* copyright (C) 1984-2019 merrill consultants dallas texas usa
Yüklə 28,67 Mb.
|
|
Change 20.164 This Analysis example caused sort errors that have been ANALHSM corrected by this change. Aug 16, 2002 Thanks to Steve Clark, California Federal Bank, USA. Change 20.163 Support for the additional four fields in subtype=15, FORMATS variables SMF15TIM, SMF15LTR, SMF15RSN, and SMF15MGT. VMACSTC Format $MGSTCVT was created to decode SMF15RSN reason. Aug 15, 2002 Thanks to Perry Lim, Union Bank of California, USA.
UTILRMFI CPU time, or to find out what is being put in 'OTHR') now Aug 15, 2002 has arguments identical to VMXGRMFI, so to run UTILRMFI reports with your definitions, just copy your tailored "%VMXGRMFI(..." text into your //SYSIN file, change that "%VMXGRMFI" to "%UTILRMFI", keeping all your tailoring arguments, and UTILRMFI reports will be produced using your definitions, with no retyping! Comments in UTILRMFI were revised. Thanks to Ada Phillips, University of Michigan Medical, USA. Change 20.161 -Support for ASG-Landmark TMON/DB2's Change in format of EXTMDBD0 the Version field. Originally a binary number, in 3.2 IMACTMDB it became an EBCDIC numeric, 'F3F2'x for 3.2, which made VMACTMDB LMRKVER=24.3. Accidentally, all of the MXG tests worked, VMAC102 but MXG code now detects and inputs the new format. VMXGINIT -INPUT of the header fields starting with HDLEN was out of Aug 14, 2002 aligment because of undocumented fields (inserted by the assembler to maintain alignment) that are now skipped. -The new DBHCDUID, DBHCEUTX, and DBHCEUWN End User USERID, Transaction Name, and Workstation Name were observed to contain lower case characters, just FYI for testing, etc. -Dataset TMDBBD, Authorization Failure, 'BD' record, which is the data segment from IBM's SMF 102 IFCID=140, now has all of the QW0140xx variables created and kept, using an extract of the VMAC102 source code.
(This design change was made in Change 17.251, but was never clearly documented, especially the zero obs part): For the SMF 102 DB2 trace records containing SQL text: T102S063,T102S124,T102S140,T102S141,T102S142,T102S145 (or for the Landmark dataset TMDBBD QW0140TX variable), the SQL text variable that contains the SQL text QW0063ST,QW01242T,QW0140TX,QW0141TX,QW0142TX,QW0145TX was broken into 100 byte chunks, with the 1st 100 output in the T102Snnn dataset, and the rest "chunked" into the T102T063,T102T124,T102T140,T102T141,T102T142,T102T145 (for Landmark, TMDBBD0) "text" datasets that contain only the text and key vars. But SAS Version 8 supports long character variables, so with V8 and (the MXG default of) COMPRESS=YES enabled, only one observation is output in the T102Snnn dataset, with the entire text stored in the SQL variable, and the "Text" T102Tnnn datasets will contain zero observations. (With archaic SAS V6, or with V8 and COMPRESS=NO, MXG reverts back and "chunks" the test into the T102Tnnn. -Some incorrect EBCDIC conversions were found in the VMAC102 code that had impact only when the code executed under ASCII SAS were relocated and corrected, and then that code was used in VMACTMDB. Thanks to Richard Jay Schwartz, State Street Bank, USA.
