3390-3 Support question

Discussion:

herc_fun

2004-12-10 15:05:04 UTC

I've seen it said "3390 mod3s MAY NOT WORK 'cause > 32K trks per
vol". So I've started doing some playing around with mod3s on
Turnkey. So far I've encountered no problems mounting a (full) mod3
volume and doing stuff like using RPF to edit PDSs well above
the "32K track line".

I was wondering if anyone has examples of mod3s NOT working. It
seems to me that CCHHR disk addressing shouldn't be an issue; perhaps
there are traps with TTR type processing? Even with TTR, if unsigned
halfword math is done, the # tracks on a mod3 can be handled fine.

Before I go any further looking for trouble, I thought I'd ask if
anyone has already found some. I am still looking at scenarios for
providing "native"[1] access to legacy mainframe data once big iron
disappears; so far, retaining CCKDDUMP'ed 3390-3 volumes from big
iron and accessing them with MVS 3.8J looks attractive.

Cheers,
Charlie

[1] Native meaning I can see PDSes complete with directory
statistics, I can manipulate sequential datasets to see whats in them
and transform them as required to port to non-MVS environment and so
on. I realise "newer" entities such as modern VSAM, PDSE, HFS etc
are all a no-go but (fortunately) there's not much of that stuff in
the legacy mainframe data I have under consideration.

------------------------ Yahoo! Groups Sponsor --------------------~-->
Make a clean sweep of pop-up ads. Yahoo! Companion Toolbar.
Now with Pop-Up Blocker. Get it for free!
http://us.click.yahoo.com/L5YrjA/eSIIAA/yQLSAA/fXFolB/TM
--------------------------------------------------------------------~->

Yahoo! Groups Links

<*> To visit your group on the web, go to:
http://groups.yahoo.com/group/turnkey-mvs/

<*> To unsubscribe from this group, send an email to:
turnkey-mvs-unsubscribe-***@public.gmane.org

<*> Your use of Yahoo! Groups is subject to:
http://docs.yahoo.com/info/terms/

Jim Morrison

2004-12-10 18:13:44 UTC

Permalink

... anyone has examples of mod3s NOT working.

Hi Charlie,

It has been a long, long time since I fooled with the
dasd mods, but just off the top of my head I think an
attempt to allocate more than X'7FFF' 3390 tracks to a
dataset fails. Check doc/results.txt in the dasd mods
for the status as I knew it back then, and maybe add:
ICKSADSF doesn't support 3390s (and probably 3380s).
There may be some good archive hits on the subject, but
the Yahoo search interface can be a bit cumbersome.

HTH.

------------------------ Yahoo! Groups Sponsor --------------------~-->
$4.98 domain names from Yahoo!. Register anything.
http://us.click.yahoo.com/Q7_YsB/neXJAA/yQLSAA/fXFolB/TM
--------------------------------------------------------------------~->

Yahoo! Groups Links

<*> To visit your group on the web, go to:
http://groups.yahoo.com/group/turnkey-mvs/

<*> To unsubscribe from this group, send an email to:
turnkey-mvs-unsubscribe-***@public.gmane.org

<*> Your use of Yahoo! Groups is subject to:
http://docs.yahoo.com/info/terms/

somitcw

2004-12-10 20:11:26 UTC

Permalink

Disk space allocation cannot handle
a free space block of over 32767 tracks.
i.e. Init a 3390-3 volume and put the VTOC
on VTOC(0,14,1) and try to allocate a
data set. Putting the VTOC in the center
of the volume should work better.

Some specific programs might also have
problems. JES2 cannot calculate the
number of track groups for SPOOL space
on a 3390-3 ( or 3390-1 ). Without SPOOL
space, JES2 does not work well. We have
the source and know where the code is, so
there is no big problem.

I have also lost labels, VTOCs, and
PDS directories on 3390 disk volumes,
but never bothered to find what was
causing it. If you use 3390 disk
volumes on MVS 3.8j, keep plenty of
back up. If you use other types of disk
volumes, still keep plenty of back up.

Post by herc_fun
I've seen it said "3390 mod3s MAY NOT
WORK 'cause > 32K trks per vol". So I've
started doing some playing around with
mod3s on Turnkey. So far I've encountered
no problems mounting a (full) mod3 volume
and doing stuff like using RPF to edit
PDSs well above the "32K track line".
I was wondering if anyone has examples of
mod3s NOT working. It seems to me that CCHHR
disk addressing shouldn't be an issue;
perhaps there are traps with TTR type
processing? Even with TTR, if unsigned
halfword math is done, the # tracks on a
mod3 can be handled fine.
Before I go any further looking for trouble,
I thought I'd ask if anyone has already found
some. I am still looking at scenarios for
providing "native"[1] access to legacy
mainframe data once big iron disappears; so
far, retaining CCKDDUMP'ed 3390-3 volumes
from big iron and accessing them with
MVS 3.8J looks attractive.
Cheers,
Charlie
[1] Native meaning I can see PDSes complete
with directory statistics, I can manipulate
sequential datasets to see whats in them and
transform them as required to port to non-MVS
environment and so on. I realise "newer"
entities such as modern VSAM, PDSE, HFS etc
are all a no-go but (fortunately) there's not
much of that stuff in the legacy mainframe
data I have under consideration.

