Mike, Did you miss the "assuming the PDSE has no free blocks and cannot be extended"?
I was just curious if the PDSE logic mimicked the PDS behavior of making a distinction in the failure response for a full and non-extendable PDSE depending on whether the no-more-space failure is first detected when trying to allocate a free block to the directory or when trying to allocate one for member data. Or do both these cases produce an identical ABEND failure along the lines of PDSE out-of-space-and-unable-to-extend? JC Ewing On 10/22/20 10:23 PM, Mike Schwab wrote: > If all the PDSE directory blocks are full it grabs another block for > the directory, it can't run out of space unless the entire data set > cannot be expanded. > > On Thu, Oct 22, 2020 at 5:30 PM Joel C. Ewing <[email protected]> wrote: >> I would assume a directory entry must be created before attempting to >> allocate space for the contents of a new PDSE member. So, assuming the >> PDSE has no free blocks and cannot be extended, do you get a different >> type of ABEND if the out-of-space condition occurs at directory entry >> creation time because a new directory block just happens to be needed vs >> finding space in an existing directory block and then hitting the >> out-of-space condition trying to allocate a block for the member data? >> With no free blocks, obviously no new members can be added to the PDSE, >> but it looks like it might still be possible that the failure could be >> reflected differently to the user or program depending on purpose for >> which a block were needed at the initial point of failure. >> >> In a pathological case where you were just adding a very large number >> of Alias directory entries pointing to existing members, I would think >> you could use all remaining free blocks in the PDSE just for directory >> blocks without allocating any new blocks for member content, so if PDSE >> block allocation failure makes any distinction between failures >> occurring when a directory block is needed vs a member data block, that >> would be another case that might be reflected as a shortage of directory >> space. >> JC Ewing >> >> On 10/22/20 10:52 AM, Charles Mills wrote: >>> Putting it differently, there is no distinction between "member data space" >>> and "directory entry space." Being out of one is being out of both. A PDSE >>> of 10 tracks could equally well hold one member of ~500K or lots and lots >>> of tiny or "null" members. A mischievous programmer adding an unbounded >>> number of empty members would be no different in effect from a mischievous >>> programmer adding one member of unbounded size. >>> >>> Charles >>> >>> >>> -----Original Message----- >>> From: IBM Mainframe Discussion List [mailto:[email protected]] On >>> Behalf Of R.S. >>> Sent: Thursday, October 22, 2020 7:29 AM >>> To: [email protected] >>> Subject: Re: emptying a PDS: was RE: [IBM-MAIN] getting XCFAS down >>> >>> W dniu 22.10.2020 o 15:12, Paul Gilmartin pisze: >>>> On Thu, 22 Oct 2020 13:50:44 +0200, R.S. wrote: >>>> >>>>> Remark: while shortage of space is possible in PDSE, then shortage of >>>>> directory blocks is not possible. >>>>> >>>> What happens if an inquisitive programmer mischievously adds an >>>> unbounded number of empty members to a small PDSE? Or adds >>>> numerous aliases to a nearly full PDSE? >>> My guess: x37 abend or next extent. This is NOT directory full, it is >>> lack of space. >>> >>> ... >> >> -- >> Joel C. Ewing >> > > -- Joel C. Ewing ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [email protected] with the message: INFO IBM-MAIN
