IRD & capacity planning
Hello, In the upcoming days we'll be activating Intelligent Resource Director in our systems (CSSPQ everywhere, and LPAR CPU management in the main production sysplex). Not being actually concerned about CSSPQ efects, as we don't expect any major changes (we've measured our SAP queues, being less than 1.5 in the maximum 15 minute average), we are about LPAR CPU management. Our capacity planning staff are currently basing their CPU utilization measurements and machine upgrades on the LSPR tables (LPAR mode with z/OS 1.4 as this is our current version). We are moving soon to z/OS 1.7, so we began to turn our eyes into IBM's zPCR tool. As IRD will be configuring processors ON/OFFline, uniprocessor effective capacity will change, and according to zPCR, this change will be more than noticeable. The problem that arises is: how can we measure CPU usage in an environment where CPU capacity may change over time? The main parallel sysplex (the one being IRD-fully-enabled) consists of 10 z/OS images running in four 2084 (soon to be upgraded to 2094) and one 2094, one or two LPARs each (total capacity reported by zPCR, based on a 450 MIPS 2084-301, being 33500 MIPS). Is it a good idea to use LSPR static tables for Capacity Planning in this environment? Is there any way to do it dinamically depending on the number of online logical processors? Should we use entire CEC's uniprocessor power as a basis for measurements? Thanks in advance -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: Job info: Wrong time in queue
Hello Kevin:
About this...
"I would expect to see the job number closer together were they both
submitted at the same time. These jobs are more than 2+ JOE apart. Are
they from different systems via NJE ?"
Yes, these jobs was submitted in different days.
I will try to explain the test in more detail:
Date: 1th March
1) To send some jobs (about 10 jobs)
2) To hold the jobs by hand
3) To free the jobs by hand
4) Quickly, to ask for job info
Date: 21th March
1) To send some jobs (about 10 jobs)
2) To hold the jobs by hand
3) To free the jobs by hand
4) Quickly, to ask for job info
At 1st test (1th March) we have obtained wrong time in queue (>23h!!) due to
these jobs were free (by hand) in few minutes from hold status. Therefore,
the correct time in queue should be few minutes. For example, we have post
the statistics of JOB01138
Then, at 2nd test (21th March) we have repeated the test following the same
steps. At this time, we have obtained correct time in queue (<2min). For
example, we have post the statistics of JOB24485.
So, with same test, we have obtained different results.
We think that it's no policy goal problem but it's a problem with SDSF job
information report.
We opened a PMR and following lab. steps we have try to reproduce it. Then
we have obtained 21th March report. We suspect some new PTF applied but we
haven't found anything.
You can try to test it:
1) To send some jobs (about 10 jobs)
2) To hold the jobs by hand
3) To free the jobs by hand
4) Quickly, to ask for job info
Jobs used:
/T751622E JOB CLASS=S,MSGCLASS=V,NOTIFY=&SYSUID
//P010 EXEC PGM=IKJEFT01,PARM='%LOOPSHRT'
//SYSTSPRT DD SYSOUT=*
//SYSTSIN DD DUMMY
//SYSPRINT DD SYSOUT=*
//SYSEXEC DD DISP=SHR,DSN=libname
//T751622E JOB CLASS=S,MSGCLASS=V,NOTIFY=&SYSUID
//P010 EXEC PGM=IKJEFT01,PARM='%LOOPSHRT'
//SYSTSPRT DD SYSOUT=*
//SYSTSIN DD DUMMY
//SYSPRINT DD SYSOUT=*
//SYSEXEC DD DISP=SHR,DSN=libname
//T751622E JOB CLASS=S,MSGCLASS=V,NOTIFY=&SYSUID
//P010 EXEC PGM=IKJEFT01,PARM='%LOOPSHRT'
//SYSTSPRT DD SYSOUT=*
//SYSTSIN DD DUMMY
//SYSPRINT DD SYSOUT=*
//SYSEXEC DD DISP=SHR,DSN=libname
...
--> libname(LOOPSHRT) <--
/* REXX */
/* "ALLOC DA('libname(LOOP1)') SHR F(I001)" */
/*DO I=1 TO 5
TEXT.I=SUBSTR('A',1,4000,'Z')
END*/
DO I=1 TO 500
TEXT.I=SUBSTR('A',1,4000,'Z')
END
A=1
B=2
T=0
DO 1500
/* T=T+1 */
C=A
A=B
B=C /*
IF T=500 THEN DO
T=0
"EXECIO 1 DISKR I001 1 (STEM A."
