Loading a VSAM data set: initial load with REPRO

Initial load is the moment an empty VSAM cluster becomes a populated production file—or the moment a conversion project proves it can. Unlike a tiny test file where you type six records by hand, real loads move millions of rows from sequential extracts, legacy dumps, or another VSAM cluster into a freshly defined structure. The mechanical tool is almost always IDCAMS REPRO, but the engineering work is everything around REPRO: proving the cluster attributes, sorting keyed input, choosing REPLACE semantics, reconciling counts, and making sure the catalog story matches what applications will open Monday morning. This page focuses on that surrounding discipline so beginners understand why “just run REPRO” is never the whole job.

What “initial load” means operationally

Initial load establishes the first complete population of records (and for a KSDS, the index structure that mirrors those keys). It usually happens once per major release, after a full redefine, or during platform migration. Steady-state processing afterward might insert or update records through COBOL programs, CICS transactions, or smaller batch extracts. Because initial load often runs with elevated authority and tight downtime windows, change tickets should list prerequisites explicitly: sort product level, temporary dataset space, restart checkpoints, and back-out steps if REPRO abends halfway.

KSDS: the sort is not optional decoration

A KSDS keeps records ordered by the primary key. REPRO writes sequentially and expects ascending key order for the input stream so the index can be built consistently as records arrive. If your extract is keyed on customer number but sorted by load timestamp, you must re-sort by customer number before REPRO. For files with duplicate keys allowed, your sort should be stable and consistent with UNIQUEKEY versus NONUNIQUEKEY expectations on DEFINE. Skipping the sort because “the data looked sorted” is a common source of weekend pages.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
//S001   EXEC PGM=SORT           (illustrative pattern)
//SORTIN DD DISP=SHR,DSN=extract.seq
//SORTOUT DD DISP=(,CATLG),DSN=&&SORTED,
//           SPACE=(CYL,(50,10)),UNIT=SYSDA
//SYSIN  DD *
  SORT FIELDS=(1,10,CH,A)
/*
//R001   EXEC PGM=IDCAMS
//SYSPRINT DD SYSOUT=*
//IN1    DD DISP=SHR,DSN=&&SORTED
//OUTVS  DD DISP=OLD,DSN=PROD.CUST.MASTER
//SYSIN  DD *
  REPRO INFILE(IN1) OUTFILE(OUTVS)
/*

ESDS and RRDS loading differences

ESDS

Entry-sequenced loads preserve arrival order. You still care about record length and block size compatibility, but you do not sort by a VSAM primary key because there is none. Logical delete flags or reorganization strategies may still apply depending on application design.

RRDS

Relative record datasets load by slot. Input ordering interacts with SKIP/COUNT and with how your application maps business keys to RRNs. Initial load jobs should document which RRNs are intentionally empty so later online programs do not treat them as errors.

Phase checklist

REPLACE, REUSE, and policy language

REPLACE tells REPRO it may overlay records in the target when keys or RRNs collide. That is powerful during reruns of a load job after a failure, but dangerous if two different business feeds accidentally share keys. REUSE relates to cluster reuse patterns on DEFINE and should be interpreted strictly per IBM text for your release. Security and audit teams may require DELETE+DEFINE instead of overlay semantics for certain datasets—follow governance, not personal taste.

Verification beyond “condition code zero”

• Record counts: Match input, REPRO messages, and application SQL or control totals when available.
• Extreme keys: Spot-check min and max keys with a read utility or program.
• Extents: LISTCAT to see whether secondary allocation triggered more than expected, signaling undersized primary.
• Smoke test OPEN: A tiny COBOL or utility read proves path, RACF, and AMP alignment for the consuming application ID.

Practice exercises

1. Write a sort card sketch for a KSDS whose key is bytes 15–24 character ascending.
2. List three reasons a load might succeed technically yet fail business validation.
3. Pair-read the main REPRO page and note two parameters you would add for a partial reload scenario.
4. Design a rollback paragraph for a change ticket if REPRO fails at 60 percent complete.

Explain like I'm five

Loading a KSDS is like filling a sticker book where every sticker has a number and the book only works if you put numbers in order. If you jump from 5 to 9 then back to 7, the book’s tabs get confused and the pages tear. Sorting first is your grown-up checking the numbers on the kitchen table before you paste. REPRO is the pasting step; it does not babysit the numbering for you.

Test your knowledge