IRLM running out of memory

Michael Ebert

IRLM running out of memory
Hello all,

as part of our Y2K preparation procedures, I had planned to take fresh
imagecopies of all our TS, and then do a QUIESCE just before midnight to
establish a recovery point - I am not concerned about DB2 failing, but this
might come in handy if an application starts chewing up data after midnight. I
used a SAS program to generate the QUIESCE job, partitioning our subsystems into
parts containing about 700-1100 Tablespaces each (QUIESCE accepts a maximum of
1165 TSs), or 1400-3500 partitions. However, I got an abend because IRLM ran out
of lock memory:

DSNT500I DSNUGBAC - RESOURCE UNAVAILABLE
REASON 00C90092
TYPE 00000905
NAME IRLM
DSNU017I DSNUGBAC - UTILITY DATA BASE SERVICES MEMORY EXECUTION ABENDED,
REASON=X'00E40006'

This is the message in the IRLMproc JESMSGLG:

STC24033 *DXR175E IRLT001 IRLM IS UNABLE TO OBTAIN STORAGE - MCSA.
STC24033 DXR176I IRLT001 IRLM STORAGE SHORTAGE RELIEVED.
STC24033 *DXR175E IRLT001 IRLM IS UNABLE TO OBTAIN STORAGE - MCSA.
STC24033 DXR176I IRLT001 IRLM STORAGE SHORTAGE RELIEVED.
STC24033 DXR101I IRLT001 STATUS SCOPE=LOCAL
SUBSYSTEMS IDENTIFIED
NAME STATUS UNITS HELD WAITING RET_LKS
DB2T UP 1 3 0 0
STC24033 DXR100I IRLT001 STOR STATS
PC: NO MAXCSA: 2M
CSA USE: ACNT: 196K AHWM: 2052K CUR: 660K HWM: 2452K
ABOVE 16M: 15 655K BELOW 16M: 3 5K
CLASS TYPE SEGS MEM TYPE SEGS MEM TYPE SEGS MEM
ACCNT T-1 2 128K T-2 1 64K T-3 1 4K
PROC WRK 3 18K SRB 1 1K OTH 1 1K
MISC VAR 6 409K N-V 6 22K FIX 1 24K
STC24033 *DXR175E IRLT001 IRLM IS UNABLE TO OBTAIN STORAGE - MCSA.

The MAXCSA parameter for our IRLMprocs is very low - 2 Megabytes. Because we are
V5, we cannot change this value dynamically (correct?). In the DB2 Admin guide
it says that each lock requires about 250 bytes. So what I want to know is, how
many locks does QUIESCE need per TS? Is this dependent on # of partitions,
LOCKPART YES/NO, and/or LOCKSIZE? From this I will have to determine whether we
should do an ALTER LOCKSIZE prior to the QUIESCE, and/or how small I have to
slice our subsystems to enable the QUIESCE to run.

Dr. Michael Ebert
DB2 Database Administrator
aMaDEUS Data Processing
Erding / Munich, Germany



[login to unmask email]

Re: IRLM running out of memory
(in response to Michael Ebert)
Hi,

If you are not looking for advice to raise your MAXCSA to 7M
(PC=NO) then you could try stopping the database(s), start for
UT only and quiesce a single tablespace which would be the
recovery point for all of the tablespaces in that database.
This could be repeated for all of the databases that you want
to create consistent recovery points for.

Regards,

Michael Levine
Premier Data Services, Inc.


