Good point! Could be "guaranteed synchronous write" in the SMS storage class definition. But what if have a non-SMS managed PDSE?
Thanks. Karl -----Original Message----- From: IBM Mainframe Discussion List [mailto:[email protected]] On Behalf Of Bernd Oppolzer Sent: Friday, March 20, 2015 5:01 PM To: [email protected] Subject: Re: Intermittent, not consistently reproducible problems with PDSEs on z/OS V2R1 (incl. infrequent S0F4-20 RSN 1C0752EE ABENDs) so there is something strange that IEBGENER does to empty PDSE members (when closing them) on your system that it does not do on others' systems? Only meant to summarize the conclusions so far ... isn't there a mechanism to cache PDSE directories system wide, to fasten up library searches? Could this have some effect here? Just a very wild guess ... Kind regards Bernd Am 20.03.2015 um 16:06 schrieb Henn, Karl: > I created the following members, in the order shown below > > DUMMY (empty) > AAAA > ZZZZ > BBBB > NNNN (empty) > EEEE > > EB1135I IEBCOPY FMID HDZ2210 SERVICE LEVEL UA74516 DATED 20140814 DFSMS > 02.01.00 z/OS 02.01.00 HBB7790 > EB1013I COPYING FROM PDSE INDD=SYSUT1 VOL=ST123B DSN=SYSADM.TEST.PDSE#A1 > EB1014I TO PDSE OUTDD=SYSUT2 VOL=ST113B DSN=SYSADM.TEST.PDSE#A2 > GW01551I MEMBER AAAA HAS BEEN COPIED > GW01551I MEMBER BBBB HAS BEEN COPIED > GW01551I MEMBER DUMMY HAS BEEN COPIED GW01551I MEMBER EEEE HAS BEEN > COPIED GW01551I MEMBER NNNN HAS BEEN COPIED GW01551I MEMBER ZZZZ HAS > BEEN COPIED GW01550I 6 OF 6 MEMBERS WERE COPIED EB147I END OF JOB - 0 > WAS HIGHEST SEVERITY CODE > > In SYSADM.TEST.PDSE#A2, members AAAA, ZZZZ, BBBB, and EEEE contain *only* > blanks. It looks as if no data was copied at all. The lexical order of the > member names does not seem to play a role. > > In addition to that, I did another test. > > (a) I created the members as shown above. > > (b) *Before* copying, I edited *all* of the empty members: DUMMY and NNN. I > just went in there with ISPF EDIT, and typed SAVE in the command line, > without having entered any data. > > (c) Ran IEBCOPY - and everything was fine! > > > So , something is wrong with the empty members in the copy source (PDSE#A1). > Once they are opened and closed once, the error is gone. I must say that I > can hardly see how this could be related to anything we have set up wrong in > our installation. On the other hand, Lizette and Leonardo have indicated that > the problem does not occur on their respective z/OS V2R1 systems. > > Thanks. > > Karl > > > -----Original Message----- > From: IBM Mainframe Discussion List [mailto:[email protected]] > On Behalf Of Bernd Oppolzer > Sent: Friday, March 20, 2015 2:31 PM > To: [email protected] > Subject: Re: Intermittent, not consistently reproducible problems with > PDSEs on z/OS V2R1 (incl. infrequent S0F4-20 RSN 1C0752EE ABENDs) > > IMO, this is a bug. > > Does the bug depend on the names of the members? > > I would think that the bug only appears, if the empty member (in this case > DUMMY) has a name lexically lower than the non-empty member (in this case > ZZZZ). > > Could you give this a try? I don't have access to a z/OS system today, so I > can't try it myself. > > What if there are non-empty members before the first empty member? > Maybe it works in this case, too. > > Would you keep us informed? I would appreciate it. > > Thank you, kind regards > > Bernd > > > > Am 20.03.2015 um 14:03 schrieb Karl Henn: >> Hello, >> >> The description of the following "problem" (if you prefer to call it such) >> sounds a little unusual, and the sample JCL looks - and is! - trivial. I >> still kindly ask the world to please read it, and maybe try it out. My >> primary goal is to make sure that I'm not simply just out of my mind. >> >> Ever since we're running z/OS V2R1, complaints about alleged