dasdseq unable to get huge data set (DSNTYPE=LARGE or DSNTYPE=EXTENDED)

[GregoryTwin]

dasdseq determines the end of a dataset based on DS1LSTAR. However, for DSNTYPE=LARGE, DS1LSTAR contains just the 2 low order bytes of the last block pointer. (The high order byte is kept on DS1TTTHI, offset 104 = x'68'.) And for DSNTYPE=EXTENDED, the first 2 bytes of the last block pointer is kept in DS1TRBAL.

dasdseq takes neither of these into consideration, and as a result, attempts to offload such data sets with dasdseq completes with RC=0, but results in an output file that is incomplete (truncated), i.e. does not contain all of the data set's data.

[GregoryTwin]

prooflink https://www.ibm.com/docs/en/zos/2.3.0?topic=dscbs-how-found

[Fish-Git]

I cannot understand your problem report. Can you explain your issue a bit more clearly please? Do you have a repro? Can you provide detailed step-by-step instructions for how someone could easily reproduce the problem? Can you at least provide us with your Hercules and guest (MVS? z/OS?) logs that shows the problem occurring?

If you want our help, you need to help us!

See: "SUBMITTING PROBLEM REPORTS". Steps 4 and 8 are the ones I'm mostly interested in right now, but all of the other information would be helpful as well.

Remember: it is never harmful to provide more information than needed, but is almost always harmful to not provide enough information.

[GregoryTwin]

Thank you very much for your answer! I hope, my description of the problem was pretty clear, but I see I'm wrong, sorry for that. Here is step-by-step instruction:

1. Use z/OS as Hercules guest.

2. Allocate data set with DSNTYPE=LARGE and fill the data set with any data, more that 65535 tracks should be used.

Either of the following jobs may be used for that:

//IEBDG    EXEC PGM=IEBDG
//SYSPRINT DD  SYSOUT=*
//LARGE    DD  DSNAME=LARGE.DATASET,UNIT=3390,DISP=(,KEEP),
//             DCB=(RECFM=FB,LRECL=80,BLKSIZE=8000),
//             SPACE=(CYL,(10000,10000)),VOLUME=SER=111111,
//             DSNTYPE=LARGE
//SYSIN    DD  *
  DSD OUTPUT=(LARGE)
  REPEAT QUANTITY=65000,CREATE=1
  CREATE QUANTITY=10000,FILL=X'FF'
  END
//

Alternatively, you could dump a large volume using ADRDSSU:

//ADRDSSU  EXEC PGM=ADRDSSU,REGION=0M
//SYSPRINT DD  SYSOUT=*
//Z24U01   DD  DISP=OLD,VOL=SER=Z24U01,UNIT=3390
//LARGE    DD  DSN=LARGE.DATASET,DISP=(,CATLG),
//             SPACE=(CYL,(10000,10000)),DSNTYPE=LARGE,
//             UNIT=3390,VOL=SER=111111
//SYSIN    DD  *
  DUMP DATASET( -
       INCLUDE(**)) -
          INDD(SYSRES1) -
          OUTDD(LARGE) -
        SHARE TOL(ENQF) ALLEXCP
 //

Before proceeding to the next step, please first verify that the data set that was created really is "LARGE" (i.e. occupies more that 65535 tracks).

3. Then use dasdseq on Windows to extract (unload) that just created data set to an output file. `dasdseq' should successfully complete without errors.

4. Finally, compare the size of the original data set LARGE.DATASET with the file that was created by dasdseq, and you'll see that the dasdseq output file has been truncated.

In my first post I tried to explain why this truncation happens.

Thank you in advance [and have Merry Christmas and Happy New Year!]

[Fish-Git]

(@GregoryTwin)

Thank you, Gregory! That helps a lot.

But unfortunately not enough.  :(

In step 2, you say:

Please verify that the data set really is "LARGE" (i.e. occupies more that 65535 tracks).

How do I do that? I know very little about z/OS! I can use z/OS a little bit, but I am extremely inexperienced with it! Should I do a ISPF function 3.4? (to get a list of datasets on that volume?) Then what? How to I display the size of the dataset? Is there a prefix command I can use? ("I" for "Information" or something??) You have to understand, Gregory, that I am a Hercules developer, not a z/OS developer!  :)

Also, on the second example JOB, I see that your ADRDSSU INDD is specified as "SYSRES1", but yet your DD statement (and VOL=SER) is labeled "Z24U01". May I presume that is a typo?