>Hello all,
>
>as part of our Y2K preparation procedures, I had planned to take fresh
>imagecopies of all our TS, and then do a QUIESCE just before midnight to
>establish a recovery point - I am not concerned about DB2 failing, but this
>might come in handy if an application starts chewing up data after midnight. I
>used a SAS program to generate the QUIESCE job, partitioning our subsystems
into
>parts containing about 700-1100 Tablespaces each (QUIESCE accepts a maximum of
>1165 TSs), or 1400-3500 partitions. However, I got an abend because IRLM
ran out
>of lock memory:
>
>DSNT500I DSNUGBAC - RESOURCE UNAVAILABLE
> REASON 00C90092
> TYPE 00000905
> NAME IRLM
>DSNU017I DSNUGBAC - UTILITY DATA BASE SERVICES MEMORY EXECUTION ABENDED,
>REASON=X'00E40006'
>
>This is the message in the IRLMproc JESMSGLG:
>
>STC24033 *DXR175E IRLT001 IRLM IS UNABLE TO OBTAIN STORAGE - MCSA.
>STC24033 DXR176I IRLT001 IRLM STORAGE SHORTAGE RELIEVED.
>STC24033 *DXR175E IRLT001 IRLM IS UNABLE TO OBTAIN STORAGE - MCSA.
>STC24033 DXR176I IRLT001 IRLM STORAGE SHORTAGE RELIEVED.
>STC24033 DXR101I IRLT001 STATUS SCOPE=LOCAL
> SUBSYSTEMS IDENTIFIED
> NAME STATUS UNITS HELD WAITING RET_LKS
> DB2T UP 1 3 0 0
>STC24033 DXR100I IRLT001 STOR STATS
> PC: NO MAXCSA: 2M
> CSA USE: ACNT: 196K AHWM: 2052K CUR: 660K HWM: 2452K
> ABOVE 16M: 15 655K BELOW 16M: 3 5K
> CLASS TYPE SEGS MEM TYPE SEGS MEM TYPE SEGS MEM
> ACCNT T-1 2 128K T-2 1 64K T-3 1 4K
> PROC WRK 3 18K SRB 1 1K OTH 1 1K
> MISC VAR 6 409K N-V 6 22K FIX 1 24K
>STC24033 *DXR175E IRLT001 IRLM IS UNABLE TO OBTAIN STORAGE - MCSA.
>
>The MAXCSA parameter for our IRLMprocs is very low - 2 Megabytes. Because
we are
>V5, we cannot change this value dynamically (correct?). In the DB2 Admin guide
>it says that each lock requires about 250 bytes. So what I want to know is, how
>many locks does QUIESCE need per TS? Is this dependent on # of partitions,
>LOCKPART YES/NO, and/or LOCKSIZE? From this I will have to determine whether we
>should do an ALTER LOCKSIZE prior to the QUIESCE, and/or how small I have to
>slice our subsystems to enable the QUIESCE to run.
>
>Dr. Michael Ebert
>DB2 Database Administrator
>aMaDEUS Data Processing
>Erding / Munich, Germany
>
>
>


>
>



HARBRY ARIZA

Re: IRLM running out of memory
(in response to Mike_Levine@TEKHELP.NET)
Michael:

I would like suggest to you use LOCK TABLE statements (like this
'LOCK TABLE TCREATOR.TNAME IN SHARE or EXCLUSIVE MODE') and I bet you won't
have problem at all. Try it out .

I hope it helps.

Harbry.


