The error messages are crucial for this kind of problem. We had a case recently of a (DF)SORT job failing on an unusually large number of records. The error message contained a reference to '64K', which turned out to be the maximum number of tracks in a conventional data set. DFSORT assumes when calculating the number of SORTWKxx datasets that each can/will be huge with the DSNTYPE=LARGE attribute so that the data set can exceed the old 64K track limit. However, because we had specified VIO=Y in the installation parms (overriding the product default), DFSORT was constrained by design to allocate only non-large work files.
I agree that the product should be allowed to analyze the input file for sizing. FILSZ should be needed only for tape input, where the actual size is indeterminate at the outset. . . JO.Skip Robinson SCE Infrastructure Technology Services Electric Dragon Team Paddler SHARE MVS Program Co-Manager 626-302-7535 Office 323-715-0595 Mobile [email protected] From: "Gibney, Dave" <[email protected]> To: [email protected] Date: 06/14/2012 12:33 PM Subject: DFSORT 180M recs Sent by: IBM Mainframe Discussion List <[email protected]> This is for a friend at another installation. I Sync rather than DF, so I'm not the best to answer the question. I have already asked for the error message detail :) I've also said to not use the FILSZ parm and let DFSORT figure it out. I am having trouble getting dfsort to sort 180,000,000 records. I think it is failing for a lack of space. The records have an average length of 261 bytes. The key is the first 25 characters of the record and there should be no duplicates. The dfsort parms I have tried include: OPTION MAINSIZE=64M,DYNALLOC=(SYSDA,20), FILSZ=E180000000,AVGRLEN=261 The DYNALLOC tells the sort to create 20 sortwork files. Maybe I need to create 50 sortwork files? FILSZ tells sort approximately how many records are to be sorted and sort is supposed to get the space it needs to do it. If you were going to sort this in syncsort how would you do it? One idea I had was to break the main file up into 30 million record groups. Sort those and then merge them back together into one file. I haven't tried that yet. It seems like I should be able to sort the large number of records. Dave Gibney Information Technology Services Washington State University ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [email protected] with the message: INFO IBM-MAIN