VMAC102 to the _V102CMN list of variables that are kept in all Aug 14, 2002 T102xxx (DB2 trace) datasets, as this is safer than redefining the _V102CMN macro, in case MXG ever needs to add variables to that common list. Thanks to Dr. Alexander raeder, Itellium, GERMANY. Change 20.159 ASMTAPES ML-25 replaces ML-24; two sites had CPU loops in ASMTAPES MXGTMNT itself, and one noticed 0C4 Logrec records. The Aug 13, 2002 error was an unexpected fall-thru after recovery of a 0E0 Aug 30, 2002 or 0C4 ABEND in one of the two cross memory environments (one, during device scan, and the second during a check for device allocation for the subtype 4, each of which is protected by a separate recovery ESTAE). The recovery logic was revised to fix the CPU loop, but ML-25 now also eliminates MXGTMNT logrec entries for those 0E0, 0C4, and any other abends that occur in cross memory (because the called address space had already been terminated before our next 2-second sample). Any "real" abends in MXGTMNT will not be suppressed from logrec. Thanks to Steve Simon, Alltel, USA. Change 20.158 First PROC DELETE DATA=ZZAIXPTX is changed to _WAIXPTX. VMACAIX
Aug 13, 2002 =======Changes thru 20.157 were in MXG 20.04 dated Aug 12, 2002========= Change 20.157 MXG 20.04 QA cleanup: -Non-used or typo-s members EXAIXAIX,EXCICEJB,EXQCSCON, VMACAIX EXQCSDIS,EXQCSJOB,EXQCSPOO,EXT120JM,EXTMTCIN,EXZRBASH, VMACNPIP EXZRBDSI,EXZRBDVH, and EXZRBENH were deleted. VMACTMO2 -In _SAIXPTX and _SNPIP02 sort macros in members VMACAIX VMAC103 and VMACNPIP, typos WAIXPTX and _WNPIP02 were corrected Aug 11, 2002 to ZZAIXPXT and ZZNPIP02. -In VMACTMO2, variables QAPRCPUT/QAPRCPUC were misspelled in MONIPA KEEP list and were not kept, and duplicate variables in KEEP list were removed, and some VMXGTIMEs were added. -Datetime variables not in %VMXGTIME() calls were found in VMAC103/SERVSTRT
EXNTMODU used for application analysis, in new MODULE dataset. VMACNTSM The new %Disk Busy, Avg Disk Service Time, and Avg Disk VMXGINIT Queue Time counters created in NTSMF 2.4.4 are now Aug 9, 2002 supported in both the Logical and Physical Disk datasets.
VMXGCICI useful, since it sums all data for all executions of an Aug 9, 2002 APPLID, but at least now it works without a zero divide error, and comments suggest it is not likely useful. If you specify INTERVAL=HOUR, etc., you must have data with SMFSTREQ='INT', i.e., you must have interval records enabled by your CICS guru, and the actual interval must be less than or equal to the INTERVAL= value you use in your CICINTRV - IBM default interval is 3 hours. Check the SAS log, as MXG will print a warning if your INTERVAL request can't be honored. Thanks to Brian Keller, ConAgra Inc, USA. Change 20.154 MXG expected LDSKNAME to be a single letter, but now data VMACNTSM with HarddiskVolume1, etc., appear in a Win/2000 server Aug 9, 2002 that shares data on a Hitachi Data Systems 7700E array. (This is a fail-over cluster, and the system that is 'in control' has the expected disk letter as LDSKNAME, but on other servers in the cluster that are not in control, this longer name value is seen). The problem is that MXG kept LDSNAME in $8, which caused all of these separate disks to be combined as HARDDISK. Increasing the length of variable LDSKNAME to $16 should be sufficient to store the longer name value. Thanks to John Compton, Bristol and West, ENGLAND. Change 20.153 Documentation only. If you have multiple systems with IMACTIME different time zones sharing a common JES spool, and you Aug 9, 2002 failed to use TIMEBILD to convert those datetimes to a common zone, your //SPIN library will fill and nothing is output to the PDB.JOBS/STEPS/PRINT datasets, because MXG compares the type 26 JES purge JTRTIME/JENDTIME with the JINITIME from the type 30 (to ensure, only for restarted jobs, that the correct (final) job termination record has been found). When you realize this is why your SPIN library is growing, you first need to enable the TIMEBILD to correct future timestamps, but you can clean out old records in the SPIN file using the old IMACTIME member, which has always externalized the time test logic. Use IF IN26 AND IN30_5 THEN OKJOB=1; to completely suppress the time check part of PDB logic. This change only added comments to member IMACTIME. Change 20.152 SMF 118 Subtype 70 caused INPUT STATEMENT EXCEEDED error. VMAC119 Four MAX(nn,xxxLEN) were changed to MIN(nn,xxxLEN). Aug 8, 2002 Three were in subtype 70 and one was found in subtype 3. Aug 9, 2002 Also, variable IFHOMEIP was corrected in subtype 6. Also, variables FCLREPLY and FSLREPLY were changed from character variables to numeric variables, because they are the 3-digit numeric FTP reply codes in RFC959. Note that FCLREPLY is a four-byte binary value, with '000000E2'x for decimal 252, but FSLREPLY is a three byte EBCDIC numeric value 'F2F5F0'x for 250, so the two fields are input with different informats and the extra byte of FSLREPLY is skipped. Thanks to Bob C. Allen, John Deere, USA. Thanks to Jonathan M. Miller, John Deere, USA. Change 20.151 This example report now shows local time rather than GMT, ANALTCP offers new %MACRO arguments to control the reports, and Aug 9, 2002 made minor corrections. See comments in the member. Oct 15, 2002 Note that IBM still does not provide the Start Date in TCP data, so with only a Start Time of Day value, the elapsed time calculation will be wrong for multi-day sessions. Oct 15: Realignment of columns on TOTAL report and clean up of comments. Thanks to Jim Hayes, Huntington Services, USA. Thanks to Judy Doherty, Florida Legislature, USA.