Finally, can you tell me what your dasdseq command line looked like? What parameters/option did you specify on your dasdseq command?

Thanks!

[GregoryTwin]

Thank you for a quick answer!

Should I do a ISPF function 3.4? (to get a list of datasets on that volume?) Then what? How to I display the size of the dataset? Is there a prefix command I can use? ("I" for "Information" or something??)

Absolutely correct. Just use ISPF 3.4, then "I" prefix command against interested data set. You'll see there both "allocated tracks" as well as "used tracks". "used tracks" shows current space utilization and should be > 65535.

Also, on the second example JOB, I see that your ADRDSSU INDD is specified as "SYSRES1", but yet your DD statement (and VOL=SER) is labeled "Z24U01". May I presume that is a typo?

Sure. However, you should use existing volume, (presumably system residence volume), which contains lot of data.

Finally, can you tell me what your dasdseq command line looked like? What parameters/option did you specify on your dasdseq command?

dasdseq  <image_file>  <dataset>

I also tried -expert mode with the same result.

Thanks!

[Fish-Git]

Absolutely correct. Just use ISPF 3.4, then "I" prefix command against interested data set.

(Re: typo):

... you should use existing volume, (presumably system residence volume), which contains lot of data.

dasdseq <image_file> <dataset>

Thanks! I'll start looking into this bug right away.

[Fish-Git]

... you should use existing volume, (presumably system residence volume), which contains lot of data.

Can you give me the name of a dataset on a z/OS system residence volume that I should use? (so I don't have to look for one?) Thanks.

[GregoryTwin]

Can you give me the name of a dataset on a z/OS system residence volume that I should use? (so I don't have to look for one?) Thanks.

My 2nd (alternative) job above specifies INCLUDE(**) as one of its ADRDSSU statements, and thus dumps all data sets that are on the volume to a single sequential output file, so you don't need to specify any specific data set name. Thus, IMHO, the 1st job would be better.

I've checked it and discovered that I'm wrong with my space calculations, and so it generated many records which I didn't required. The updated (corrected) version is below:

//IEBDG    EXEC PGM=IEBDG
//SYSPRINT DD  SYSOUT=*
//LARGE    DD  DSNAME=IBMUSER.LARGE,UNIT=3390,DISP=(,CATLG),
//             DCB=(RECFM=FB,LRECL=80,BLKSIZE=8000),
//             SPACE=(TRK,(150000,150000)),DATACLAS=BASE,
//             DSNTYPE=LARGE
//SYSIN    DD  *
  DSD OUTPUT=(LARGE)
  REPEAT QUANTITY=65000,CREATE=1
  CREATE QUANTITY=1000,FILL=X'FF'
  END
//

This job writes 65000000 records (65000 * 1000), which consumes 108334 tracks, which is enough. On my home system (Intel i5 760, 2.80Ghz, windows 7, Hyperion 4.6.0 SDL, z/OS 2.4) this job ran 7 minutes.

If you have non-SMS volumes available, you can specify VOL=SER on LARGE DD. On my home z/OS, there are no non-SMS volumes (all volumes are SMS-managed), so I've replaced VOL=SER with DATACLAS=BASE instead.

P.S. Size of this data sets in bytes is 80 x 1000 x 6500 = 5 200 000 000.

P.P.S. After running dasdseq, I have a file called IBMUSER.LARGE which is 2 054 272 000 bytes (2GB) rather than 5 200 000 000 bytes (5GB).

[Fish-Git]

Gregory (or anyone else!):   another Quick Question:

How can you tell whether a dataset is normal, large or extended? I'm guessing there is a flag somewhere in the DSCB?

(Remember, I know nothing about z/OS or MVS!)

[GregoryTwin]

Exactly, it can be recognized by flag in DSCB F1:

61(X'3D') DS1FLAG1 Flag byte
.... 1... DS1LARGE Large format data set.

78(X'4E') DS1SMSFG System managed storage indicators.
.... .1.. DS1STRP Sequential extended-format data set.

See https://www.ibm.com/docs/en/zos/2.3.0?topic=dscbs-how-found for other details

"Large" data set may exists on both SMS-managed and non SMS-manged volumes, but "extended" data set may exists on SMS-managed volume only.

[GregoryTwin]

I suppose all what we need to know to correctly process all kind of sequential data set are the following DSCB F1 fields:

61(X'3D') DS1FLAG1(1) Flag byte. 
.... 1... (X'08') DS1LARGE Large format data set.