problems with >> PDSEs were forwarded to me, sporadically. The error descriptions were fairly >> fuzzy, ranging from intermittent S0F4-20 RSN 1C0752EE ABENDs, problems with >> empty members, as well as incomplete copies of PDSEs with IEBCOPY CC 00. >> Since none of the problems were really consistently reproducible, I kind of >> just brushed away the issue for a while. >> >> The S0F4 ABENDs I could not ignore, though. The complaints didn’t stop, >> either. So, in order to narrow down a possible problem, I started >> experimenting - yielding a surprising result. >> >> Please take a look at the following JCL. Don’t laugh, but as simple as it >> may look, it *sometimes* does not work. >> >> >> //*------------------------------------------------------------------* >> //DELETE EXEC PGM=IDCAMS,COND=(0,NE) >> //SYSPRINT DD SYSOUT=* >> //SYSIN DD DATA >> DELETE <hlq>.TEST.PDSE#A1 NVSAM SCRATCH >> DELETE <hlq>.TEST.PDSE#A2 NVSAM SCRATCH >> SET MAXCC=00 >> /* >> //*------------------------------------------------------------------* >> //ALLOC EXEC PGM=IEFBR14,COND=(0,NE) >> //SYSPRINT DD SYSOUT=* >> //PDSE#A1 DD DISP=(NEW,CATLG),DSN=<hlq>.TEST.PDSE#A1, >> // RECFM=FB,LRECL=80,SPACE=(CYL,(1,1,1)),UNIT=SYSALLDA, >> // DSNTYPE=LIBRARY >> //PDSE#A2 DD DISP=(NEW,CATLG),DSN=<hlq>.TEST.PDSE#A2, >> // RECFM=FB,LRECL=80,SPACE=(CYL,(1,1,1)),UNIT=SYSALLDA, >> // DSNTYPE=LIBRARY >> //*------------------------------------------------------------------ >> * //CRE#A1D EXEC PGM=IEBGENER,COND=(0,NE) //SYSPRINT DD SYSOUT=* >> //SYSIN DD DUMMY >> //SYSUT2 DD DISP=SHR,DSN=<hlq>.TEST.PDSE#A1(DUMMY) >> //SYSUT1 DD DSN=NULLFILE,RECFM=FB,LRECL=80,DSORG=PO >> //*------------------------------------------------------------------ >> * //CRE#A1Z EXEC PGM=IEBGENER,COND=(0,NE) //SYSPRINT DD SYSOUT=* >> //SYSIN DD DUMMY >> //SYSUT2 DD DISP=SHR,DSN=<hlq>.TEST.PDSE#A1(ZZZZ) >> //SYSUT1 DD DATA >> ###ZZZZ### >> /* >> //*------------------------------------------------------------------ >> * >> //COPY#A1 EXEC PGM=IEBCOPY,COND=(0,NE) //SYSPRINT DD SYSOUT=* >> //SYSUT1 DD DISP=SHR,DSN=<hlq>.TEST.PDSE#A1 >> //SYSUT2 DD DISP=SHR,DSN=<hlq>.TEST.PDSE#A2 >> //SYSIN DD DUMMY >> //*------------------------------------------------------------------ >> * >> // >> >> All steps usually end with CC 00. So far, so good. Everything as expected. >> >> However, if you BROWSE member ZZZZ in the two data sets involved, the result >> is surprising, to say the least. >> >> BROWSE SYSADM.TEST.PDSE#A1(ZZZZ) >> Command ===> >> ******************************** >> -------------------------------- >> ###ZZZZ### >> 777EEEE7774444444444444444444444 >> BBB9999BBB0000000000000000000000 >> -------------------------------- >> >> >> BROWSE SYSADM.TEST.PDSE#A2(ZZZZ) >> Command ===> >> ******************************** >> -------------------------------- >> >> 44444444444444444444444444444444 >> 00000000000000000000000000000000 >> -------------------------------- >> >> Even though IEBCOPY (in step COPY#A1) ended with CC 00, and target member >> ZZZZ is *empty*. Even more intriguing: This *only* happens if an empty >> member (DUMMY in my example) already exists in the source PDSE, otherwise >> everything is fine and good. >> >> In my view, there are three possibilities. >> >> (1) I'm crazy. >> >> (2) In my 28 year as a Mainframe Systems Programmer, I somehow managed to >> miss a crucial piece of information. >> >> (3) There is a bad bug somewhere. >> >> Once I can rule out No. 1 and No. 2, I might turn to IBM. >> >> Thanks a lot in advance, everybody! >> >> Best Regards >> >> Karl >> >> --------------------------------------------------------------------- >> - 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 ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [email protected] with the message: INFO IBM-MAIN
