8
Troubleshooting

This chapter provides information in the following area:

• Troubleshooting General Problems

Troubleshooting General Problems

This section contains information to assist with general problems and performance issues. The probable cause and a corrective action are provided for each problem.

Table 8-1  Troubleshooting General Problems

Problem	Probable Cause	Corrective Action
An out-of-memory exception is raised when running Oracle Files.	The maximum Java heap size is too low.	Increase the heap size by modifying the -Xmx setting for that node configuration.
Server is generally slow for read and write activity.	Probable cause #1: Server memory is overcommitted. The server is excessively swapping memory blocks to disk.	Run system monitoring tools, such as vmstat (UNIX) and look for excessive page swapping to verify the problem. Adjust the following parameters in your database's init.ora file: Reduce processes. Reduce open_cursors. Reduce db_block_buffers. Stop unneeded Java VMs or other unneeded processes. You may also need to add memory to your server or, if you are running a single-tier configuration, reconfigure your Oracle Files server into a two-tier configuration. For more information on adjusting the parameters in the init.ora file, see the Oracle Collaboration Suite Installation and Configuration Guide.
	Probable cause #2: CTXHX is using 100 percent of your CPU.	See Appendix A, "Oracle Text Reference".
Server is slow only on read or search activity.	Large volumes of data have been loaded but the CBO statistics weren't updated.	If the Oracle9i Cost-Based Optimizer is using stale statistics data, performance suffers. Run the analyze.sql script located in the $ORACLE_HOME/ifs/files/admin/sql directory to refresh the statistics.
Server is slow only on content-based search activity.	Oracle Text tablespaces are on the same disk as other database files.	Move the Oracle Text tablespaces to other disks. See the Oracle Collaboration Suite Installation and Configuration Guide and the Oracle9i Database Administrator's Guide for more information on moving tablespaces.
	Oracle Text indexes have become fragmented.	Regularly optimize the Oracle Text Oracle index IFS_TEXT. For more information, see "Maintaining the IFS_TEXT Index By Using the Oracle Text PL/SQL Packages" in Appendix A, "Oracle Text Reference".
Server is slow only on write activity.	Large amounts of documents are being loaded and the Redo logs are too small.	Add two or more 100 MB or larger Redo logs. See the Oracle9i Database Administrator's Guide for more information. In general, Redo logs should be switching every hour or less frequently. See the $ORACLE_HOME/rdbms/<sid>/bdump directory for the latest logs which indicate the frequency of Redo log switching.
	Large amounts of documents are being loaded and the Redo logs are on the same disk as the database files.	Place the Redo logs on a separate disk from the database files. See the Oracle9i Database Administrator's Guide and the Oracle9i Database Performance Tuning Guide and Reference for more information. For optimal performance, dedicate one or more disks (and, if possible, a disk controller) exclusively to the Redo logs, and optimize the disks for sequential write activity. For example, on Sun SPARC Solaris, you may choose raw partitions or UNIX file systems for the disks. If you choose UNIX file systems on Solaris 2.6 or above, use the "forcedirection" option when mounting the file systems. These options should only be used if the file systems are dedicated exclusively to the Redo logs.
Content queries through the Web and Windows return no rows.	Oracle Text indexing of the documents has not occurred.	See "Maintaining the IFS_TEXT Index By Using the Oracle Text PL/SQL Packages" in Appendix A, "Oracle Text Reference".
Problems with outbound FTP on UNIX.	You are using /usr/bin/ftp on UNIX and the default port number in /etc/services is a port other than 21, such as 2100.	Specify the port number explicitly, for example, "ftp ifs.us.oracle.com 21", where "21" is the port assigned.
Want to reset the Oracle Files system password without knowing what it is.	Have forgotten the Oracle Files system password.	You will need to contact Oracle Support or, if not, you must re-install the database. If you choose to re-install the database, you will need to make sure that you have copied any files stored in the Oracle Files database out to a local hard drive before starting the re-install process. The re-install will result in the loss of all information currently stored in the database.
The administrator has uploaded files and removed them and doesn't see the space retrieved in the tablespace.	The Initial Time of Day and Activation Period has been set incorrectly for the Content GarbageCollectionAgent.	Use Oracle Enterprise Manager Web site to look at the Initial Time of Day and Activation Period entries for the Content Garbage Collection Agent. Also check the Node.log and see if the Content Garbage Collection Agent is getting activated at the periodic intervals.
Cannot access Oracle Files from outside the firewall using SMB.	The required port is not open.	Windows clients can only access SMB if it is running on port 139, so you will need to open up port 139 to the external machine.
	You are using Windows 95/98/ME.	Windows 95/98/ME are not very good at connecting to SMB servers across the internet. You have to modify the local lmhosts file for these operating systems to connect to the server at all. Windows NT/2000/XP do not have this problem.
Accessing Oracle Files from outside the firewall using SMB is very slow.	SMB is inherently slow over the Internet.	All Windows SMB clients are very chatty, and will therefore amplify any latency that exists between the client and the server. The latency over the internet could be enough to make SMB appear much slower than normal.
Users fail to be provisioned, or newly provisioned users cannot be added to workspaces.	Required user attributes were not set in Oracle Internet Directory.	The following Oracle Internet Directory user attributes must be non-null for all users: sn, givenName, mail. In addition, all users must have a non-null username. The username is specified by the orclCommonNickname Attribute in the Subscriber's OracleContext. See the Oracle Internet Directory Administrator's Guide for more information on viewing the orclCommonNickname Attribute.
"503 Service Unavailable" error message.	The OC4J instance was not restarted after starting the Oracle Files Domain.	If the domain is not started, run ifsctl start. From $ORACLE_HOME/opmn/bin/, run opmnctl restartproc gid=OC4J_iFS_files. Check the HTTP Node log file to ensure that the HTTP node has started.
Cannot log in to FTP.	Protocol Access is not enabled for that user.	Log in to Oracle Files with the user account that cannot access FTP and go to the Protocol Access page, where you can set an Oracle Files-specific password. You can then log in to FTP using the Oracle Files-specific password.
When using NTFS, nothing gets written into the node log.	The Log Level of the node configuration is not set high enough.	NTFS will not write into the node log unless the Log Level is set to 6. See "Changing a Node Configuration" for information about how to set the Log Level.