78(X'4E') DS1SMSFG(1) System managed storage indicators.
.... .1.. (X'04') DS1STRP Sequential extended-format data set.


82(X'52') DS1DSORG(2) Data set organization.
0100  000. (X'FE'=X'40') DS1DSGPS Physical sequential (PS)
organization. (I guess, 'dasdseq' already check this field)


98(X'62') DS1LSTAR(3) Last used track and block on track (TTR).


101(X'65') DS1TRBAL(2) Space remaining on last track.
If NOT extended format, this is the value from TRKCALC indicating
space remaining on last track used. For extended format data sets
this is the high order two bytes of the four-byte TTTT last used
track number. (see DS1LSTAR.) Zero for VSAM, PDSE, and HFS.


104(X'68') DS1TTTHI(1) High order byte of track number in DS1LSTAR.
The first (high-order) byte of the TTT last block pointer for "large"
data set. Valid only if DS1LARGE is on.

[Fish-Git]

THANK YOU, Gregory! Much appreciated.

Please check me on this. I want to be sure I'm understanding things correctly.

Are the following comments correct/accurate?

/*                     PROGRAMMING NOTE

    For NORMAL datasets, the last block pointer (TTR) is in the
    DS1LSTAR field. So the size of the dataset in number of tracks
    is in the first two bytes (TT) of DS1LSTAR. (The last right-
    most byte of DS1LSTAR being the 'R' part of the 'TTR'.)

    For DSNTYPE=LARGE however, the size of the dataset is 3 bytes
    in size (i.e. TTT, not just TT). So you use the first two
    high-order bytes of DS1LSTAR (just like you do for for normal
    format datasets), but in addition to that, the high-order byte
    of the 3-byte TTT is kept in the DS1TTTHI field.
   
    For DSNTYPE=EXTENDED, the size of the dataset in tracks is of
    course 4 bytes in size (TTTT), with the low-order 2 bytes of
    that 4-byte TTTT coming from the high-order two bytes of the
    DS1LSTAR field (just for for normal/large format datasets),
    but the two HIGH-order bytes of the 4-byte TTTT is in DS1TRBAL.

            SUMMARY OF DATASET SIZE IN NUMBER OF TRACKS:

    Normal:      TT   =                high-order 2 bytes of ds1lstar
    Large:      TTT   =   ds1ttthi(1), high-order 2 bytes of ds1lstar
    Extended:  TTTT   =   ds1trbal(2), high-order 2 bytes of ds1lstar

*/

Yes?

[GregoryTwin]

excellent!

[Fish-Git]

HELP!  :-(

I've made the necessary changes to dasdseq, but I am unable to verify they are good, due to my inability to create a Large or Extended data set with your JCL.

Attached are the output listings of my 3 attempts. They all failed:

• S0W1 IBMUSER FISH FISHUNLD JOB00267 2024-01-15 17.54.59.pdf
• S0W1 IBMUSER FISH FISHUNLD JOB00269 2024-01-15 18.02.34.pdf
• S0W1 IBMUSER FISH FISHUNLD JOB00270 2024-01-15 18.05.09.pdf

 
As you can see, in my 1st attempt, I tried using IEBDG, but it failed with: IGD01014I DATA SET ALLOCATION REQUEST FAILED - SPECIFIED DATACLAS BASE DOES NOT EXIST.

In my 2nd attempt, I tried using ADRDSSU, but it failed with: IEF257I FISHUNLD ADRDSSU LARGE - SPACE REQUESTED NOT AVAILABLE.

Finally, in my 3rd attempt, I changed my output VOL=SER to C5USR1 (since it appears to have very little data on it and thus should have the most space). But it too failed with an even weirder (more unusual) error:

HISTORIC RETURN CODE IS 192 DIAGNOSTIC INFORMATION IS 04160055
IGD306I UNEXPECTED ERROR DURING IGGDAC02 PROCESSING
RETURN CODE 4 REASON CODE 85
THE MODULE THAT DETECTED THE ERROR IS IGDVTSDA
SMS MODULE TRACE BACK - VTSDA VTSCR SSIRT
SYMPTOM RECORD CREATED, PROBLEM ID IS IGD00000

So I'm completely stuck and need Help!  :-(

(Remember, I know almost nothing about z/OS!!)

