deferring index creation with an option to not have to log the rebuild

This was originally opened in Aug 2013 RIMS database (as 8556)

In Aug 2013 one of the ids-System customers, an insurance company, had to do an offline table reorg, which had to rebuild index at the

end. This is an HADR customer, so index is fully logged, except this runs into log full and index is marked bad. After this, any

access triggers rebuild, but always fails with log full. This repeats throughout the evening. The system was set up years ago, with

an active log configuration that would be big enough for the table / index size. But now things are much bigger. Not only the index

is much bigger now, but more significantly is the concurrent transactions during the time index was being rebuild. The rebuild is a

long operation, so all the transactions performed concurrently during the rebuild would require significantly more log space.

Extreme steps were taken (customer set logsecond to -1, change config so it is not fully logged, stopped HADR and reinitialized the

standby after wards) to get customer out of this situation. After this, they have increased active logs (so the largest index is now

1/4 of the total active log space). This will buy them some time, until table / index grows significantly again.

They are looking for improvements, in particular, they are asking for what z/OS has for deferring index creation with an option to not

have to log the rebuild, but generate a copy image for the index, much like load does.