------------------------ Yahoo! Groups Sponsor --------------------~-->
$4.98 domain names from Yahoo!. Register anything.
http://us.click.yahoo.com/Q7_YsB/neXJAA/yQLSAA/fXFolB/TM
--------------------------------------------------------------------~->

Yahoo! Groups Links

<*> To visit your group on the web, go to:
http://groups.yahoo.com/group/turnkey-mvs/

<*> To unsubscribe from this group, send an email to:
turnkey-mvs-unsubscribe-***@public.gmane.org

<*> Your use of Yahoo! Groups is subject to:
http://docs.yahoo.com/info/terms/

herc_fun

2004-12-13 14:35:42 UTC

Permalink

Great suggestion,

Post by somitcw
Disk space allocation cannot handle
a free space block of over 32767 tracks.
i.e. Init a 3390-3 volume and put the VTOC
on VTOC(0,14,1) and try to allocate a
data set. Putting the VTOC in the center
of the volume should work better.

That certainly makes sense. I'll try that approach. It also makes
it clear to me that I'm not going to get anything larger than a mod3
going. At least there's 16 3390s gen'ed and that makes a reasonable
size 3.8j DASD farm without having to get into a sysgen.

Post by somitcw
Some specific programs might also have
problems. JES2 cannot calculate the
number of track groups for SPOOL space
on a 3390-3 ( or 3390-1 ).

I was just looking at mod3s for "user data", no intention to muck
around with the the OS environment if I can avoid it.

Post by somitcw
I have also lost labels, VTOCs, and
PDS directories on 3390 disk volumes,
but never bothered to find what was
causing it.

Haven't encountered any such problems but will certainly keep alert.
Shadow files really help minimise the danger of major data loss.

Appreciate the help; I'll post back any concrete conclusions I come
to regarding mod3s after I do a wack of experimenting.

Regards,
Charlie

------------------------ Yahoo! Groups Sponsor --------------------~-->
$4.98 domain names from Yahoo!. Register anything.
http://us.click.yahoo.com/Q7_YsB/neXJAA/yQLSAA/fXFolB/TM
--------------------------------------------------------------------~->

Yahoo! Groups Links

<*> To visit your group on the web, go to:
http://groups.yahoo.com/group/turnkey-mvs/

<*> To unsubscribe from this group, send an email to:
turnkey-mvs-unsubscribe-***@public.gmane.org

<*> Your use of Yahoo! Groups is subject to:
http://docs.yahoo.com/info/terms/

somitcw

2004-12-13 20:07:25 UTC

Permalink

The usermod to add 3390 support may have
hardcoded the number of cylinders per disk
volume. You may have to up the number from
3390-1 size to 3390-3 ?

If the DOS contamination bit ever gets
turned on in the format 4 DSCB in the VTOC,
any space over what MVS thinks is on the
volume will have the data sets deleted.

Post by herc_fun
Great suggestion,

That certainly makes sense. I'll try that
approach. It also makes it clear to me
that I'm not going to get anything larger
than a mod3 going. At least there's
16 3390s gen'ed and that makes a reasonable
size 3.8j DASD farm without having to get
into a sysgen.

Post by somitcw
Some specific programs might also have
problems. JES2 cannot calculate the
number of track groups for SPOOL space
on a 3390-3 ( or 3390-1 ).

I was just looking at mod3s for "user data",
no intention to muck around with the the OS
environment if I can avoid it.

Post by somitcw
I have also lost labels, VTOCs, and
PDS directories on 3390 disk volumes,
but never bothered to find what was
causing it.

Jim Morrison

2004-12-13 20:29:35 UTC

Permalink

Post by somitcw
The usermod to add 3390 support may have
hardcoded the number of cylinders per disk
volume.

My recollection is that the dasd mods specify
a 3390-1 size, as does OS/390. There may be
a few instances where there are multiple DVCT
entries for different models but I think IBM
stopped doing it that way when they started to
run short of dasd device type bits.

Post by somitcw
You may have to up the number from 3390-1 size
to 3390-3 ?

My guess would be that isn't necessary. I believe
the VTOC describes how many tracks appear on the
volume (so VM minidisks work properly). It might
be in the volume label rather than the VTOC, I don't
recall exactly.