>From: [login to unmask email]
>Reply-To: DB2 Data Base Discussion List <[login to unmask email]>
>To: [login to unmask email]
>Subject: IRLM running out of memory
>Date: Wed, 29 Dec 1999 18:18:46 +0100
>
>Hello all,
>
>as part of our Y2K preparation procedures, I had planned to take fresh
>imagecopies of all our TS, and then do a QUIESCE just before midnight to
>establish a recovery point - I am not concerned about DB2 failing, but this
>might come in handy if an application starts chewing up data after
>midnight. I
>used a SAS program to generate the QUIESCE job, partitioning our subsystems
>into
>parts containing about 700-1100 Tablespaces each (QUIESCE accepts a maximum
>of
>1165 TSs), or 1400-3500 partitions. However, I got an abend because IRLM
>ran out
>of lock memory:
>
>DSNT500I DSNUGBAC - RESOURCE UNAVAILABLE
> REASON 00C90092
> TYPE 00000905
> NAME IRLM
>DSNU017I DSNUGBAC - UTILITY DATA BASE SERVICES MEMORY EXECUTION ABENDED,
>REASON=X'00E40006'
>
>This is the message in the IRLMproc JESMSGLG:
>
>STC24033 *DXR175E IRLT001 IRLM IS UNABLE TO OBTAIN STORAGE - MCSA.
>STC24033 DXR176I IRLT001 IRLM STORAGE SHORTAGE RELIEVED.
>STC24033 *DXR175E IRLT001 IRLM IS UNABLE TO OBTAIN STORAGE - MCSA.
>STC24033 DXR176I IRLT001 IRLM STORAGE SHORTAGE RELIEVED.
>STC24033 DXR101I IRLT001 STATUS SCOPE=LOCAL
> SUBSYSTEMS IDENTIFIED
> NAME STATUS UNITS HELD WAITING
>RET_LKS
> DB2T UP 1 3 0
> 0
>STC24033 DXR100I IRLT001 STOR STATS
> PC: NO MAXCSA: 2M
> CSA USE: ACNT: 196K AHWM: 2052K CUR: 660K HWM: 2452K
> ABOVE 16M: 15 655K BELOW 16M: 3 5K
> CLASS TYPE SEGS MEM TYPE SEGS MEM TYPE SEGS
>MEM
> ACCNT T-1 2 128K T-2 1 64K T-3 1
> 4K
> PROC WRK 3 18K SRB 1 1K OTH 1
> 1K
> MISC VAR 6 409K N-V 6 22K FIX 1
>24K
>STC24033 *DXR175E IRLT001 IRLM IS UNABLE TO OBTAIN STORAGE - MCSA.
>
>The MAXCSA parameter for our IRLMprocs is very low - 2 Megabytes. Because
>we are
>V5, we cannot change this value dynamically (correct?). In the DB2 Admin
>guide
>it says that each lock requires about 250 bytes. So what I want to know is,
>how
>many locks does QUIESCE need per TS? Is this dependent on # of partitions,
>LOCKPART YES/NO, and/or LOCKSIZE? From this I will have to determine
>whether we
>should do an ALTER LOCKSIZE prior to the QUIESCE, and/or how small I have
>to
>slice our subsystems to enable the QUIESCE to run.
>
>Dr. Michael Ebert
>DB2 Database Administrator
>aMaDEUS Data Processing
>Erding / Munich, Germany
>
>
>
>
>

______________________________________________________
Get Your Private, Free Email at http://www.hotmail.com



[login to unmask email]

Re: IRLM running out of memory
(in response to HARBRY ARIZA)
HI

The online V5 books include

>>__MODIFY__irlmproc,SET___,CSA=nnn_________________________________>< |


http://www.s390.ibm.com/bookmgr-cgi/bookmgr.cmd/BOOKS/DSNCR0F5/2%2e42?SHELF=DSNSH0F5




I haven't tried it and Ii don't know if you need PTF's ???


Tom







[login to unmask email] on 12/29/99 12:18:46 PM

Please respond to DB2 Data Base Discussion List <[login to unmask email]>








To: [login to unmask email]

cc: (bcc: Thomas G Taylor/ChubbMail)



Subject: IRLM running out of memory








Hello all,

as part of our Y2K preparation procedures, I had planned to take fresh
imagecopies of all our TS, and then do a QUIESCE just before midnight to
establish a recovery point - I am not concerned about DB2 failing, but this
might come in handy if an application starts chewing up data after midnight. I
used a SAS program to generate the QUIESCE job, partitioning our subsystems into
parts containing about 700-1100 Tablespaces each (QUIESCE accepts a maximum of
1165 TSs), or 1400-3500 partitions. However, I got an abend because IRLM ran out
of lock memory:

DSNT500I DSNUGBAC - RESOURCE UNAVAILABLE
REASON 00C90092
TYPE 00000905
NAME IRLM
DSNU017I DSNUGBAC - UTILITY DATA BASE SERVICES MEMORY EXECUTION ABENDED,
REASON=X'00E40006'

This is the message in the IRLMproc JESMSGLG:

