* copyright (C) 1984-2019 merrill consultants dallas texas usa
Yüklə 28,67 Mb.
|
|
changes have NOT been validated yet: PRCCUP SYTCUG SYTCUM SYTCUP Thanks to Diana L. Bodner, Progressive, USA. Change 35.091 -Sometimes failed with a two level dataset name (unknown ANALCNCR cause). VGETOBS logic modified. Now will run SGPLOT Apr 22, 2017 if your SAS version is GE 9.3. -New example to count/plot concurrent TELNET sessions from the TYP11921 dataset.
VMACOPSS These datasets are created: Apr 22, 2017 May 9, 2017 DDDDDD MXG MXG DATASET DATASET DATASET SUFFIX NAME LABEL OPSS01 TYPOPSS1 SS TERMINATION SUMMARY OPSS02 TYPOPSS2 SS OSF SERVER TERMINATION OPSS03 TYPOPSS3 SS AOF RULE DISABLEMENT OPSS04 TYPOPSS4 SS GLOBAL VARIABLE OPSS05 TYPOPSS5 SS SQL STATISTICS OPSS06 TYPOPSS6 SS IMS BMP STATISTICS OPSS07 TYPOPSS7 SS OSF TRANSACTION OPSS08 TYPOPSS8 SS EPI STATS Thanks to Bruce Sloss, PNC, USA.
IHDRNDM which NDM Record Types are output with this logic: VMACNDM //SYSIN DD * VMXGINIT %LET MACNDMH= %QUOTE( IF NDMRTYPE='CT';) ; Apr 19, 2017 %INCLUDE SOURCLIB(TYPSNDM);
BUILDPDB can free the SMF allocation when SMF read is completed Apr 15, 2017 with this tailoring in your //SYSIN: %LET EPDBOUT=%QUOTE( FILENAME SMF CLEAR; ); If your SMF data is a GDG, this will unblock the base GDG name as soon as possible. -If running MXG on z/OS, add FREE=CLOSE to the //SMF DD to also free the allocation when the read is complete.
ANALFTP GENERATE REPORTS FROM PDB Apr 14, 2017 GENERATE REPORTS FROM SMF GENERATE REPORTS FROM SMF AND STORE DATA IN PDB GENERATE REPORTS FROM SMF LOOKING FOR A USER GENERATE REPORTS FROM SMF LOOKING FOR A DATASET Reports are from TYPE119 records; see also ANAL119 and ANALCNCR for additional reports. Change 35.086 New variable FSBYTERATE='TRANSMISSION*BYTE*RATE' is added VMAC119 to TY119070 dataset. ANALFTP Apr 13, 2017 Change 35.085 Possible exposure with too long a code line generated by VMXGPRNT VMXGPRAL print with variable name and label as heading. Apr 13, 2017 With 32 character variable name and 40 character label, the line generated could be 109 characters, exceeding the z/OS limit of 72 (S=72,S2=72). Two lines are now created and the label truncated (no more than 5 lost) to trim if needed.
UTILCOPY undefined macro variable NUMMEM. Now it tells you that Apr 10, 2017 it did not find anything to copy.
