DB2ACCTP TRNDDB2R TRNDDB2R ASUMDB2R DB2ACCTR -Additional corrections: -ASUMDB2A. Two variables QTXANPL and QXCASCDP were in the SUM list are moved to the MAX list. They were incorrect in datasets built by the old ASUM/TRND members. -TRNDDB2A. Looked for DB2ACCT first, but now it uses ASUMDB2A first if it exists. -VMACDB2 common variables added to new DB2ACCTR "Remote". -Dec 28: ASUMDB2R INTERVAL now set by &INTASUM. Thanks to Chuck Hopf, Independent Consultant, USA.
VMXGDUR statement blew away a Windows session so you never saw VMXGSUM the error message. Now, the error is printed and the Sep 22, 2010 invocation is ended without the ABORT. This error can Oct 7, 2010 only occur when testing a new VMXGSUM invocation; that DATETIME argument is always required, except when the INTERVAL=NONE argument is specified. -VMXGDUR, Oct 7: WARNING message for invalid interval= values updated To include all possible valid values. Change 28.218 Includes of ASUMDBAA/TRNDDBAA added in MXG 28.05 caused BLDSMPDB errors and were replaced by includes of ASUMDB2A/TRNDDB2A Sep 21, 2010 and ASUMDB2B/TRNDDB2B. Oct 5, 2010 -Oct 5: If AUTOALOC EQ NO and FIRSTRUN EQ YES then all of the directories needed are tested and created if the do not already exist. Thanks to Cletus McGee, ALFA Insurance, USA. Thanks to Mary Kay Pettengill, MSI, USA. Change 28.217 The CPUID section is now protected for a truncated ID=70 VMAC7072 subtype 1 record (created by B37 ABEND in SMF dumping). Sep 21, 2010 Other sections were already protected.
VMACTNG large positive or negative values. While the error is in Sep 16, 2010 the NSM cube itself, a possible circumvention is to us IF RH018001 GT 0 AND RH018002 GT 0 THEN DO; OUTPUT _WTRH018; /* RH018 , DEFINED IN VMACTNG*/ END;
in the EXTRH018 exit member; this will cause the DELTATM between intervals calculated by MXG to be twice DURATM, so these intervals can be identified. Thanks to Xiaobo Zhang, FISERV, USA. Change 28.215 NOTES: FIRST/LAST UNINITIALIZED with BAR=TIME had no GRAFLPAR impact as they are not used with that option, but they Sep 15, 2010 are no longer printed. Arguments DEVICE/DISPLAY were removed, as your current site defaults will be used. Thanks to Cletus McGee, ALFA Insurance, USA.
ANALDB2R Starting with MXG 28.05, ANALDB2R passed those selection Sep 14, 2010 criteria to READDB2 so that if PDB=SMF was used (so that READDB2 was called) BEGTIME/ENDTIME were used for select. Now, PDB=PDB or PDB=SMF honors BEGTIME/ENDTIME selection. Thanks to Betty Wong, Bank of America, USA. Change 28.213 This Change was replaced by 28.220. Change 28.212 -New (very small) TYPE74ID dataset is now created when the ANALRAID TYPE74 or TYPE74CA are output, so the mapping format that ASUMCACH ASUMCACH requires can be automagically built in ASUMCACH, EXTY74ID eliminating a full pass of TYPE74CA dataset. VMAC74 -PDB.ASUMCACH now contains variables character DEVMODEL VMXGINIT and its numeric version in variable MODEL. VMXGSUM -ANALRAID now uses 8-character DEVMODEL, instead of MODEL Sep 13, 2010 with 3-byte numeric in HEX6 formatted value, to support Oct 29, 2010 newer, longer (3390A) control unit model names. -If TYPE74ID is does not exist, ASUMCACH uses TYPE74CA, as before, but DEVMODEL will contain 3390??. -In testing this change, numerous Missing Value messages were printed on the SAS log, all from VMXGSUM NORMn= arguments that had valid missing values. VMXGSUM was revised to test and suppress those cosmetic messages. -In testing this VMXGSUM revision for the NORMn arguments: -It was discovered that NORMn=A/B syntax, with only one variable A that had only one character name, was never being normalized. Fortunately, there was no instance of this syntax in MXG code, but it is now corrected. -Also, if B was missing or zero (with any length "A"), so the division was never executed, the value output for "A" output was SUM(A). Now, A is a missing value. -Oct 29: %VMXGOPTR relocated to prevent warning message. Change 28.211 Using COPYONLY argument wrote ALL SMF records; the READDB2 selection criteria were not used for the COPYONLY writes. Sep 11, 2010 Thanks to Larry Stahl, IBM Global Services, USA. Change 28.210 DB2 V8.1-DB2 V9, MXG 28.02-28.05, DB2ACCT and DB2ACCTP, VMACDB2 these variables added by APAR PK62161, MXG Change 28.051: Sep 10, 2010 QXRWSFET QXRWSINS QXRWSUPD QXRWSDEL were wrong because the prior field, QXSTXMLV, was increased to 8 bytes but that increase was overlooked, causing the subsequent fields to be misaligned/trashed. Thanks to Betty Wong, Bank of America, USA.