END */
END
...
--> libname(LOOP1) <--
/* REXX */
/* "ALLOC DA('libname(LOOP1)') SHR F(I001)" */
/*DO I=1 TO 5
TEXT.I=SUBSTR('A',1,4000,'Z')
END*/
A=1
B=2
T=0
DO WHILE 1=1
/* T=T+1 */
C=A
A=B
B=C /*
IF T=500 THEN DO
T=0
"EXECIO 1 DISKR I001 1 (STEM A."
END */
END
On 3/22/06, Clark, Kevin D, HRC-Alexandria/EDS <
[EMAIL PROTECTED]> wrote:
>
> With that being the case.
>
> Now is the policy goal high enough to cause WLM to start another INIT. On
> the report there is always 1 job active.
>
> Go for 00:00:30 goal at 90%
>
> -Original Message-
> From: Clark, Kevin D, HRC-Alexandria/EDS
> Sent: Wednesday, March 22, 2006 2:06 PM
> To: 'Christian Blesa'
> Subject: RE: Job info: Wrong time in queue
>
>
> Christian,
>
> I would expect to see the job number closer together were they both
> submitted at the same time.
> These jobs are more than 2+ JOE apart. Are they from different systems
> via NJE ?
>
> -Original Message-
> From: Christian Blesa [mailto:[EMAIL PROTECTED]
> Sent: Wednesday, March 22, 2006 1:52 PM
> To: IBM-MAIN@BAMA.UA.EDU; Kevin Clark
> Subject: Re: Job info: Wrong time in queue
>
>
> Yes, of course, these are 2 diff. jobs but both has been waiting for a
> init
> for 2min!! Therefore time in queue of 1138 was incorrect.
>
> Anyone could test it?
>
>
> On Wed, 22 Mar 2006 12:04:49 -0500, Clark, Kevin D, HRC-Alexandria/EDS
> <[EMAIL PROTECTED]> wrote:
>
> >What is the problem ?. These are two different jobs (1138 & 4485).
> >
> >1138 has been waiting for a initiator for 23 1/2 hours, is that the
> >issue ?
> >
> >Kevin
> >
> >-Original Message-
> >From: IBM Mainframe Discussion List [mailto:[EMAIL PROTECTED] On
> >Behalf Of Christian Blesa
> >Sent: Wednesday, March 22, 2006 11:34 AM
> >To: IBM-MAIN@BAMA.UA.EDU
> >Subject: Job info: Wrong time in queue
> >
> >
> >Hello
> >
> >We have had a "ghost" and we're finding it...
> >
> >
> >We're migrating from JES2 inits to WLM inits and in our test
> >environment we have done following test:
> >
> >1) To send some jobs
> >2) To hold the jobs by hand
> >3) To free the jobs by hand.
> >4) To ask for job info.
> >
> >1th March - Report
> >
> >Job Information
> > Job name T751622E Job class limit exceeded? NO
> > Job ID JOB01138 Duplicate job name wait? NO
> > Job schedulable? YESTime in queue 23:32:29 !!!
> > Job class mode WLMAverage time in queue 00:00:00
> > Job class held? NO Position in queue
Re: SYS1.IMAGELIB doubts!
Eric, I also think that only the catalogued SYS1.IMAGELIB will be used but I would like to see if our premise is true! I had revising proclibs/parmlibs/nucleus but I haven't found where SYS1.IMAGELIB is defined... Thanks On Wed, 15 Feb 2006 10:04:28 -0600, Eric Bielefeld wrote: >Chris, > >I'm pretty sure the answer to your question is that only the catalogued >SYS1.IMAGELIB will be used. I dug out my z/OS V1.4 CDs, and looked in >the System Dataset Definitions, which I gather you already checked. It >did say to catalog the dataset. That brings up a question that I had >while looking for this. > >Where is SYS1.IMAGELIB defined? Or does it just have to exist and be >catalogued with that name. I noticed that on our 1.2 system, it isn't >found in either proclib or parmlib. The System Dataset definition >doesn't mention how to make it a different name if you wanted to do >that. > >Eric Bielefeld >Sr. Systems Programmer >P&H Mining Equipment >414-671-7849 >Milwaukee, Wisconsin -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: JES2 proclibs allocated???