CONFIGV9 20.04-20.05, the CONFIGV9 member had no SEQENGINE option, MXGSASV9 so the V9 default V9SEQ engine was used. But MXG 20.06 Aug 8, 2002 added SEQENGINE=V6SEQ when it was discovered that the V9 Oct 3, 2002 sequential engine still compresses tape dataset. Mar 1, 2003 -The JCL Procedure example member MXGSASV9 is the same as MXGSASV8, but points to the CONFIGV9 member, just in case you look for the MXGSASV9 JCL example. Note: Updated Oct 3, 2002 here, without the reason: The SEQENGINE=V6SEQ was added to CONFIGV9 because both the V8SEQ and V9SEQ engine compress tape datasets. SEQENGINE=V6SEQ is the only safe sequential engine. Do NOT use the default SEQENGINE=V9SEQ. See Newsletter FORTY-ONE, SAS Technical Notes, for V9. Change 20.149 If you enabled TIMEBILD/VMXGTIME to change timestamps, VMAC26J2 datasets TYPE26J2/TYPE26J3/PDB.JOBS had variables VMAC26J3 JRDRTIME JCNVTIME JCNETIME JSTRTIME JENDTIME JPRNTIME and Aug 8, 2002 JFINTIME incorrect if that event occurred on a system which has a different GMT offset than the SYSTEM of the purge event. Variable SYSTEM was used to get the GMT offset. Now, the correct event system variable is used (SYSREAD SYSCVRT SYSEXEC SYSOUTP) to get the GMT offset. Change 20.148 Multiple CMODNAME='USER' CMODHEAD='USER' fields are now UTILEXCL correctly supported, creating a single INCLUDE statement Aug 8, 2002 for the IMACICUS member, with the sum of CMODLENG from the multiple fields used for detection/protection logic. Previously, multiple INCLUDEs were created, with notes for you to remove those duplicates, but the length tests were wrong, so the output dataset was trashed. Thanks to Victoria Lepak, Aetna, USA. Change 20.147 Variable PLATFORM was added to all TNG datasets so you VMACTNG can identify the software version that created each Aug 7, 2002 observation in each dataset. Thanks to Gunner Jacobsen, Nordea Bank Sverige AB, SWEDEN. Thanks to Erling Andersen, SMT Data A/S, DENMARK.
ASMTAPES -Fix to eliminate filling logrec due to 0E0 ABENDs: Aug 7, 2002 the 0E0 Software ABENDs still occur when the mount is so fast that the ASID has gone away, and MXGTMNT still recovers, but now it prevents the unneeded and unwanted logrec record from ever being created. -Enhanced recovery logic. Previously, the 0E0 recovery routine would skip all remaining devices after a 0E0. Now it only skips the DEVNR with the 0E0. -Fix for high CPU loop and/or high SMF record write loop (only in ML-21 thru ML-23). We're pretty certain this is corrected, but please look closely and verify all is well with the CPU time used by MXGTMNT. -Correction for a pre-existing programming error in memory management (a "storage leak") in the recovery logic found while testing.
solution to the original problem of filling logrec shows this was a non-trivial redesign!
mounts, since some are as fast as tens of milliseconds. But reducing the 2 second sample pop to 1/2 second can improve the percent of mounts captured, and you can see if there was a significant cost in the CPU time of the MXGTMNT monitor (and let me know your results!). However, by using the ASUMTAPE program to merge TYPETMNT mount, TYPETALO allocate, and (always-there) TYPE21 dismount record to create the PDB.ASUMTAPE dataset, and use it instead of the PDB.TYPETMNT dataset, then your total mount activity will be captured; fast mounts will just have missing/blank values for MNTTM (and other variables only available when the mount is captured).
