However, HSM Partial Release should consolidate extents, so a properly sized 
(big) secondary should keep the abends at bay

> -----Original Message-----
> From: IBM Mainframe Discussion List <[email protected]> On
> Behalf Of Paul Feller
> Sent: Wednesday, February 21, 2024 12:39 PM
> To: [email protected]
> Subject: Re: Something keeps releasing space on a large (annual) DS
> 
> [EXTERNAL EMAIL]
> 
> Bob as Steve said you might want to talk to your storge management team.
> What I think is happening is your dataset is getting a management class that
> has Partial Release set and then HSM is doing space management and releasing
> any unused space.  I've seen this happen before and it has happened to me.
> 
> Good luck.
> 
> Paul
> 
> -----Original Message-----
> From: IBM Mainframe Discussion List <[email protected]> On
> Behalf Of Steve Thompson
> Sent: Wednesday, February 21, 2024 1:59 PM
> To: [email protected]
> Subject: Re: Something keeps releasing space on a large (annual) DS
> 
> Ask the storage guys about the preferred method to allocate a file that will 
> get
> very large during production runs. And you don't want production to fail with
> a storage ABEND. And let them know about the current behavior. They may
> have made a change that is the cause of your problem.
> 
> Then you can modify the REXX code to include the class info they give you.
> 
> Also, I suggest you allocate in CYLS not TRKS in the case they are doing
> compression of space on data sets allocated in tracks...
> 
> I've seen various odd things done to recover space for files that can't be on 
> the
> tracks after 65xxx (forgot the last track accessible by the old NOTE/POINT 
> that
> causes products like Panvalet to break).
> 
> And if I remember correctly, compression/decompression is done by the
> access method. Double check to make sure this isn't a VSAM thing.
> 
> I've not had to deal with these features for a few years.... So things slip 
> from
> one's mind.
> 
> Steve Thompson
> 
> On 2/21/2024 2:33 PM, Bob Bridges wrote:
> > Ooh, now that's interesting!  The content of this file would lend
> > itself well to compression - all alphanumeric with a few parens,
> > colons and the like.  But what happens when someone needs to view it?
> > Does it compress automatically or is another step required?
> >
> > It's not something I can bring up now, because everyone's busy with a
> > z/OS upgrade.  But next month...
> >
> > ---
> > Bob Bridges, [email protected], cell 336 382-7313
> >
> > /* For Sale: Parachute.  Only used once, never opened, small stain. */
> >
> > -----Original Message-----
> > From: IBM Mainframe Discussion List <[email protected]> On
> > Behalf Of Michael Oujesky
> > Sent: Wednesday, February 21, 2024 13:49
> >
> > You might consider SMS compression to reduce the physical size of the file.
> > If you do, change the BLKSIZE to 32760 as SMS compression writes full
> > tracks and the BLKSIZE becomes logical (the size of the buffer used in
> > passing date to/from the application).
> >
> > --- At 11:44 AM 2/21/2024, Bob Bridges wrote:
> >> I'm not a sysprog (just a security geek), but I can at least allocate
> >> datasets, and at the start of this year it fell to me to allocate a
> >> new dataset in which are logged all changes made in the security system.
> >> Past year's log are in the 12000-track range, so I started with a
> >> smaller allocation while I took the time to talk to our sysprog about
> >> space requirements.  It's populated from a daily production job, by
> >> the way.
> >>
> >> When I re-allocated it, on his advice I tried a multi-volume and
> >> extended allocation (PS-E).  Almost immediately the job started
> >> bombing, claiming that the first four volumes it tried didn't have
> >> the necessary space to add an extension.  The sysprog is puzzled -
> >> says it should have looked in volumes that DO have the space, not the
> >> ones that don't.
> >>
> >> Second attempt (I don't count the temporary smaller allocation) I
> >> kept PS-E but dropped the multi-volume requirement.  I've never done
> >> one of those anyway, and don't trust 'em.  The system promptly
> >> dropped the extra tracks I allocated, and a day or two later the job
> >> started bombing with a B37-04.
> >>
> >> Third attempt: Forget PS-E (I'm unfamiliar with that too) and just
> >> used SPACE=(TRK,(9000,1000)).  That seemed to work for a whole week,
> >> but I just noticed that something, somewhere, has released extra
> >> space AGAIN;
> >> 3.4 tells me it's now 1960 tracks and 83%.  The job isn't bombing
> >> yet; some time later in the year I'm guessing it's going to.
> >>
> >> Pardon my frustration: WHAT THE HECK IS GOING ON?  Why does it keep
> >> releasing space although I never specified RLSE?  The sysprog doesn't
> >> know either - but he's an external contractor who just took over the
> >> system a few months ago and if it's something simple he may not be
> >> aware yet of ... I dunno, something in SMS maybe?
> >>
> >> Some wrinkles that may or may not be relevant:
> >>
> >> 1) The dataset is written using a REXX exec that calculates the DSN
> >> by reference to the current year.  This relieves folks from having to
> >> update the JCL every year, but maybe something about the way the exec
> >> does the allocate is causing the problem?  I'm guessing not, because
> >> as far as I now this job has run correctly for years.  But just in case:
> >>
> >>    "ALLOC DDN(CHG$$OT) DSN('<dsn>') MOD CATALOG REUSE",
> >>        "SPACE(300,30) CYLINDERS RECFM(V,B) LRECL(304) BLKSIZE(27998)"
> >>
> >> 2) I don't know anything about SMS, but could something there be
> >> releasing space?
> >>
> >> 3) What IS extended PS, anyway?  I'm told it allows more than 16
> >> extents, but a) how many more? And b) how else is it different?
> >>
> >> 4) I allocated the dataset each time using not batch JCL but 3.2 ...
> >> expecting there's no difference.
> > ----------------------------------------------------------------------
> > For IBM-MAIN subscribe / signoff / archive access instructions, send
> > email to [email protected] with the message: INFO IBM-MAIN
> 
> ----------------------------------------------------------------------
> For IBM-MAIN subscribe / signoff / archive access instructions, send email to
> [email protected] with the message: INFO IBM-MAIN
> 
> ----------------------------------------------------------------------
> For IBM-MAIN subscribe / signoff / archive access instructions, send email to
> [email protected] with the message: INFO IBM-MAIN

----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO IBM-MAIN

Reply via email to