Yes but... Report of JES2 TIOT shows: JES2 JES2 JES2 PRTY 254 PAGE I/O'S 11,162 DD PROC00UNIT VOLUME sysres EXCP 376 UNIT VOLUME volse1 EXCP78 UNIT VOLUME volse2 And older lib03 is allocated in other volser (volse3)... -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: Rule of thumb about paging / page fault rate
On Fri, 23 Dec 2005 15:33:47 -0600, Mark Zelden <[EMAIL PROTECTED]> wrote: >The ROTs are fine (old and current), but there is really only one >thing that matters and you may be paging, or you may not be > >Are you meeting your SLAs? > >Mark Hello Mark, of course, to meet SLAs is the first and last break point! But... in our production sysplex we have installed between 14Gb to 27Gb and we think that, maybe, it's too bigger. This sysplex (7 "normal" systems + 2 GDPS) have IMS, DB2 and MQ. Regards Chr. -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Rule of thumb about paging / page fault rate
Good morning, we are revising central storage due to our systems not paging and, therefore, it's possible that we have installed too central storage. We have seen to find values or some rule of thumb but we don't find anything. At 1991 the recommendation was 300-500pag/sec of total paging but now this recommendation has been disappeared. Could you help us? Thanks -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: STRNO in DB2 catalog
Hi folks, as I mentioned later, I post IBM answer at PMR: DFSMS : Managing Catalogs explains in topic "Specifying the Number of Concurrent Requests" the paameter STRNO. This is the number of concurrent read requests for a BCS and can have from 2 to 255. >From this manual: . Initially define the catalog with STRNO(3). You can monitor the adequacy of this value by using the Resource Measurement Facility (RMF) to watch for enqueues on the resource SYSZRPLW.catname, and I/O contention on catalog volumes. SYSZRPLW is the major name of the resource, and catname (the name of the catalog) is the minor name. An indication of enqueue contention is given by the I/O service times, not necessarily the number and frequency of enqueues on this resource. . Seeing reported pmrs, in calalog heavy used, STRNO=3 may be too low and it is recommeded to set STRNO of 5 - 7. This is also indicated on apar II10752. Regards Christian Blesa DMR-Qualitas -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: DFSORT: from hiperspaces to dataspaces!!
On Thu, 1 Dec 2005 07:46:54 -0800, Frank Yaeger <[EMAIL PROTECTED]> wrote: >[EMAIL PROTECTED] wrote on 12/01/2005 01:14:36 AM: > >> Hello, >> >> I'm investigating little "ghost" in a sort job batch due to this job was >> working with hiperspaces and, without any changes, it has started to use >> dataspaces. >>... >> Any idea to continue working? > >If you want to know why hipersorting couldn't be used for this particular >run, I need to see the ICE898I message information. > >But what exactly is the concern here? Did the job using dataspace fail? >DFSORT can choose memory object sorting, data space sorting, or >hipersorting for the same job at different times depending on what's going >on in the system at the time. This doesn't mean that the job isn't >working, unless it actually failed. > >Frank Yaeger - DFSORT Team (IBM) > Specialties: ICETOOL, IFTHEN, OVERLAY, Symbols, Migration > => DFSORT/MVS is on the Web at http://www.ibm.com/storage/dfsort/ > Hello Frank, as I mentioned at mvshelp forum we'll sent you output of executions, one of them using hiperspc. and another using dataspaces. Thank you for your help Christian Blesa DMR-Qualitas -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: DFSORT: from hiperspaces to dataspaces!!