VMAC102 now correctly converted to a datetime value. Apr 10, 2017
Apr 2, 2017 Change 35.081 DB2ACCTP dataset, these "truncated" variables VMACDB2 QPACLOCN QPACCOLN QPACPKID QPACASCH QPACAANM Apr 2, 2017 were increased to $128 LENGTH, but the longer length text was not input when QPACOFFn was non-zero due to incorrect circumvention for prior invalid length in Change 31.015. Thanks to Rachel Holt, Fidelity Systems, USA. Change 35.080 *New z/OS 2.2 Changes found in Jan 2017 SMF Manual. VMAC30 -VMAC30. Apr 1, 2017 New variable SMF30JF1='JOB/SESS*ID' *This change is incomplete. Change 35.079 Some accumulated z/VM 6.3 SMT fields in VXSYTPRP dataset VMACVMXA weren't deaccumulated, and the below new unaccumulated Mar 31, 2017 counters contain an error code '80'x in first bit when Apr 5, 2017 the counter cannot be populated that is now decoded and Apr 11, 2017 the first two instances of each error is printed on the FORMATS SAS log, although there is nothing you can do for these these error conditions, and the variable is set to a missing value for these intervals. SYTPRP_CAL_CAPBYTYPE SYTPRP_CAL_MAXCAPBYTYPE SYTPRP_CAL_MTUTILBYCORE SYTPRP_CAL_MTUTILBYTYPE SYTPRP_CAL_PRODBYCORE SYTPRP_CAL_PRODBYTYPE -Dataset VXAPLSL0 was "hosed" because my loop was DO CPUNR=1 TO NRCPUS, but first CPU is CPUNR=0, and the CPUNR at the end of the segment had been overlooked. -PFXPRKWT is now deaccumulated. -The VXAPLSLx dataset only has observations output when there was activity by the Linux machine; the LINXTIME is the "wake up" time in this interval and is used to create DELTALINXTM=MRHDRTOD-LINXTIME with the maxiumum duration of an interval that that VMDUSER could have been active. -These variables added to VXMTRSYS: RCCCCAPF RCCSCAPF SYSMTFLG RCCMTRSM RCCMTCFM RCCMTPMT RCCMTTDW RCCMTFRS RCCCOMXT RCCCOALL CAL_RCCACMNT1-4='CPUTYPE-1-4*ACTIVATED*THREADS' CAL_CPUTYPE1-4='CPUTYPE-1-4*CPU*TYPE' CAL_RCCSYMNT1-4='CPUTYPE-1-4*MAXIMUM THREADS*SOFTWARE' CAL_RCCHWMNT1-4='CPUTYPE-1-4*MAXIMUM*THREADS*HARDWARE' CAL_RCCCOMNT1-4='CPUTYPE-1-4*REQUESTED*THREADS' CAL_RCCCRMNT1-4='CPUTYPE-1-4*REQUESTED*THREADS' CAL_RCCSMMNT1-4='CPUTYPE-1-4*SPECIFIED*THREADS' -These variables added to VXSYTCUP: LCXCMTIT='MT*IDLE*TIME' LCXCHPCP='LPAR GROUP*ABS CAPACITY*CAP VALUE' CALGCAPV='ABSOLUTE*CAP*AMOUNT' -Variable RDEVCTRG is removed from the DIDIO test to create VXIODDEV because it is always non-zero and the only non-zero field for non-DASD records. -The SMT Busy Time calculation was revised to LCXCMTBY=(2*LCUCACTM)-LCXCMTIT Thanks to Graham Harris, RBS, ENGLAND. Change 35.078 If you specified SORTEDBY=YES and you run the daily and BLDSMPDB weekly/monthly processing in the same SAS execution of Mar 31, 2017 BLDSMPDB the SORTEDBY= was ignored because VMXGSUM used the same MACRO variable name making it a GLOBAL macro variable to VMXGSUM and overriding the BLDSMPDB used parameter. NO data was lost - just not in the order you may have expected. Circumvented by holding the value in a LOCAL macro variable and the reinstating it after DAILY processing is completed.
UTILBLDP and a redundant WANTSMF in one example was removed. Mar 31, 2017 Thanks to John Compton, WPS, ENGLAND. Change 35.076 Some long RMF III variables have '00'x null characters VMACRMFV instead of blanks at the end; they are converted to Mar 29, 2017 blanks.
FORMATS datasets with a Data Class. For records without subtype 5 VMAC1415 the below flag variables are now set to 'FF'x and will Mar 29, 2017 print "NOT AVAILABLE" with new $MGNOTAV format: SMF14BFG SMF14FLGS SMF14FLG2 DEB2XUPF EADSCBOK DCBEEX31 XTIOTYES Variable SMF14ALIAS will be blank and SMF14LBS will be a missing value. Thanks to Michael Oujesky, DTCC, USA.