missed mounts actually are, we may begin to investigate if there is any alternative to sampling, but at the very best, that would be a many-month, many tests, redesign. Let me know your thoughts. Thanks to my new ASMTAPES support person, who chooses to be anon. Change 20.145 SAS Version 9 Beta prints a warning when it finds syntax VMAC74 with a quoted string that is not followed by a blank: Aug 6, 2002 NOTE 49-169: THE MEANING OF AN IDENTIFIER AFTER A QUOTED STRING MAY CHANGE IN A FUTURE SAS VERSION.... Only one instance was detected in MXG, in VMAC74, where a blank was needed before the OR in IF R74ETYPE='B'OR .... A blank was not previously required, and even with V9, the program compiled correctly, but the warning will let you identify and correct any similar "unwise" syntax in your SAS programs, so you can revise your code now, long before SAS tightens up the syntax requirements. Thanks to MP Welch, SPRINT, USA. Change 20.144 MXG incorrectly read NTSMF PROCESS object records with VMACNTSM NRDATA=20 or NRDATA=27 (NTSMF 2.2.2 with NT 4.0 or 5.0). Aug 6, 2002 Variable CREATPID was not INPUT, causing the next three three variables (POOLPGBY, POOLNPBY, and HANDLES) to be wrong, and for NRDATA=20, the new READYTHR variable was not INPUT. The IN() test for CREATPID now tests for both 20 and 27, and the IN() test for READYTHR now tests for 20 to correct these errors. Thanks to Xiaobo Zhang, Insurance Services Office, USA.
could attend the Terrapin Station concert at Alpine Valley, Wisconsin, the first reunion of "The Other Ones" as the remaining four originals from "The Grateful Dead" now call themselves. Fantastic performances, peace and love among deadheads prevailed; it was a "grate show".
VMACNPIP the documentation for BYTESSNT and BYTESRCV variables in Jul 30, 2002 dataset NPMIP02; previously "ACCUMulated", now "DELTA" values, so the DIF() logic in the _SNPIP02 Sort Macro was removed.
VMXGSUM only appeared in the SUMBY= argument, because VMXGSUM Jul 30, 2002 used the original SORTBY= string as the SORTEDBY= value. (For example, if variable OPERATOR was not in CICSTRAN, your TRNDCICX now failed). Code was relocated so now the variables that actually exist are used for both SORTBY= and SORTEDBY= values, protecting dropped variables. Thanks to John Gebert, Office Depot, USA. Thanks to Justin Peterson, Office Depot, USA.
VMACMPLX five new datasets for Protocol Monitor of FTP, NTT, EE, Jul 26, 2002 OSPF, and MIBs. No SMF records from new version have Feb 3, 2003 been tested. Feb 3: This change supports Version 2.3. Change 20.140A Actual data values for SMF70CSF,SMF70ESF show they are in VMAC7072 megabytes and not frames, so their 4096 multiplier needs Jul 26, 2002 to be 1048576. IBM acknowledged the documentation error. Thanks to Allan Winston, MBNA, USA. Change 20.140 MXG 20.02-20.03 only. ALL Landmark TMON/CICS datetimes VMACTMO2 are wrong, if your GMT Offset value is non-zero. MXG Jul 25, 2002 Change 20.072 incorrectly assumed that TODSTAMP values were in GMT and incorrectly added the GMTOGMTO offset to every datetimestamp variable. Close examination of data with non-zero GMT offset shows all timestamps are local, except for TMMDCCLK, which is not kept nor converted, so all of the xx=xx+GMTOGMTO; statements were removed. -Variable TKLSTTOD was incorrectly treated as a duration, but now is input and formatted as a datetimestamp. -The one statement GMTOGMTO=GMTOGMTO/4096; was deleted. Thanks to Dean Ruhle, J C Penny, USA. Change 20.139 If your GMTOFFTM is non-zero, most datetimes in TMON/DB2 VMACTMDB will be wrong; Change 19.288 assumed all STCKs were GMT, Jul 25, 2002 but data shows that only REGNTIME, DBKSCB/DBKSCE, and (only for LMRKREC=:'B') HSSTCK and HDTSTP are GMT and now only those variables have +GMTOFFTM conversion to local. Thanks to Martin Legendre, Regie des Rentes du Quebec, CANADA. Change 20.138 CIMSDBDS dataset had obs with DBORG='00'X that should not VMACCIMS have been output. Those DBD segments are now documented Jul 22, 2002 as dummy, when DB2 was called but no DB2 data collected; previously, documented as ALLDBS Sync Point Writes. This change no longer outputs to CIMSDBDS if DBORG='00'X AND FLGOVERF='00'X, added to ensure that only dummy segments are skipped. OBSERVATION COUNT CHANGE: CIMSDBDS has fewer obs. Note: 02Jan02: See Change 20.306 revision.