[GregoryTwin]

Hello.
Firstly. please take this file:

• https://drive.google.com/file/d/1vuJgyPvgBN1q7p1vBXtR1nww7vgGhaXG/view?usp=drive_link

This is zipped image of 3390-27 disk with 2 dat sets: LARGE1.DATASET and LARGE2.DATASET. The 1st data set is a result of IEBDG job shown above - 5 200 000 000 bytes x'FF'. The 2nd data set created with slightly different job;

//IEBDG    EXEC PGM=IEBDG
//SYSPRINT DD  SYSOUT=*
//LARGE    DD  DSNAME=LARGE2.DATASET,UNIT=3390,DISP=(,KEEP),
//             DCB=(RECFM=FB,LRECL=80,BLKSIZE=8000),
//             SPACE=(TRK,(150000,150000)),VOL=SER=111111,
//             DSNTYPE=LARGE
//SYSIN    DD  DATA
  DSD OUTPUT=(LARGE)
  FD  NAME=SEQ,LENGTH=15,INDEX=1,FORMAT=ZD
  REPEAT QUANTITY=65000,CREATE=1
  CREATE QUANTITY=1000,NAME=(SEQ),FILL=X'40'
  END

All records here are different: the first 15 positions contains a sequential number.

Secondly, about your failed jobs: the 1st and 2nd failures are quite clear. The 3rd was caused by missed space. I do not know, what C5USR1 volume is, which type/model? You need 3390-27 to allocate 150000 tracks, or at least, empty 3390-9 (3390-9 has 10,017 cylinders e.g. 150255 tracks.

[Fish-Git]

Secondly, about your failed jobs: the 1st and 2nd failures are quite clear.

It might be quite clear to YOU!   But it isn't clear to me!  ;-)

The 3rd was caused by missed space.

Where? I do not see any missed space.

I do not know, what C5USR1 volume is, which type/model?

Volume C5USR1 is from the z/OS 2.5c ADCD distribution. It is a 3390-9.

You need 3390-27 to allocate 150000 tracks, or at least, empty 3390-9 (3390-9 has 10,017 cylinders e.g. 150255 tracks.

A cckdmap summary of volume C5USR1 reports as follows:

CCKDMAP -i "Q:/CCKD64/zOS 2.5c (ADCD)/c5usr1.cckd64"
12:14:48.482 CCKDMAP -i c5usr1.cckd64 started; process-id = 8748 (0x0000222C)
12:14:48.636 HHC02499I Hercules utility CCKDMAP - Compressed dasd file map - version 4.7.0.11047-SDL-DEV-g8a6cfd87-modified
12:14:48.636 HHC01414I (C) Copyright 1999-2023 by Roger Bowler, Jan Jaeger, and others
12:14:48.637 HHC01417I ** The SDL 4.x Hyperion version of Hercules **
12:14:48.637 HHC01415I Build date: Jan 15 2024 at 12:27:40
12:14:48.637 HHC03020I
12:14:48.637 HHC03021I CCKDMAP of:     "Q:/CCKD64/zOS 2.5c (ADCD)/c5usr1.cckd64"
12:14:48.640 HHC03007I File size:      (2,424,016 bytes)
12:14:48.647 HHC03022I
12:14:48.647 HHC03022I   dh_devid:      CKD_C064        (64-bit CCKD64 base image)
12:14:48.647 HHC03022I   dh_heads:      15
12:14:48.647 HHC03022I   dh_trksize:    56832
12:14:48.648 HHC03022I   dh_devtyp:     0x90            (3390-9)
12:14:48.648 HHC03022I   dh_fileseq:    0x00
12:14:48.648 HHC03022I   dh_highcyl:    0
12:14:48.648 HHC03022I   dh_serial:     278896508142
12:14:48.648 HHC03023I
12:14:48.648 HHC03023I   cdh_vrm:       0.3.1
12:14:48.648 HHC03023I   cdh_opts:      0x40
12:14:48.648 HHC03023I   num_L1tab:     587
12:14:48.648 HHC03023I   num_L2tab:     256
12:14:48.649 HHC03023I   cdh_cyls:      10017           (150,255 tracks)
12:14:48.649 HHC03023I   cdh_size:      0x000024FCD0    (2,424,016 bytes)
12:14:48.649 HHC03023I   cdh_used:      0x000024FCD0    (2,424,016 bytes)
12:14:48.649 HHC03023I   free_off:      0x0000000000    (old format)
12:14:48.649 HHC03023I   free_total:    0x0000000000    (0 bytes)
12:14:48.650 HHC03023I   free_largest:  0x0000000000    (0 bytes)
12:14:48.650 HHC03023I   free_num:      0
12:14:48.650 HHC03023I   free_imbed:    0
12:14:48.650 HHC03023I   cdh_nullfmt:   0               (ha r0 EOF)
12:14:48.650 HHC03023I   cmp_algo:      1               (zlib)
12:14:48.650 HHC03023I   cmp_parm:      -1              (default)
12:14:48.722 HHC03020I
12:14:48.723 HHC03043I Total active tracks    = 75 tracks
12:14:48.723 HHC03044I Avg. L2-to-track seek  = 0.011 MB
12:14:48.723 HHC03020I
12:14:48.733 CCKDMAP -i c5usr1.cckd64 ended; rc=0