EXXAMPRC 30 (SEGLEN should be 684) are valid and a problem is open EXXMVPID with Barton. FORMATS -New VSIPID Process Segment creates XMVISPID dataset. IMACXAM -New PRCCPU LIMPOOL Segment creates XAMSYPRC dataset, but VMACXAM values for FLAGSPRC have '60'x for both limited cpuaffon. VMXGINIT Mar 28, 2017 Change 35.073 DB2ACCT variables QWACALOG and QWACALCT are now always VMACDB2 missing values; they have been reserved for years, but Mar 28, 2017 MXG code had still INPUT them causing confusion. Thanks to Peter Gray, HPE Australia, AUSTRALIA. ====== Changes thru 35.072 are in this MXG 35.03 dated Mar 27, 2017===== Change 35.072 -MXG 35.03. VMAC1415, hex dumps but no error, because VGETJESN line 962 in VMAC1415 IF VOLSER NE VOLSER1 THEN LIST; VMAC1415 left from debugging needs to be deleted. Mar 24, 2017 -WARNING TYPETASK NOT DECODED, JCTJOBID=A0000022 expected Mar 27, 2017 either 'ASCH' or 'OMVS' in to be stored in TYPETASK from variable SUBSYS, but this task has SUBSYS blank. Now, if SUBSYS is blank, TYPETASK='APPC' is stored. Thanks to Paul Naddeo, Fiserv, USA. Change 35.071 Reserved Change Number. Mar 23, 2017 Change 35.070 New fields are added to zPROTECT SMF records: VMACZPRO New variable in ZPROT05: Mar 23, 2017 ZPRRAUSR='ALTERNATE*USERID' New variable in ZPROT16: ZPRTMINP='MIN*REQUEST*PERFORMANCE*TIME' ZPRTMAXP='MAX*REQUEST*PERFORMANCE*TIME' ZPRTAVGP='MEAN*REQUEST*PERFORMANCE*TIME' ZPRTAUSR='ALTERNATE*USERID' ZPRNPROT='PROTECT*OPERATIONS' ZPRNACCE='ACCESS*OPERATIONS' ====== Changes thru 35.069 are in this MXG 35.03 dated Mar 22, 2017===== Change 35.069 Support for CICS Version TS/5.4 Beta 11 adds three new UTILEXCL variables to CICSTRAN: VMAC110 LPARNAME='LPAR*NAME' Mar 21, 2017 MPSRACT='TIMES WHEN*POLICY*EVALUATED*AND TRIGGERED' MPSRECT='TIMES WHEN*POLICY*RULES WERE*EVALUATED' Thanks to Andy Wharmby, IBM CICS Hursley, ENGLAND. Change 35.068 Support for MQ Version 9.1 SMF 115 new Subtype 201 record EXTY115Y creates new dataset: FORMATS dddddd dataset description IMAC115 TY115Y MQ115201 MQ SUBTYPE 201 PAGESET STATS VMAC115 Mar 27: FORMATS MG115EX and MG115PS added. VMXGINIT
Mar 20, 2017 Mar 27, 2017 Thanks to Giuseppe Giacomodonato, EPVTECH, ITALY.
SMFINTRV needed to be &PDBMXG..SMFINTRV, but MXG invocations in Mar 14, 2017 BUILDPDB/BUILDPD3 were correct so no error there. Apr 18, 2017 -Three new interval START datetimes are created so you can directly summarize to that interval by your choice of START15INT='FIFTEEN*MINUTE*INTERVAL*START' START30INT='THIRTY*MINUTE*INTERVAL*START' STARTHRINT='HOUR*INTERVAL*START' Apr 11: Revised. Those three variables are now created by VMXGDUR, with the DEFAULT SMFINTSYNC59 "SYNC59" option defaulting to YES, to be consistent with RMFINTRV. If your records are NOT "SYNC59", i.e., are written :00 use %LET SMFINTSYNC59=NO; in your //SYSIN. Unlike other "INTRV" programs that invoke VMXGDUR to create new summary datasets, SMFINTRV does NOT summarize TYPE30_V records; instead it only combines the multiple SMF 30 records (MULTIDD='Y' steps with LOTS of DDs) into one observation per interval with all EXCPs totaled. If you do want to summarize PDB.SMFINTRV across intervals the ANALSMFI program provides that example. Change 35.066 APAR OA59593 adds new flag variable BUIL3005 SMF30CAS_INELIGHONOR='ELIG WORK*IS NOT*OFFLOADED*TO CP?' BUILD005 to SMF 30 TYPE30_V/TYPE30_4/TYPE30_V datasets to identify VMAC30 jobs whose eligible work was NOT offloaded to CPs for Mar 19, 2017 help. The variable is also added to PDB.SMFINTRV and PDB.STEPS datasets. Change 35.065 Almost "cosmetic": READDB2 could create dataset DB2STATR READDB2 even though you did not request it, due to Change 34.265 Mar 14, 2017 that overlaid a token for DB2NETZA. No obs were output.
