I recently found myself recreating an Oracle 10g database on Red Hat Linux remotely. Long story. I found these two resources to be very helpful:
Here is what I mostly ended up with (I pieced it back together when I finally got it to work). I've removed paths, passwords, etc. or anything that is specific to the actual database I was working on. This is pretty much a combination of the above resources with a couple changes.
connect / as sysdba
shutdown abort
startup nomount
CREATE DATABASE oradb
controlfile reuse
MAXINSTANCES 8
MAXLOGHISTORY 226
MAXLOGFILES 16
MAXLOGMEMBERS 4
MAXDATAFILES 1022
DATAFILE 'system01.dbf'
SIZE 540M REUSE
AUTOEXTEND ON NEXT 10240K MAXSIZE UNLIMITED
EXTENT MANAGEMENT LOCAL
SYSAUX DATAFILE 'sysaux01.dbf'
SIZE 350M REUSE
AUTOEXTEND ON NEXT 10240K MAXSIZE UNLIMITED
DEFAULT TEMPORARY TABLESPACE TEMP TEMPFILE 'temp01.dbf'
SIZE 25M REUSE
AUTOEXTEND ON NEXT 1024K MAXSIZE UNLIMITED
UNDO TABLESPACE "UNDOTBS1"
DATAFILE 'undotbs01.dbf'
SIZE 100M REUSE
AUTOEXTEND ON NEXT 5120K MAXSIZE UNLIMITED
CHARACTER SET UTF8
NATIONAL CHARACTER SET utf8
LOGFILE
GROUP 1 ('redo01a.log',
'redo01b.log') SIZE 20M,
GROUP 2 ('redo02a.log',
'redo02b.log') SIZE 20M,
GROUP 3 ('redo03a.log',
'redo03b.log') SIZE 20M
USER SYS IDENTIFIED BY "password"
USER SYSTEM IDENTIFIED BY "password";
@?/rdbms/admin/catalog.sql
@?/rdbms/admin/catproc.sql
connect system/password
@?/sqlplus/admin/pupbld