On Thu, 1 Dec 2005 06:07:59 -0500, John S. Giltner, Jr. <[EMAIL PROTECTED]> wrote: >DMR-Qualitas Outsourcing wrote: >> Hello, >> >> I'm investigating little "ghost" in a sort job batch due to this job was >> working with hiperspaces and, without any changes, it has started to use >> dataspaces. >> > >< Being Snip > >< End Snip > > >> >> Thanks in advance >> >> -- > >When you say no changes, do you really mean NO changes. That is: > > No maintence to z/OS > No maintance to DFSORT > >IIRC you can specific "default" DFSORT system parameters, these did not >change either? > Yes John, no maintenance to z/OS or DFSORT. DFSORT system parameters are different from default but it hasn't been changed between executions! Thx -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: STRNO in DB2 catalog
Thanks Terry, we have received similar answer from another forum: "I would have thought for a heavily hit DB2 there will be a heavy load on the Catalog at startup and at peak load times, depending on how many table spaces and index yor database comprises. I think the default is still 3, so I would be surprised to find a big kicker like a DB2 catalog to be so low, maybe up around the 7 or 8 mark. The downside of having multiple strings is that you use more storage (big deal), so yes, increasing this should be worthwhile. But as always with an exercise like this, always monitor the before and after impact on a known or simulated workload." Also we have opened a PMR to ask IBM about it. When we obtain a response we post it here. Regards Christian Blesa DMR-Qualitas -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: DFSORT: from hiperspaces to dataspaces!!
Another important difference: Before: ICE752I FSZ= 26843782528 BC IGN= 14280193 E AVG= 1000 0 WSP= 34865459 C DYN= 630120 56664 Now: ICE752I FSZ= 26663697126 BC IGN= 14119737 E AVG= 1000 0 WSP= 34631560 C DYN= 1389344 24544 Thanks Chr. On 12/1/05, DMR-Qualitas Outsourcing <[EMAIL PROTECTED]> wrote: > > Hello, > > I'm investigating little "ghost" in a sort job batch due to this job was > working with hiperspaces and, without any changes, it has started to use > dataspaces. > > This job executes DFSORT: > //P015 EXEC PGM=SORT,PARM='DYNALLOC=(,90),FILSZ=E0014119737' > > Statistics before: > RECORDS - IN: 14240593, OUT: 14240593 > NUMBER OF BYTES SORTED: 25405217912 > AVERAGE RECORD LENGTH = 1784 BYTES > TOTAL WORK DATA SET TRACKS ALLOCATED: 700875 , TRACKS USED: 522495 > HIPERSPACE STORAGE USED = 1048464K BYTES > DATA SPACE STORAGE USED = 0K BYTES > > Statistics now: > RECORDS - IN: 14346472, OUT: 14346472 > NUMBER OF BYTES SORTED: 25594106048 > AVERAGE RECORD LENGTH = 1784 BYTES > TOTAL WORK DATA SET TRACKS ALLOCATED: 628785 , TRACKS USED: 459660 > HIPERSPACE STORAGE USED = 0K BYTES > DATA SPACE STORAGE USED = 202216K BYTES > > > We have compared execution messages and we have seen following > differences: > >ICE750I >Before: > DC 26843782528 TC 0 CS DSVT0 KSZ 46 VSZ 46 >Now: > DC 26663697126 TC 0 CS DSVHH KSZ 46 VSZ 46 > > ICE886I > Before: > SYS 2956616 > TSTG 4674693 > FS 6553658 > INIT 36949 > MAX 50803 > LEN 50803 > Now: This message not appears > > ICE165I > Before: >TOTAL WORK DATA SET TRACKS ALLOCATED: 630120 >TRACKS > USED: 514800 > Now: >TOTAL WORK DATA SET TRACKS ALLOCATED: 686775 >TRACKS > USED: 514800 > > Following messages has appeared now: > > ICE880I 0 QP=64 QA=149 HI=465 LI=393 MI=541 TZ=262144 N1=262144 > N2=262144 SZ=65 HN=1ICE897I 0 > AVAL=1172189,BVAL=1880648,CVAL=1880648,ASV=1749576,HSZ=0,HM=M0,HN=2 ICE096I > 0 SUCCESSFUL RECOVERY FROM B37 ABEND(S) FOR WORK DATA SET(S) ICE881I 0 > EQ=I2 DX=0 D2=0 D3=63 D4=1 AS=0 SA=0 SB=0 SC=0 HN=1 > > Available storage is not changed and not in/decrease capacity sort pool!! > > Any idea to continue working? > > > Thanks in advance > -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
DFSORT: from hiperspaces to dataspaces!!