VMAC7072 -ZIPACTTM in PDB.ASUMCELP is the best source of per-LPAR Mar 13, 2017 hardware zIIP CPU busy, created in BUILDPDB/ASUM70PR. This change adds SMT_NUM to PDB.ASUMCELP to identify the SMT mode. -Variable ZIPACTTM in PDB.TYPE70 could be too small for an LPAR when in SMT_NUM=2; in rare cases the last LPAR's was not included. -Correction for ZIPACTTM in TYPE70 also caused PLATBUSY LPARSHAC LPARSHAR TOTSHARC TOTSHARE ZIPSHARC ZIPSHARE LZIPSHAC LZIPSHAR to also be corrected/changed values in compares. -TYPE70/TYPE70PR variables CPUID SMT_CORE_ID LCPUADDR and (new) SMT_THREAD are now formatted numeric HEX2. to to match RMF reporting formats. -In the ASID TYPE30 and Service Class TYPE72GO data, the recorded MT=2 CPUZIPTM/ZIPUNITS values are "inflated" above the actual hardware zIIP time, and the hardware equivalent can not be calculated using R723MCFS, the Maximum Multi-Threading Capacity Factor. LPAR with 7 zIIP engines in SMT_NUM=2 MT=2 mode: UPTIME: 1:45 ZIPACTTM: 1:31 72-CPUZIPTM: 2:05 hh:mm 105 min 91 min 125 min "Above Inflation Factor" 125/91=1.37 Interval R723MCFS =1.17871 "MCSF Equivalent zIIP CPU= 125/1.17871 = 106 min but that is as large as the UPTIME of 105 minutes. And IBM's range of R723MCFS values is 1.1 to 1.4, with a theoretical max of 2.0.
now in SMF 30/72 records in MT=2 mode (AND ONLY in ASID and SRVCLASS records): NO INFLATED VALUES IN RMF 70 SMF data that are used for zIIP capacity metrics. So, while the values are too large, their interval sum can be used to determine the proportion of the MT=2 zIIP usage for each workload, job, or service class.
confirmed my conclusions that the values are inflated. August: IBM confirmed that at very high or very low util the SMT values are inflated; it's unclear if that will ever change.
Systems z13 Simultaneous Multi-Threading R(Evolution)" discusses the MT=2 metrics.
inflation of 30/72 CPUZIPTM when SMT Mode is enabled. Change 35.064A Multi-Volume DCOLDSET records populate some fields only VMACDCOL in the first (DCDVOLSQ=1) record. When TYPSDCOL program Mar 10, 2017 is used, these fields are retained from the first record and are now output in PDB.DCOLDSET. -Records with DCDVOLSQ=0 were created in WORK.TYPEDCOL but were then not output in the first record logic, but now they are output to PDB.DCOLDSET.
VMACXAM are wrong, protected by changing SYTNLPS to 5 while the Mar 10, 2017 problem is opened with Barton now to resolve. -XMTCPSYS dataset variable NAMENODE was blank because the 128-byte CONTACT was reduced to 64, then NAMENODE, then 64 bytes are inserted to keep the original SEGLEN. Thanks to Matthew L. Rennebohm, State of Wisconsin, USA.