STC24033 *DXR175E IRLT001 IRLM IS UNABLE TO OBTAIN STORAGE - MCSA.
STC24033 DXR176I IRLT001 IRLM STORAGE SHORTAGE RELIEVED.
STC24033 *DXR175E IRLT001 IRLM IS UNABLE TO OBTAIN STORAGE - MCSA.
STC24033 DXR176I IRLT001 IRLM STORAGE SHORTAGE RELIEVED.
STC24033 DXR101I IRLT001 STATUS SCOPE=LOCAL
SUBSYSTEMS IDENTIFIED
NAME STATUS UNITS HELD WAITING RET_LKS
DB2T UP 1 3 0 0
STC24033 DXR100I IRLT001 STOR STATS
PC: NO MAXCSA: 2M
CSA USE: ACNT: 196K AHWM: 2052K CUR: 660K HWM: 2452K
ABOVE 16M: 15 655K BELOW 16M: 3 5K
CLASS TYPE SEGS MEM TYPE SEGS MEM TYPE SEGS MEM
ACCNT T-1 2 128K T-2 1 64K T-3 1 4K
PROC WRK 3 18K SRB 1 1K OTH 1 1K
MISC VAR 6 409K N-V 6 22K FIX 1 24K
STC24033 *DXR175E IRLT001 IRLM IS UNABLE TO OBTAIN STORAGE - MCSA.

The MAXCSA parameter for our IRLMprocs is very low - 2 Megabytes. Because we are
V5, we cannot change this value dynamically (correct?). In the DB2 Admin guide
it says that each lock requires about 250 bytes. So what I want to know is, how
many locks does QUIESCE need per TS? Is this dependent on # of partitions,
LOCKPART YES/NO, and/or LOCKSIZE? From this I will have to determine whether we
should do an ALTER LOCKSIZE prior to the QUIESCE, and/or how small I have to
slice our subsystems to enable the QUIESCE to run.

Dr. Michael Ebert
DB2 Database Administrator
aMaDEUS Data Processing
Erding / Munich, Germany








[login to unmask email]

Re: IRLM running out of memory
(in response to ttaylor@CHUBB.COM)
The PTF enabling this is UQ17535, and was available at the 9806 level.
PQ12126 has the full description of what's included.

> -----Original Message-----
> From: [login to unmask email] [SMTP:[login to unmask email]
> Sent: Wednesday, December 29, 1999 12:45 PM
> To: [login to unmask email]
> Subject: Re: IRLM running out of memory
>
> HI
>
> The online V5 books include
>
> >>__MODIFY__irlmproc,SET___,CSA=nnn_________________________________>< |
>
>
> http://www.s390.ibm.com/bookmgr-cgi/bookmgr.cmd/BOOKS/DSNCR0F5/2%2e42?SHEL
> F=DSNSH0F5
>
>
>
>
> I haven't tried it and Ii don't know if you need PTF's ???
>
>
> Tom
>
>
>
>
>
>
>
> [login to unmask email] on 12/29/99 12:18:46 PM
>
> Please respond to DB2 Data Base Discussion List <[login to unmask email]>
>
>
>
>
>
>
>
>
> To: [login to unmask email]
>
> cc: (bcc: Thomas G Taylor/ChubbMail)
>
>
>
> Subject: IRLM running out of memory
>
>
>
>
>
>
>
>
> Hello all,
>
> as part of our Y2K preparation procedures, I had planned to take fresh
> imagecopies of all our TS, and then do a QUIESCE just before midnight to
> establish a recovery point - I am not concerned about DB2 failing, but
> this
> might come in handy if an application starts chewing up data after
> midnight. I
> used a SAS program to generate the QUIESCE job, partitioning our
> subsystems into
> parts containing about 700-1100 Tablespaces each (QUIESCE accepts a
> maximum of
> 1165 TSs), or 1400-3500 partitions. However, I got an abend because IRLM
> ran out
> of lock memory:
>
> DSNT500I DSNUGBAC - RESOURCE UNAVAILABLE
> REASON 00C90092
> TYPE 00000905
> NAME IRLM
> DSNU017I DSNUGBAC - UTILITY DATA BASE SERVICES MEMORY EXECUTION ABENDED,
> REASON=X'00E40006'
>
> This is the message in the IRLMproc JESMSGLG:
>
> STC24033 *DXR175E IRLT001 IRLM IS UNABLE TO OBTAIN STORAGE - MCSA.
> STC24033 DXR176I IRLT001 IRLM STORAGE SHORTAGE RELIEVED.
> STC24033 *DXR175E IRLT001 IRLM IS UNABLE TO OBTAIN STORAGE - MCSA.
> STC24033 DXR176I IRLT001 IRLM STORAGE SHORTAGE RELIEVED.
> STC24033 DXR101I IRLT001 STATUS SCOPE=LOCAL
> SUBSYSTEMS IDENTIFIED
> NAME STATUS UNITS HELD WAITING
> RET_LKS
> DB2T UP 1 3 0
> 0
> STC24033 DXR100I IRLT001 STOR STATS
> PC: NO MAXCSA: 2M
> CSA USE: ACNT: 196K AHWM: 2052K CUR: 660K HWM: 2452K
> ABOVE 16M: 15 655K BELOW 16M: 3 5K
> CLASS TYPE SEGS MEM TYPE SEGS MEM TYPE SEGS
> MEM
> ACCNT T-1 2 128K T-2 1 64K T-3 1
> 4K
> PROC WRK 3 18K SRB 1 1K OTH 1
> 1K
> MISC VAR 6 409K N-V 6 22K FIX 1
> 24K
> STC24033 *DXR175E IRLT001 IRLM IS UNABLE TO OBTAIN STORAGE - MCSA.
>
> The MAXCSA parameter for our IRLMprocs is very low - 2 Megabytes. Because
> we are
> V5, we cannot change this value dynamically (correct?). In the DB2 Admin
> guide
> it says that each lock requires about 250 bytes. So what I want to know
> is, how
> many locks does QUIESCE need per TS? Is this dependent on # of partitions,
> LOCKPART YES/NO, and/or LOCKSIZE? From this I will have to determine
> whether we
> should do an ALTER LOCKSIZE prior to the QUIESCE, and/or how small I have
> to
> slice our subsystems to enable the QUIESCE to run.
>
> Dr. Michael Ebert
> DB2 Database Administrator
> aMaDEUS Data Processing
> Erding / Munich, Germany
>
>
>
> the
>
>
>
>
>
>
>