As you can see, the volume is a 3390-9 with 150,255 total tracks, of which only 75 tracks are in use. Thus it has over 150,000 tracks not in use. I would think that would have been plenty.

BUT NONE OF THAT IS IMPORTANT NOW, since you have kindly provided me with a copy of the exact same volume you were trying to unload with dasdseq. THANK YOU!

I will try my dasdseq on your "111111.cckd" image and let you know whether the problem has now been fixed or not.

Please Stand By......

[GregoryTwin]

Hi,

Are you sure that nothing else allocated on the C5USR1 at the execution of this job? What about temporary data (SYSOUT, SYSIN)? Does C5USR1 mounted as PRIVATE? (Use D U,DASD console command to check).

Anyway, 150000 track is very close to limit, so you can try reduce space to SPACE=(TRK,(110000,110000))
BTW. Expected IEBDG statement would be: CREATE QUANTITY=65000,FILL=X'40'

But this is not correct, because IEBDG limits QUANTITY to 64535 (2 bytes he-he). To avoid this limitation, I should use REPEAT+CREATE. However, I would say that IEBDG (aka "Data Generator") is a really wonderful program. It's written more than 50 years ago, but is still very useful and works perfectly...

Thank you very much for your time spent!

[Fish-Git]

Please Stand By......

DONE!  It appears to be working perfectly!  The proof is below:

• aaa.txt
• bbb.txt

"aaa.txt" is the output of the old/current broken version of dasdseq.
"bbb.txt" is the output of the new/fixed version of dasdseq.

As you can see, in the old version, it shows "startrack(42797)" and "track(A72D)" on the "DS1LSTAR" messages, and "25678400" for the "Records written" message. (I added the "verbose 1 1" option to the dasdseq command.)

With the new fixed version, it shows "startrack(108333)" and "track(1A72D)" on the "DS1LSTAR" messages, and "65000000" for the "Records written" message.

Additionally, you can also see in the new version a new HHC02696I Data set format is Large message too, and that a directory display shows both files now having their proper size.

To be extra thorough, I also visually inspected the "large2.dataset" output file with a hex editor too, and verified that the last record in the "large2.dataset" output file had "65000000" as its sequence number.

So everything looks good to me!  :)

So I've committed the fix. For reference, the commit hash is f2bbab8

Feel free to verify the fix for yourself if you wish.

THANK YOU for your help, Gregory!   I could not have done it without your kind help!  :-D

Closing issue.

[Fish-Git]

Are you sure that nothing else allocated on the C5USR1 at the execution of this job? What about temporary data (SYSOUT, SYSIN)? Does C5USR1 mounted as PRIVATE? (Use D U,DASD console command to check).

I don't know. But it's all moot. I was able to complete my testing using your "111111.cckd" dasd image.

Anyway, 150000 track is very close to limit, so you can try reduce space to SPACE=(TRK,(110000,110000))
BTW. Expected IEBDG statement would be: CREATE QUANTITY=65000,FILL=X'40'

Unnecessary. I was able to complete my testing using your "111111.cckd" dasd image.

... IEBDG (aka "Data Generator") is a really wonderful program. It's written more than 50 years ago, but is still very useful and works perfectly...

In the hands of someone like you who knows what they're doing? I'm sure it is!

In my hands though?  Not so much.  :(

But again, it doesn't matter. The point is moot. I was able to complete my testing using the "111111.cckd" dasd image that you kindly uploaded. Thank you for that! Much appreciated.

[GregoryTwin]

Great!