5 Oracle Ultra Search Postinstallation Information

Step 1: Tune the Oracle Database

Increase the Size of the Oracle Redo Logs, if necessary

Every instance of an Oracle database has an associated online redo log, which is a set of two or more online log files that record all committed changes made to the database. Online redo logs protect the database in the event of an instance failure. The size of redo log files determines the frequency of redo log file switches. This, in turn, significantly impacts text indexing speed. To reduce the frequency of log file switches, ensure that the redo log files are each 100MB or more.

The following section lists some tips on how to increase the redo log file sizes, if necessary. Enter the statements in the following section with the appropriate Oracle administrator privileges.

1. Locate redo log files and determine their sizes:

   SELECT v$logfile.member, v$logfile.group#, v$log.status, v$log.bytes
   FROM v$log, v$logfile
   WHERE v$log.group# = v$logfile.group#;
   
   

2. Add larger redo log files:

   ALTER DATABASE ADD LOGFILE 'redo_log_directory/newredo1.log' size 100m;
   ALTER DATABASE ADD LOGFILE 'redo_log_directory/newredo2.log' size 100m;
   ALTER DATABASE ADD LOGFILE 'redo_log_directory/newredo3.log' size 100m;
   
   

   A production database should have more log members for each log group, and different storage devices should be used to increase performance and reliability.

3. Drop the old log files. For each old redo log file, enter the ALTER SYSTEM SWITCH LOGFILE statement until that log file's status is INACTIVE. This is necessary to ensure that Oracle is not using that log file when you try to drop it.

   Then, drop the old redo log file with the following statement:

   ALTER DATABASE DROP LOGFILE 'redo_log_directory/redo01.log';
   ALTER DATABASE DROP LOGFILE 'redo_log_directory/redo02.log'; 
   ALTER DATABASE DROP LOGFILE 'redo_log_directory/redo03.log';
   
   

4. Manually delete the old log files from the file system. For each old redo log file, use the appropriate operating system statement to delete the unwanted log file from the file system.

Increase the Size of the Undo Space

Every Oracle database must have a method of maintaining information that is used to roll back or undo changes to the database. Such information consists of records of the transactions, primarily before they are committed. Oracle refers to these records collectively as undo. The undo space created by the Oracle Installer may be too small. It is recommended to use automatic undo management and increase the undo space.

Tune Oracle Initialization Parameters

Set the following values in the initialization file:

• PROCESSES: Set this to 50 or more.

• SORT_AREA_SIZE: Set this to 5MB or more.

• SORT_AREA_RETAINED_SIZE: Set this to 5MB or more.

• JOB_QUEUE_PROCESSES: Set this to three or higher. (Set it to at least one.) This is needed because the Oracle Ultra Search crawler is launched by scheduling a database job. If this is zero, then no database jobs are run. As a result, any attempts to launch the Oracle Ultra Search crawler fail. Also consider other requirements for job queue processes when you set this value.

Step 2: Create and Assign the Temporary Tablespace to the CTXSYS User

The starter database created by the Oracle Installer may create a temporary tablespace that is too small. Oracle Ultra Search uses the Oracle Text engine intensively. Therefore, a large temporary tablespace must be created for the Oracle Text system user CTXSYS. If you want greater read and write performance, create the tablespace on raw devices.

When you have created the temporary tablespace, assign it as the temporary tablespace for the CTXSYS user. For this, you must log on as the SYSTEM or SYS user. Assign the temporary tablespace to the CTXSYS user with the following statement:

ALTER USER CTXSYS TEMPORARY TABLESPACE new_temporary_tablespace;

Step 3: Create a Large Tablespace for Each Oracle Ultra Search Instance User

For each Oracle Ultra Search instance, you must create a tablespace large enough to contain all data obtained during the crawling and indexing processes. This amount is subject to the amount of data you intend to crawl and index. However, it is often not possible to know in advance how much data you intend to collect. Try to obtain an estimate of the cumulative size of all data you want to crawl.

If you cannot estimate the size, then try to allocate as much space as possible. If you run out of disk space, then to resume Oracle Ultra Search add more datafiles to the instance tablespace.