UTILBLDP the PDB.JOBS/STEPS/PRINT/NJEPURGE/SPUNJOBS datasets, Sep 9, 2010 the TYPE30xx,TYPE6,TYPE26J2 inputs for BUILD005 were Jul 7, 2011 sorted into the PDB library, which caused BUILD005 to fail, as it expected inputs to be in WORK library. But since the purpose of INCLAFTR=BUILD005 should be to create the above PDB datasets, this change removes the sort of those input datasets into the PDB data library. Jul 2011: If you really do want any or all of those raw TYPE30xx, TYPE6, and/or TYPE26J2 to be kept in the PDB, you can add these PROC SORTs to the EXPDBOUT= argument: PROC SORT NODUP DATA= _WTY30U1 OUT= _LTY30U1 ; BY _BTY30U1 ; PROC SORT NODUP DATA= _WTY30U4 OUT= _LTY30U4 ; BY _BTY30U4 ; PROC SORT NODUP DATA= _WTY30U5 OUT= _LTY30U5 ; BY _BTY30U5 ; PROC SORT NODUP DATA=_WTY6 OUT=_LTY6 ; BY _BTY6 ; PROC SORT NODUP DATA= _WTY26J2 OUT= _LTY26J2 ; BY _BTY26J2 ; Or, if INCLAFTR=BUIL3005, for TYPE26J3, use: PROC SORT NODUP DATA=_WTY25 OUT=_LTY25 ; BY _BTY25 ; PROC SORT NODUP DATA= _WTY26J3 OUT= _LTY26J3 ; BY _BTY26J3 ; Change 28.208 MXG incorrectly set R7451TIM to 16 microseconds when VMAC74 it should have been 16 milliseconds; the variables Sep 9, 2010 R7451CT3, R7451CT4, R7452PRT, R7452PWT are divided by R7451TIM, and thus were wrong. Thanks to Rick Southby, Insurance Australia Group, AUSTRALIA.
VMAC102 offset of '7532'x for QBMCRQLO in APPTUNE Segment 9 for Sep 7, 2010 Long Names, causing INPUT STATEMENT EXCEEDED ERROR. Thanks to Christa Neven, KBC Global Services, BELGIUM. Change 28.205 Julian Date variables with FORMAT 6. were increased to 7. VMACACF2 so the full seven digits can be seen (e.g. 2010099): VMACEPIL -VMACEPIL - OMJULDAT VMAC6367 -VMACACF2 - LIDCDATE LIDIPDAT LIDADATE LIDDXPDT VMACHSM - LIDATIME is now a datetime variable when both Sep 6, 2010 LIDADATE and LIDATIME are GT 0. Sep 18, 2010 -VMACHSM - MCDJDATE -VMAC6367 - DSCRDT DSEXTD Change 28.204 Support for local installation optional USERSTR field in IMACICUM CICSTRAN dataset from SMF 110 subtype 1 modified records. IMACEXCL VMAC110
Sep 3, 2010 Change 28.203 Variable LIDATIME was not formatted, but now contains VMACACF2 the datetimestamp value (DATETIME21.2) of last access. Sep 1, 2010 Date variables LIDCDATE LIDIPDAT LIDADATE LIDDXPDT all contain Julian Dates YYYYDDD and are cannot be formatted Yüklə 28,67 Mb. Dostları ilə paylaş:
|