VMACMVCI changes that caused INPUT STATEMENT EXCEEDED error. Mar 6, 2017 Thanks to DJ Chen, AST/Southwood Shared Resource Center, USA. Change 35.061 Enhancement for PDB.ASUMCELP (per-LPAR CEC data) adds the VMXG70PR variable SMT_NUM to identify the SMT Mode of zIIPs, from Mar 7, 2017 the PDB.TYPE70EN dataset, and protection if that dataset was not copied to the PDB data library. SMT_NUM will be a missing value for PHYSICAL and IFL-Only LPARS, or LPARS with no zIIPs. Change 35.060 Enhancement for SMF 120 Subtype 11 TYP120BL new variables VMAC120 containing TOTAL, CP ONLY, and ZIP ONLY CPU times: Mar 5, 2017 SM120BCPUTM='TOTAL*CPUTIME*USED*BCA1-BBZ1' SM120BCPCPUTM='CP ONLY*CPUTIME*USED*BCA2-BBZ2' SM120BZIPCPUTM='ZIP ONLY*CPUTIME*USED' are calculated, thanks to the IBM WebSphere Developer who educated me that the 16-byte binary TIMEUSED format used in this record for the START and END Accumulated CPU data contains two 8-byte CPU times: the TOTAL CP+ZIIP in the first 8 bytes and the CP ONLY CPU time in the second 8. Thanks to Giuseppe Giacomodonato, EPVTECH, ITALY.
VMACSVIE 32.170 has now been tested with actual data, and these Mar 4, 2017 variables had missing values that are now corrected. Mar 27, 2017 IMTR_CLK_SUBQ06_TIME IMTR_CLK_MPP_CPU IMTR_CLK_SMB_ENQ IMTR_CLK_CNT_ENQ IMTR_CLK_MXG_END IMTR_CLK_CNT_GU IMTR_CLK_UOW_END IMTR_CLK_UOW_START IMTR_CLK_SCHEDULE_TIME -Mar 23: Several TODSTAMP variables were not converted from GMT to local time zone. Thanks to Denise Williers, Wipro, USA. Change 35.058 Support for IMS LOG 67D0 DIAGNOSTIC RECORD for DC '02'x EX67D002 created new IMS67D002 dataset. Note that both D0TIME and FORMATS IMSSTCK are both GMT because there is no GMT offset in IMACIMS the 67D0 log record. VMACIMS VMXGINIT
Mar 2, 2017 Thanks to Rosa Maria Martinez Alonso, Bustia, SPAIN. Change 35.057 Reserved Change. Mar 1, 2017 Change 35.056 -If you used KEEPMNTH= (very rare) an MDY() could fail. VMXGSUM -If your INCODE= contains a DATA step, the CLASSNWAY Mar 1, 2017 option failed, but now a DATA step's existence in the INCODE is parsed, and if found, SUMBYCLASS is reset. Change 35.055 Support for Mainview for IP PTF BPN2331 that adds VMACMVIP variable TNDSTATX='CONNECTION*STATE*ACTIVE*CLOSED?' Mar 1, 2017 variable to the I490 dataset.
ADOCRMFV table character data filtering and usability. ASMRMFV -These filters are intended for building ad hoc MXG RMF VMACRMFV III PDBs for studies avoiding the overhead of generating Mar 1, 2017 a full OPD table-based PDB. They control which OPD table entries are output to the RMFBSAM file. -Five new filters are added to support OPD entry selection from this table to the RMFBSAM output file. These filters are effective only if the OPD table is selected.
