Compressing a Partitioned Data Set

A partitioned data set will contain unused areas (sometimes called gas) where a deleted member or the old version of an updated member once resided. This unused space is only reclaimed when a partitioned data set is copied to a new data set, or after a compress-in-place operation successfully completes. It has no meaning for a PDSE and is ignored if requested.

The simplest way to request a compress-in-place operation, is to specify the same ddname for both the OUTDD and INDD parameters of a COPY statement, without specifying the SELECT / EXCLUDE statement.

However, a compress is actually performed when both the input and output is the same data set on the same volume. For example, this job step will compress data set Pacanowska:

//COMPRESS    EXEC    PGM=IEBCOPY
//A    DD  DSNAME='Pacanowska',DISP=OLD
//B    DD  DSNAME='Pacanowska',DISP=OLD
//SYSIN DD *
       COPY OUTDD=B,INDD=A

If multiple entries are made on the INDD statement, a compress-in-place occurs when any of the input ddnames matches the OUTDD name. The compress operation is performed in the same relative order as the ddnames in the INDD list.

For example, consider the COPY statement:

       COPY OUTDD=B,INDD=(A,B,C,B)

The data set for ddname A is copied to ddname B
The data set B is compressed
ddname C is copied to ddname B
The data set B is compressed again.

It is a good idea to make a copy of the data set that you intend to compress-in-place before you actually do so. You can use the same execution of IEBCOPY to do both, and a following job step could delete the backup copy if the IEBCOPY job step ends with a return code of 0.

Attention: A partitioned data set can be destroyed if IEBCOPY is interrupted during processing, for example, by a power failure, abend, TSO attention, or I/O error. Keep a backup of the partitioned data set until the successful completion of a compress-in-place.

Attention: Do not compress a partitioned data set currently being used by more than one user. If you do, the other users will see the data set as damaged, even though it is not. If any of these other users update the partitioned data set after you compress it, then their update will actually damage the partitioned data set.