stephen.rodgers_at_turner.com (Tony Rodgers) wrote in message news:<a3d391e0.0110091143.1f0312a1_at_posting.google.com>...
> We've got an IOT which is partitioned by date range. We were able to
> test some loads (6M records) using sqlloader (not direct path) into
> the IOT before we partitioned it and the timing seemed acceptable. We
> have partitioned the IOT now and we're trying to load 12M rows into
> the IOT. The load has been running over 27 hours now and still not
> complete. When I look at v$session_wait for the session doing the
> load, it's always showing db file sequential read as the wait for the
> session. The files being read are the data files for this partitioned
> IOT. I can't figure out why it's doing so much reading, nor why it's
> running for so long. There isn't any other activity against this IOT
> and it's isolated onto its own disk array. I was just wondering if
> anyone else has seen something like this or had any ideas about things
> I could check. This is on an 8.1.6.0.0 database.
Have you tried sorting the ascii data on the partitioned column before
starting the load. I have the feeling it is going thru index spliting
because of unsorted data. If you sort it based on partitioned key then
leaf blocks will grow to the right. I am able to load 23M rows into
partitioned IOT (although only 4 columns) in 3 hours.
Received on Thu Nov 01 2001 - 13:06:53 CST