VMXGPRAL had minor corrections to eliminate a note about TITLEs, Jul 22, 2002 and to reset OBS=MAX after its execution, and defaults were changed so %VMXGPRAL(DDNAME=WORK,NOBS=10); now will only PROC PRINT the first ten obs of each dataset in the work file that had any observations, which I find quite useful in validating datasets I've just created.
VMACQACS %LET DDNAME=QAPMPOLL; to read from the correct file. Jul 18, 2002 Thanks to John Gebert, Office Depot, USA. Change 20.135 An 0C4 ABEND reading RMFVSAM was caused by a bad branch ASMRMFV in old code in ASMRMFV, which was exposed only if the CSR Jul 18, 2002 Common Storage Remaining table was (unexpectedly) empty. Jul 22, 2002 The CSRG3 table was empty because the DIAG member in parmlib was set to OFF on one LPAR, but the correction protects for an empty table. Thanks to Jerry for his ASM skills, and to his management for allowing him to examine Betty's dump to diagnose and correct the error. And an error if only ASI was selected was also corrected. Thanks to Jerry Urbaniak, Acxiom CDC, USA. Thanks to Betty Wong, Bank of America, USA.
VMAC42 so Buffer Type (NSR/RLS/LSR/GSR) values were incorrect. Jul 18, 2002 The corrected decoding of VSAM Buffer Type is: S42DSBUF=FLOOR(INPUT(S42DSFL1,&PIB.1.)/64); Thanks to Don Deese, (CPExpert), Computer Management Sciences, USA.
VMAC26J2 and to be consistent with other instances of TYPETASK. Jul 18, 2002
VMACNTSM like DCOLLECT for NT! The new FS Statistics object will Jul 17, 2002 capture the size, type, name, owner, etc., of all or some NT disk files with the create, last access, and last use datetimes, so you can measure and chargeback disk space. The new dataset FSSTATS will have an observation for each file that was captured (you can set thresholds on size, etc., to determine if all or only some files are reported by this new collector.
VMACNTSM tested to be non-missing, to suppress the Missing Values Jul 17, 2002 SAS Note if you do not collect that field in NTSMF data. Thanks to Bob Gauthier, Albertsons, USA. Change 20.130 Support for APAR PQ61349 adds SERVSTRT (HTTP Server Start VMAC103 Datetime) to both TYPE1031 and TYPE1032 datasets, and Jul 16, 2002 JOB and ASID to TYPE1031 dataset. When not running in single server mode, the entity name is not unique, these fields provide unique tokens so the subtype 1 and subtype 2 can be merged together. Change 20.129 Support for APAR OW54007, which adds SMF42ESA (UCBSTAT) VMAC42 and SMF42EFL (UCBSFLS) variables to TYPE42AU dataset. Jul 16, 2002
ANALDB2K DB2ACCT Accounting variables and creates dataset DB2ACCTK EXDB2ACB with both Package and Accounting data. Comments in the EXDB2ACP ANALDB2K document what was discovered about Packages. VMACDB2 -This analysis uncovered errors in MXG; observations were Jul 16, 2002 output into DB2ACCTP and DB2ACCTB for Parallel RollUp records that were not output in DB2ACCT (See discussion in text of Change 19.027). The logic added to EXDB2ACC to delete if DB2PARTY='Y' AND QWACPARR='Y' was added to the EXDB2ACB and EXDB2ACP members so all rollups will be deleted. This required relocation of code inside VMACDB2 so those variables would exist for those exits. OBSERVATION COUNT CHANGE: DB2ACCTP has fewer obs. OBSERVATION COUNT CHANGE: DB2ACCT has fewer obs. Thanks to Nicholas Ward, Northern Territory Government, AUSTRALIA.
FORMATS going to be provided by IBM (!!!), but Al used Cheryl's Jul 16, 2002 input and his best guess to estimate an MSU capacity in this update to the MXG format $MG070CP, used to create the MSU capacities in dataset RMFINTRV, for the 7060 Yüklə 28,67 Mb. Dostları ilə paylaş:
|