Here is an example of how to create a new tablespace:

CREATE TABLESPACE lmtbsb DATAFILE '/u02/oracle/data/lmtbsb01.dbf' SIZE 150M;

The amount of data to be stored in the tablespace can be very large. This can cause the Oracle server to progressively allocate many new extents when more storage space is needed. If the extent management clause specifies that each new extent is to be larger than the previous extent (that is, the PCTINCREASE setting is nonzero), then you could encounter the situation where the next extent that the Oracle server wants to allocate is larger than what is available. In such a situation, indexing halts until new extents can be added to the tablespace. Ensure that the STORAGE clause in your CREATE TABLESPACE statement specifies a value that can be allocated by the system.

To mitigate this problem, certain instance-specific tables have explicit storage parameter settings. The initial extent size, next extent size, and PCTINCREASE setting are defined for these tables. These tables are created when a new instance is created. The tables and their storage clause settings are as follows:

DR$WK$DOC_PATH_IDX$I 
                (initial extent size 5M, next extent size 50M, PCTINCEASE 1)
DR$WK$DOC_PATH_IDX$K 
                (initial extent size 5M, next extent size 50M, PCTINCEASE 1)

If you want greater read and write performance, then create the tablespace on raw devices.

Create a new large tablespace for each Oracle Ultra Search instance user.

Step 4: Create and Configure New Users for Oracle Ultra Search Instances

Oracle Ultra Search uses Oracle's fine grained access control feature to support multiple Oracle Ultra Search instances within one physical database. This is especially useful for large organizations or application service providers (ASPs) which want to host multiple disjoint search indexes within one installation of Oracle.

You must grant the WKUSER role to database users hosting new Oracle Ultra Search instances.

Enter the following statements to create and configure a new user. Run these statements as the WKSYS, SYSTEM, or SYS database user.

CREATE USER username 
              IDENTIFIED BY password DEFAULT TABLESPACE default_tbs 
              TEMPORARY TABLESPACE temporary_tbs QUOTA UNLIMITED 
              ON default_tbs;

where username = name of the Oracle Ultra Search instance owner

and password = password of the Oracle Ultra Search instance owner

and default_tbs = default tablespace for the Oracle Ultra Search instance created in Step 3

and temporary_tbs = temporary tablespace created in Step 2

GRANT WKUSER TO username;

After these steps are completed, WKSYS or an Oracle Ultra Search super-user can create an Oracle Ultra Search instance on this user schema.

If you want this user to have the general administrative privilege or the super-user privilege, then log on as an Oracle Ultra Search super-user or WKSYS and go to the Users page in the administration tool to grant the appropriate privilege.

Step 5: Alter the Index Preferences

This step is optional. An empty index is created when an Oracle Ultra Search instance is created. The existing index preferences, such as language-specific parameters, are defined in the $ORACLE_HOME/ultrasearch/admin/wk0pref.sql file.

You can modify these preferences so that all new Oracle Ultra Search instances use the modified preferences, or you can alter the index using your own preferences immediately after an instance is created. Alter the index using SQL.

Default Oracle Ultra Search Stoplist

During the installation process, a default stoplist is created for the Oracle Ultra Search product. Subsequently, when an Oracle Ultra Search instance is created, a copy of the default stoplist is created for the Oracle Ultra Search instance.

The default stoplist is created under the WKSYS schema. The default stoplist name is wk_stoplist. (This list is defined in the file $ORACLE_HOME/ultrasearch/admin/wk0pref.sql, which is run during installation). The stoplist is created only in English.

Modifying Instance Stoplists

You can modify the default stoplist by adding or removing stopwords. However, these modifications do not affect existing Oracle Ultra Search instances. They only affect Oracle Ultra Search instances that are created after the modifications are made.

Modifying instance stoplists should be done as a last resort. Use one of the following methods:

• Modify the default stoplist before creating the instance.

• Replace the instance stoplist immediately after creating the instance.

Replacing the instance stoplist immediately after creating the instance affects only that instance. You must first create a user-defined stoplist.

In both the cases, the result is that the Oracle Ultra Search instance stoplist is modified and defined before initial crawling. This means that all documents collected by the Oracle Ultra Search crawler are evaluated against the correct stoplist. It is important to modify the stoplist before initial crawling to avoid having to recrawl all documents.