------------------------ Yahoo! Groups Sponsor --------------------~-->
$4.98 domain names from Yahoo!. Register anything.
http://us.click.yahoo.com/Q7_YsB/neXJAA/yQLSAA/fXFolB/TM
--------------------------------------------------------------------~->

Yahoo! Groups Links

<*> To visit your group on the web, go to:
http://groups.yahoo.com/group/turnkey-mvs/

<*> To unsubscribe from this group, send an email to:
turnkey-mvs-unsubscribe-***@public.gmane.org

<*> Your use of Yahoo! Groups is subject to:
http://docs.yahoo.com/info/terms/

herc_fun

2005-01-10 17:43:05 UTC

Permalink

I found the problems I went looking for; the bottom line is 3.8J
doesn't really support mod3s. In my testing, it seems you can mount
mod3s, mod9s and probably even mod27s under 3.8J and have no problem
accessing the datasets that are already on those volumes. The
problems all start when you try to create datasets on anything other
than a mod1.

What kind of problems? Bad things happen -
413-20 abends, 614-08 abends, command rejects on crazy cylinder
addresses, VTOC errors, and lots of other fun.

But, being the tinkering type, I had a look at Jim's 3380/3390 mods
and saw that the only thing that said anything about the size of
volumes was the IECZDTAB CSECT. After playing around with zapping
the 3390 size (in cylinders) it became clear to me that MVS3.8J space
management gets very confused if the volume contains more than 32767
tracks. When you try to allocate above the 32K track limit, you get
really wild cylinder addresses generated in the fmt4 DSCB and
everything goes downhill from there.

So, forget the mod3s and LONG LIVE THE MOD 1.96s!

I zapped IECZDTAB to say a 3390 has 2184 cylinders which would yield
32760 tracks. This doesn't appear to harm any existing mod1 volumes
in any way. Then I created my "mod1.96" volumes by doing a dasdinit
of a 3390 specifically requesting 2184 cylinders. Next, I used
ICKDSF to initialise the volume and everything was fine after that.
I have created thousands of datasets on my mod1.96 vols and have had
no problems.

What is the point of having a mod1.96 volume? It facilitates larger
disk storage farms under 3.8J with a more manageable number of
devices (and, given there are only 48 3390s in the Turnkey system, it
gives the potential for about 94GB of disk space without resorting to
a SYSGEN).

Here is the formula I used in case any of you kids want to try this
at home:

//AMASPZAP JOB
//ZAPPER EXEC PGM=AMASPZAP,REGION=2048K
//SYSPRINT DD SYSOUT=*
//SYSLIB DD DSN=SYS1.NUCLEUS,DISP=SHR
//SYSIN DD *
NAME IEANUC01 IECZDTAB
VER 94 0459
REP 94 0888
//

dasdinit -z -a c:\mvs38j\dasd\PUB100.vol 3390-3 PUB100

init unit(xxx) vtoc(0,1,44) noindex noverify owner(charlie)

Best regards,
Charlie

Post by Jim Morrison

Post by somitcw
The usermod to add 3390 support may have
hardcoded the number of cylinders per disk
volume.

Post by somitcw
You may have to up the number from 3390-1 size
to 3390-3 ?

Yahoo! Groups Links

<*> To visit your group on the web, go to:
http://groups.yahoo.com/group/turnkey-mvs/

<*> To unsubscribe from this group, send an email to:
turnkey-mvs-unsubscribe-***@public.gmane.org

<*> Your use of Yahoo! Groups is subject to:
http://docs.yahoo.com/info/terms/

somitcw

2005-01-11 04:44:58 UTC

Permalink

--- In turnkey-mvs-***@public.gmane.org,
"herc_fun" <***@h...> wrote:
- - - snipped - - -

Post by herc_fun
dasdinit -z -a c:\mvs38j\dasd\PUB100.vol 3390-3 PUB100

- - - snipped - - -

Why not:
dasdinit -z -a c:\mvs38j\dasd\PUB100.vol 3390-2 PUB100

Yahoo! Groups Links

<*> To visit your group on the web, go to:
http://groups.yahoo.com/group/turnkey-mvs/

<*> To unsubscribe from this group, send an email to:
turnkey-mvs-unsubscribe-***@public.gmane.org

<*> Your use of Yahoo! Groups is subject to:
http://docs.yahoo.com/info/terms/

herc_fun

2005-01-11 15:48:49 UTC

Permalink

Thanks for catching that error; actually both 3390-3 and 3390-2 are
wrong for what I was describing. The objective was to make a non-
standard size 3390 which I was jokingly calling a mod1.96 since it is
1.96 times the size of a mod1.

The command in my previous post should have read:

dasdinit -a -z c:\mvs38j\dasd\PUB100.vol 3390 PUB100 2184

i.e. It should NOT have a 3390 model number in the command and it
should explicitly state 2184 cylinders.

Sorry for any confusion, I'll watch my cut'n'paste closer next time.

Charlie

Post by somitcw
- - - snipped - - -

Post by herc_fun
dasdinit -z -a c:\mvs38j\dasd\PUB100.vol 3390-3 PUB100

- - - snipped - - -
dasdinit -z -a c:\mvs38j\dasd\PUB100.vol 3390-2 PUB100