Roger Miller

Re: IRLM running out of memory
(in response to aylswop@WELLSFARGO.COM)
The IRLM release is the same for DB2 V4, V5 and V6. It's still IRLM V2R1.
The changes which were provided after DB2 V5 general availability are
documented in the Version 6 Release Guide, chapter 6 User Productivity,
section 4 IRLM Enhancements. If you are V4, V5 or V6 DB2, you can put
on the service and use these changes.

This section is about four pages, noting the APARs needed to deliver
the function. Before you increase the MAXCSA substantially, please be sure
the the ECSA specification on the OS/390 image is adequate. We have
seen some customers switch from PC=NO to PC=YES to avoid some
issues with limited storage sizes.

Roger Miller



Robert A. (Cameron of Pittsburgh at Alcoa) Knight

Re: IRLM running out of memory
(in response to Roger Miller)
I had this problem going from v3 to v4 on a test LPAR. I changed
the parameter to PC=YES and was able to run. I was not to
concerned because it was the test lpar and not the production
lpar. Then upon migration to v5 the ECSA was expanded on that
lpar and I changed back to PC=NO to keep consistent.

> ----------
> From: Roger Miller[SMTP:[login to unmask email]
> Reply To: DB2 Data Base Discussion List
> Sent: Wednesday, December 29, 1999 5:38 PM
> To: [login to unmask email]
> Subject: Re: IRLM running out of memory
>
> The IRLM release is the same for DB2 V4, V5 and V6. It's still IRLM V2R1.
> The changes which were provided after DB2 V5 general availability are
> documented in the Version 6 Release Guide, chapter 6 User Productivity,
> section 4 IRLM Enhancements. If you are V4, V5 or V6 DB2, you can put
> on the service and use these changes.
>
> This section is about four pages, noting the APARs needed to deliver
> the function. Before you increase the MAXCSA substantially, please be
> sure
> the the ECSA specification on the OS/390 image is adequate. We have
> seen some customers switch from PC=NO to PC=YES to avoid some
> issues with limited storage sizes.
>
> Roger Miller
>
>
>
>
>
>