Hello, I'm investigating little "ghost" in a sort job batch due to this job was working with hiperspaces and, without any changes, it has started to use dataspaces. This job executes DFSORT: //P015 EXEC PGM=SORT,PARM='DYNALLOC=(,90),FILSZ=E0014119737' Statistics before: RECORDS - IN: 14240593, OUT: 14240593 NUMBER OF BYTES SORTED: 25405217912 AVERAGE RECORD LENGTH = 1784 BYTES TOTAL WORK DATA SET TRACKS ALLOCATED: 700875 , TRACKS USED: 522495 HIPERSPACE STORAGE USED = 1048464K BYTES DATA SPACE STORAGE USED = 0K BYTES Statistics now: RECORDS - IN: 14346472, OUT: 14346472 NUMBER OF BYTES SORTED: 25594106048 AVERAGE RECORD LENGTH = 1784 BYTES TOTAL WORK DATA SET TRACKS ALLOCATED: 628785 , TRACKS USED: 459660 HIPERSPACE STORAGE USED = 0K BYTES DATA SPACE STORAGE USED = 202216K BYTES We have compared execution messages and we have seen following differences: ICE750I Before: DC 26843782528 TC 0 CS DSVT0 KSZ 46 VSZ 46 Now: DC 26663697126 TC 0 CS DSVHH KSZ 46 VSZ 46 ICE886I Before: SYS 2956616 TSTG 4674693 FS 6553658 INIT 36949 MAX 50803 LEN 50803 Now: This message not appears ICE165I Before: TOTAL WORK DATA SET TRACKS ALLOCATED: 630120 TRACKS USED: 514800 Now: TOTAL WORK DATA SET TRACKS ALLOCATED: 686775 TRACKS USED: 514800 Following messages has appeared now: ICE880I 0 QP=64 QA=149 HI=465 LI=393 MI=541 TZ=262144 N1=262144 N2=262144 SZ=65 HN=1ICE897I 0 AVAL=1172189,BVAL=1880648,CVAL=1880648,ASV=1749576,HSZ=0,HM=M0,HN=2 ICE096I 0 SUCCESSFUL RECOVERY FROM B37 ABEND(S) FOR WORK DATA SET(S) ICE881I 0 EQ=I2 DX=0 D2=0 D3=63 D4=1 AS=0 SA=0 SB=0 SC=0 HN=1 Available storage is not changed and not in/decrease capacity sort pool!! Any idea to continue working? Thanks in advance -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
STRNO in DB2 catalog
Hi folks, I'm revising catalog parameters to improve performance and I have found that DB2 catalog has STRNO=2 (concurrent read request). This catalog was created in 1991 and perhaps this value could be increased to decrease SYSZRPLW enqueues. Thanks in advance -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
How to simulate Status Update Missing to test SFM?
Good morning, we are trying to test SFM policy in our test sysplex and we have done following to simulate a "Status Update Missing" of one system: 1) Start one system without CF and we would like to put offline then all paths but last of them has been restarted automatically: V 6208,OFFLINE,FORCE 0925 IEE800D CONFIRM VARY FORCE FOR 6208 - REPLY NO OR YES R 925,YES IEE600I REPLY TO 0925 IS;YES IXC308I STOP PATHIN REQUEST FOR DEVICE 6208 WAS 541 NOT SUCCESSFUL: AN UNCONDITIONAL STOP PREEMPTED THIS REQUEST DIAG073:08220003 000C 000C 0001 IEE793I 6208 PENDING OFFLINE AND BOXED IXC467I RESTARTING PATHIN DEVICE 6208 537 USED TO COMMUNICATE WITH SYSTEM SYSA RSN: I/O ERROR WHILE WORKING 2) Start one system with CF, then we have put offline all paths but when we have executed last stop of STR we have obtained following message: IXC302I SETXCF STOP PATHIN REQUEST FOR STRUCTURE IXC4 REJECTED: LAST PATH TO SYSTEM Anybody knows another type of test to verify SFM policy? Thanks -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
SMF61,65,67 & frequency
Hello, we would like to know the relevance of frequency to access an usercatalog. We have obtained a report of SMF 61, 65 and 67 and it shows values from 1 to 9000. Do you have experience about which is the threshold to swap a usercatalog? Thanks -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
BUFND & large GDGs
Good morning, we are working in catalog performance and we have read some post, zOS DFSMS-Managing Catalogs and Cheryl Watson Tuning Letter 2001 - N1. About following paragraph: "BUFND Specifies the number of buffers for transmitting data between virtual and auxiliary storage. The default, STRNO+1, is usually adequate. When a catalog contains large GDGs or other spanned records, the number should be increased accordingly to a minimum of MAXLRECL/CISIZE+STRNO." before to try and error, we would like to know experiences about usercatalogs with "large" GDGs and which will be, orientative, the amount of GDGs to consider 2nd suggested rule? Thanks -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Reply if GDPS scripts ends correctly