different keywords are used. New Keyword Aliases ------------ ------------------------------------------ OPDPROCNAME= OPDPROCNA=, OPDPROCNM=, OPDPROC=, OSDPROC= OPDPN=
OPDJOBNAME= OPDJOBNA=, OPDJOBNM=, OPDJOB=, OPDJN= OPDUSERNAME= OPDUSERNA=, OPDUSERNM=, OPDUSER=, OPDUN= OPDAND None OPDOR None The order of OPD filter application is: 1) OPDPROCNAME= 2) OPDJOBNAME= 3) OPDUSERNAME= -TUTORIAL: Ranges of the form keyword=first:last may be used with any of the above keywords except OPDAND and OPDOR. The colon character ':' is required for a paired range specification. All entries GE the first value and LE the last value are selected for output to the RMFBSAM file. The first value may not exceed the last value in EBCDIC collating sequence or an error is flagged. Single unpaired values may be specified for a range simply as keyword=first and in this case the colon ':' is omitted. Patterns may also be used with any of the above keywords except OPDAND and OPDOR and include one or more Wild Card characters to match the respective OPD data field. A pattern contains one or more special Wild Card characters as follows: Wild Card Matches ---- ------------------------------------------------ * 0 or more characters % 1 Non-blank character + 1 Numeric character (0-9) _ 1 Alphabetic character or _ (a-z, A-Z, _) . 1 National character (@, #, $) ! 1 Special character (not a-z, A-Z, 0-9, @, #, $) ? A blank string if used by itself ? 1 Blank character (X'40') if used with any other characters Ranges may not be wild carded. If wild carded the range value becomes a pattern instead and is processed as such. See Section 25 "Ranges and Patterns" in the ADOCRMFV member or ASMRMFV source code for more details on usage of ranges and patterns. -OPDPROCNAME= selects OPD tables by 1-8 character z/OS JCL Procedure Name. Proc Name characters are validated to those allowed by JCL syntax. Both ranges and patterns with wild cards may be specified. Up to 16 ranges and 16 patterns are supported. The default is OPDPROCNAME=ALL. NOTE: There is only ONE OSDPROC field per OPD table. If the OPDPROCNAME= value does not match, then the ENTIRE OPD table with all entries is excluded. Use OPDPROCNAME= with care and discretion and only if the OSDPROC contents are well understood. Almost always OSDPROC is simply 'OMVS'.
-OPDJOBNAME= selects OPD entries by 1-8 character z/OS Job Name. Job Name characters are validated to those allowed by JCL syntax. Both ranges and patterns with wild cards may be specified. Up to 32 ranges and 32 patterns are supported. The default is OPDJOBNAME=ALL. Job Names must be 1-8 characters in length and may include any characters A-Z, #, $, or @. Numeric digits (0-9) may be used only after the first character. -Examples for OPDJOBNAME= : OPDJN=PROD1234:PROD5678 selects only address spaces with a z/OS Job Name GE 'PROD1234' and LE 'PROD5678' as a range. Note use of the keyword alias OPDJN for coding convenience. OPDJOBNAME=.* is a pattern that selects only address spaces with a Job Name that begins with a national character. OPDJOBNAME=*++ is a pattern that selects only address spaces with a Job Name that ends with 2 numeric digits. OPDJOBNAME=ABC:ABC88888 is a range that selects only address spaces with a Job Name that is GE 'ABC ' and LE 'ABC88888'. -OPDUSERNAME= selects OPD entries by 1-8 character z/OS User Name. User Name characters are validated to those allowed by JCL syntax. Both ranges and patterns with wild cards may be specified. Up to 32 ranges and 32 patterns are supported. The default is OPDUSERNAME=ALL. User Ids must be 1-8 characters in length (1-7 characters for TSO Ids) and may include any characters A-Z, #, $, or @. Numeric digits (0-9) may be used only after the first character. -Examples for OPDUSERNAME= : OPDUSERNAME=JOE8888 selects only address spaces with a login User Name of 'JOE8888'. OPDUSERNAME=JOE:JOE8888 selects only address spaces with a login User Name that is GE 'JOE' and LE 'JOE8888'. OPDUSERNAME=.* selects only address spaces with a login User Name that begins with a national character (@, #, $).
OPDUSERNAME=*++ selects only address spaces with a login User Name that ends with 2 numeric digits (00-99). -OPDAND (default) indicates that selection results from the two different OPD filter keywords are logically ANDed. -OPDOR indicates that selection results from the two different OPD filter keywords are logically ORed. Example 1 with OPDAND in effect: OPDJOBNAME=ABC* OPDUSERNAME=SAM* only selects Address Spaces in the RMF Monitor III OPD table that have a Job Name starting with 'ABC' AND a User Name beginning with 'SAM'. Otherwise the Address Space is filtered and will NOT appear in the result MXG PDB. Yüklə 28,67 Mb. Dostları ilə paylaş:
|