I have a 9.2.0.8 database that I do some bulk loads in once a month. Generally this is about 750,000 records. I wrote a control file for sqlldr and all appears to work fine. However, the issue is that it still generates redo even though I have the unrecoverable option in it. Here is the CTL file. options (DIRECT=TRUE,ROWS=5000) unrecoverable load data infile 'c:\loadarrest\part1.txt' truncate into table dcsodba.jms_arrests fields terminated by "," OPTIONALLY enclosed by '"' (jms_number, control_number, admitted_date DATE "MM/DD/YYYY hh24:mi", released_date DATE "MM/DD/YYYY hh24:mi", warrant_number, arrested_charge, convicted_charge, place_of_birth, foreign_born, blank_null, suppress) The data loads with no errors and it loads in about 15 or 20 secs. No problem with the speed. But, today I checked and I noticed that I generated about 3 or 4 archive logs when I ran the load (I only have 10meg logs.) I execute the script in the directory where the infile is loaded. The command I use is sqlldr control=loadarrest_ctl I am then prompted for username and password and I enter those. So, all things appear to work...accept that I generated redo. I checked back a few months and realized I've always been generating it. So, I'm stumped. This appears to be the correct format. What am I missing?