Good morning, we need to implement a console reply when a GDPS script ends. We believe that is only possible obtain a reply if script fails. May be we can put an ASSIST at the end of the script but also appears if this scripts fails and we program to continue. Do you know if there is possible to implement this type of reply? Thanks -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: Negative implications to performance due to scratch files pro cess
On Tue, 23 Aug 2005 11:51:48 -0400, Bruce Black <[EMAIL PROTECTED]> wrote: >> >> >>"... SYSVTOC is generally not a good candidate for reserve conversion >>because reserves are of short duration and I/O intensive >>" >> >That is true if you are using GRS Ring configuration (CTCs) but not if >you are using GRS Star (CFs in a parallel sysplex) > >-- Yes, it's correct. We are using mode Star through CFs in a Parallel Sysplex (9 systems). Tks! -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: Negative implications to performance due to scratch files pro cess
Hello, yes but we can't have any reserve due to Hyperswap (GDPS feature) and therefore all reserves must be converted to ENQ. Tks for your help -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: Negative implications to performance due to scratch files process
Hello Bruce, this ASM module uses following macros: NAME (to locate), UCATDX (to descatlg) and SCRATCH (to delete). This 'long process' (12hour) is a batch that executes thousands of file deletes and it loads this module and it passes differents parameters. This module is prepared to execute different function with every paramater passed (delete scratch, hdelete, delete noscratch). We are converting SYSVTOC to global ENQ and we have an active VTOCIX on the disk volume. Perhaps the 'bug' is at architecture of this module. We will revise it. Thanks -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: Region requested by MQBroker
We have found following explanation: http://www-306.ibm.com/software/data/db2/db2olap/v81docs/zosinst/olapz81tfrm.htm?olapz8115.htm Programming the IEFUSI user exit Memory should not be limited by the IEFUSI user exit. IBM recommends that you bypass normal IEFUSI processing when the subsystem is OMVS. DB2 OLAP Server spawns processes for each OLAP application. According to the MVS Installation Exits manual, at the time of the IEFUSI processing, the kernel has not yet propagated the parent's region size to the child. Therefore, IEFUSI has nothing to work with. If IEFUSI modifies the region size of the child process, the kernel will honor that region size, and it will not propagate the region size from parent to child. This can result in a failure of a fork if the region size is insufficient for the storage requirements of the forked process. If you must run with the IEFUSI exit, be sure that it correctly handles the region size available to the OMVS address space. The IEFUSI exit should be coded to check for OMVS and should not set the x'80' bit in REGFLAGS for OMVS regions. The REGFLAGS x'80' bit means that you use the values that IEFUSI has set into the region size values. This can cause an inappropriate and unintended region size limit for OMVS; it is often 54 MB, but the value can vary by system. If your IEFUSI exit has not been updated in several years, it is likely that the OMVS address space runs with a 54 MB region size, even though it looks like it is not limiting the OMVS region size. When IBM added UNIX (Open Edition, then UNIX System Services), they made OMVS a special case that behaves differently from other region types and documented it as being different. Some customers did not amend their IEFUSI exits that predated the introduction of