Ron
I'm guessing that one of the loads receives the ORA-01555 and the other
doesn't. Does the load commit occasionally? There are some pretty good
papers on Metalink for this error.
Dennis Williams
DBA, 80%OCP, 100% DBA
Lifetouch, Inc.
[EMAIL PROTECTED]
-Original Message-
Sent:
Ron,
Are the sizes of the data manipulated by the batch load the same?
Is the db_block_size the same on each database?
What is the size of the rbs during processing? What is the hwmsize, shrinks,
extends from v$rollstat?
You may be encountering a situation where
They are both processing the same data, the same way.
-Original Message-
Sent: Tuesday, July 22, 2003 10:15 AM
To: Multiple recipients of list ORACLE-L
Ron
I'm guessing that one of the loads receives the ORA-01555 and the
other doesn't. Does the load commit occasionally? There are
Snapshot too old also occurs when there is a commit across an open cursor.
Keep the commit on the outermost cursor in your PL/SQL (if you are using
PL/SQL) IN other words committing on the inside of a cursor FOR loop can
cause the cursor snapshot to be overwritten in rollback by activity on the
Ron - I think you are feeling that one system is good and the other is bad.
My suggestion is to consider one system at the ragged edge of failing and
the other system over the edge and in the ditch. Next week with a different
set of data, they may both well fail or both succeed. With ORA-01555
Title: Message
You
can try committing less frequently in the batch load e.g. if you are committing
every 1 rows then try every 3 (assuming of course that the you have
space left in the RBS to allow for this) or you can try adding rollback
segments.
Metalink docs give some good ideas
