We have PRODUCTION database on the IBM Mainframe platform (OS-390, MVS/ESA).
Last Sunday a heavy update (batch job) abended with

ORA-01562:
failed to extend rollback segment number 5 ORA-01628

The following are the data extracted from V$ROLLSTAT

SEG-NAM----XTNTS---RSSIZE---OPTSIZE----HWMSZ-- XTNDS-SHRNK
SYSTEM----------4------241664---------------241664-----0---0
PRB002----------2---10481664--10485760--15724544-----1---1
NEWROL---------2---10481664--10485760--10481664-----0---0
PRB003----------2---10481664--10485760--10481664-----0---0
PRB001----------2---10481664--10485760--10481664-----0---0
PRB004----------2---10481664--10485760--10481664-----0---0
PRB005----------2---10481664--10485760-304082944---56---6

It looks like the segment-troublemaker (#5) was expanded 56 times:
initial size is 5Mb, increment size is 5Mb,optimal size set to (5*2)Mb. (56+2)*5Mb=290Mb=304087040
Maximum Number of extents (249) apparently was not reached.

My question is - why? Why ORACLE responded with "ORA-01562"?

It's obvious, that the total highwatermark size (even if it was reached at the same point in all segments) is less than 350Mb.

On the other hand, the total size available in ROLLBACK tablespace is 620Mb.
(We have 3 VSAM datasets with total 13260 TRKs on 3390 triple density DASD)

What did I miss ?

[Edited by jrvladr on 01-09-2001 at 02:52 PM]