the OMVS region type and did not implement the change that IBM suggested. The suggestion became a requirement when maintenance caused IEFUSI behavior to limit the OMVS region size to 54 MB. OLAP is likely to require more memory than 54 MB and will fail to execute correctly. New variables were added to the REGION DSECT in recent years, including REGNAME, which can be used to detect the OMVS region. Here is some sample IEFUSI logic: L BASENAME,REGNAMEcheck for region type parm CLC 0(4,BASENAMEN),=CL4'OMVS' is it OMVS region? BE Return OI REGFLAGS,X'80' SET THAT IEFUSI CONTROLS NON-OMVS REGION You can find more information about the IEFUSI exit in MVS Installation Exits and DOC APAR OW38477. On 7/21/05, DMR-Qualitas Outsourcing <[EMAIL PROTECTED]> wrote: > Hi folks, > > we are analyzing an incident about MQBroker and we have seen that > region requested is 54M but we don't find where region is codified. > > Could you help us to find it? > > >RPT VERSION V1R2 RMF TIME 16.33.28 > >PRIVATE AREA SUMMARY > JOB NAME - Q0T1BRK3REGION REQUESTED 54.0M > STEP NAME - STEP1 REGION ASSIGNED (BELOW 16M) 6400K > PROGRAM NAME - BPXPRFC REGION ASSIGNED (ABOVE 16M) 500M > NUMBER OF SAMPLES -17 > > > PRIVATE STORAGE MAP > BELOW 16M EXTENDED (ABOVE 16M) > 6F > 7FFF > | LSQA/SWA | | LSQA/SWA | > | 229/230 276K|BOTTOM OF| 229/230 114M| > 6BB000 |_16.33.29_| ALLOCATED AREA > |_16.33.29_| 78E0 > | UNUSED 212K| | UNUSED 304M| > 686000 |__| GETMAIN LIMIT > |__| 65E0 > | UNUSED 6568K| | UNUSED 77.5M| > 1C000 |_16.33.29_| TOP OF > |_16.33.29_| 6107D000 > | USER | ALLOCATED AREA | USER | > | REGION88K| | REGION | > 6000 |__| | 423M| > | SYSTEM REGION 16K| | | > 2000 > 46A0 > > Thank you > -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Region requested by MQBroker
Hi folks, we are analyzing an incident about MQBroker and we have seen that region requested is 54M but we don't find where region is codified. Could you help us to find it? RPT VERSION V1R2 RMF TIME 16.33.28 PRIVATE AREA SUMMARY JOB NAME - Q0T1BRK3REGION REQUESTED 54.0M STEP NAME - STEP1 REGION ASSIGNED (BELOW 16M) 6400K PROGRAM NAME - BPXPRFC REGION ASSIGNED (ABOVE 16M) 500M NUMBER OF SAMPLES -17 PRIVATE STORAGE MAP BELOW 16M EXTENDED (ABOVE 16M) 6F 7FFF | LSQA/SWA | | LSQA/SWA | | 229/230 276K|BOTTOM OF| 229/230 114M| 6BB000 |_16.33.29_| ALLOCATED AREA |_16.33.29_| 78E0 | UNUSED 212K| | UNUSED 304M| 686000 |__| GETMAIN LIMIT |__| 65E0 | UNUSED 6568K| | UNUSED 77.5M| 1C000 |_16.33.29_| TOP OF |_16.33.29_| 6107D000 | USER | ALLOCATED AREA | USER | | REGION88K| | REGION | 6000 |__| | 423M| | SYSTEM REGION 16K| | | 2000 46A0 Thank you -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: MCAT cache hits
Hello, we don't found any answer about 'I have found comments in my current customer's COFVLFxx member stating "never place master catalogs under VLF" by Kelvin Wells. Tks. -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: VLF Trimming - how much is good?
Hello Fred, we are also analyzing the reports of VLF and we have seen high trimming for 2 hours/day. At COFVLF we have MAXVIRT(16384) -> 64Mb!! Which is your MAXVIRT? Regards Christian Blesa Operating Systems -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
How to add, dynamically, a catalog to VLF?
Hi folks, we need to know how to add, dynamically, a catalog to VLF. Actually, this catalog resides at ISC. The sequence executed is: 1) Remove this catalog from ISC with 'F CATALOG,ALLOCATE(catname),NOISC' 2) Add this catalog to VLF with 'F CATALOG,VLF(catname) Then, this catalog is closed. We force to open this catalog (browse to any file of it) and it's open but at report of 'F CATALOG,ALLOCATED' shows this catalog with '-' at third position and therefore it isn't at ISC or VLF!!! FLAGS -VOLSER-USER-CATALOG NAME YS--R- S00013 0001 CATALOG.XXIN Y/N-ALLOCATED TO CAS S-SMS V-VLF I-ISC C-CLOSED D-DELETED, R-SHARED A-ATL E-ECS SHARED K-LOCKED Could you help us? Tks. -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
