Oracle® Database

3.1 Displaying and Modifying Resource Control Settings on Solaris 10

On Solaris 10, use the following procedure to display the current value specified for resource controls, and to change them if necessary:

1. To display the current values of the resource control, enter the following commands:

   $ id -p // to verify the project id
   uid=100(oracle) gid=100(dba) projid=1 (group.dba)
   $ prctl -n project.max-shm-memory -i project group.dba
   $ prctl -n project.max-sem-ids -i project group.dba
   

2. If you must change any of the current values, then:

   1. To modify the value of max-shm-memory to 6 GB:

      # prctl -n project.max-shm-memory -v 6gb -r -i project group.dba
      

   2. To modify the value of max-sem-ids to 256:

      # prctl -n project.max-sem-ids -v 256 -r -i project group.dba
      

Use the following procedure to modify the resource control project settings, so that they persist after a system restart:

1. By default, Oracle instances are run as the oracle user of the dba group. A project with the name group.dba is created to serve as the default project for the oracle user. Run the command id to verify the default project for the oracle user:

   # su - oracle
   $ id -p
   uid=100(oracle) gid=100(dba) projid=100(group.dba)
   $ exit
   

2. To set the maximum shared memory size to 2 GB, run the projmod command:

   # projmod -sK "project.max-shm-memory=(privileged,2G,deny)" group.dba
   

   Alternatively, add the resource control value project.max-shm-memory=(privileged,2147483648,deny) to the last field of the project entries for the Oracle project.

3. After these steps are complete, check the values for the /etc/project file using the following command:

   # cat /etc/project
   

   The output should be similar to the following:

   system:0::::
   user.root:1::::
   noproject:2::::
   default:3::::
   group.staff:10::::
   group.dba:100:Oracle default
   project:::project.max-shmmemory=(privileged,2147483648,deny)
       
   

4. To verify that the resource control is active, check process ownership, and run the commands id and prctl, as in the following example:

   # su - oracle
   $ id -p
   uid=100(oracle) gid=100(dba) projid=100(group.dba)
   $ prctl -n project.max-shm-memory -i process $$
   process: 5754: -bash
   NAME                    PRIVILEGE     VALUE     FLAG     ACTION    RECIPIENT
   project.max-shm-memory  privileged    2.00GB     -       deny 
   

   Note:

   The value for the maximum shared memory depends on the SGA requierments and should be set to a value greater than the SGA size.

   For additional information, refer to the Solaris Tunable Parameters Reference Manual.

3.2 Configure Shell Limits

Oracle recommends that you set shell limits and system configuration parameters as described in this section.

The ulimit settings determine process memory related resource limits. Verify that the shell limits displayed in the following table are set to the values shown:

To display the current value specified for these shell limits enter the following commands:

ulimit -t
ulimit -f
ulimit -d
ulimit -s
ulimit -n
ulimit -v

4.1 Latest Upgrade Information

For late-breaking updates and best practices about preupgrades, postupgrades, compatibility, and interoperability discussions refer to note 466181.1 on My Oracle Support (formerly OracleMetaLink) (https://support.oracle.com) that links to "10g Upgrade Companion" page.

4.2 Determining Size of Configured Swap Space

To determine the size of the configured swap space, enter the following command:

# /usr/sbin/swap -l

4.3 Oracle Storage Compatibility Program Obsolete

Oracle Storage Compatibility Program (OSCP) is no longer valid. Disregard the OSCP content in the section C.1, "General Configuration Guidelines for NAS Devices" of Appendix C, " Using NAS Devices" in Oracle Database Installation Guide for Solaris Operating System (SPARC 64-Bit).

4.4 Installing Enterprise Security Manager

To install Oracle Security Manager, install Oracle Client and then select the Administrator installation type.

4.5 Installing Oracle Database on a Computer That has an Automatic Storage Management Instance

Oracle Universal Installer displays an error message that reads 0. This message is displayed after you specify the database home and path in the Specify Home Details screen and click Next. Ignore the error message whenever it is displayed, and continue the installation.

4.6 extjob Executable Required Directory Permissions

To enable the extjob executable to locate required libraries, the $ORACLE_HOME/lib directory and all of its parent directories must have execute permissions for group and other.

4.7 Modifying a Virtual IP Address Node Application

When modifying the name, IP address, or netmask of an existing virtual IP address (VIP) resource, use the srvctl modify nodeapps command and include the existing interfaces for the VIP in the -A argument. For example:

srvctl modify nodeapps -n mynode1 -A 100.200.300.40/255.255.255.0/eth0

This issue is tracked with Oracle bug 4500688.

4.8 Running the Rootpre.sh Script on x86-64 with Sun Cluster

On x86-64 platforms running Solaris, if you install Sun Cluster in addition to Oracle Clusterware, then complete the following task:

1. Switch user to root:

   $ su - root
   

2. Complete one of the following steps, depending on the location of the installation

   • If the installation files are on a DVD, then enter a command similar to the following,where mountpoint is the disk mount point directory or the path of the database directory on the DVD:

     # mountpoint/clusterware/rootpre/rootpre.sh
     

   • If the installation files are on the hard disk, change directory to the directory /Disk1/rootpre and enter the following command:

     # ./rootpre.sh
     

3. Exit from the root account:

   # exit
   

   Note:

   Do not run the rootpre.sh script if you have a later release. If the Oracle Database software already installed on this system.

4. Repeat steps 1 through 3 on all nodes of the cluster.

4.9 Installing Oracle Database Client into an Existing Oracle Home

Oracle Database Client can be installed in the same Oracle Database home if both products are at the same release level. For example, you can install Oracle Database Client 10g Release 2 (10.2) into an existing Oracle Database 10g Release 2 (10.2) home. If you apply a patch set before installing the client, then you must apply the patch set again.

4.10 Database Installation Types

If you perform a Custom installation, then ensure that you install only the components covered by your license. You cannot install Standard Edition using Custom installation.

5.1 Materialized View and Partition Change Tracking Rewrite

When the query has single column in-lists, the materialized view has ranges, and partition change tracking rewrite is used, you might get stale results with Query Rewrite. Turn off the fresh partition containment rewrite to avoid this problem by using the following command:

SQL> alter session set "_query_rewrite_fpc" = false;

5.2 Host-Based Mirroring

The host-based mirroring is not supported with ASM. Note that there is no issue with resilvering for storage based mirroring.

Workaround: Use ASM redundancy.

This issue is tracked with Oracle bug 4466206.

5.3 Oracle ODBC Driver Limitations

Oracle ODBC driver for Solaris does not work if you use the REAL data type for a column and the application tries to retrieve data as native double or float data type through PL/SQL stored procedure.

Workaround: Instead of the REAL data type, use the NUMBER(s,p) data type while creating a table.

This issue is tracked with Oracle bug 4551566.

5.4 Importing a User Certificate Into Oracle Wallet Using Oracle Wallet Manager

Oracle Wallet Manager will fail to recognize the user certificate as valid if you capture the user certificate contents from the Oracle Certificate Authority site and save this as a flat file. This is because when saving as a flat file, all the lines of the user certificate are run together with no line breaks.

Workaround: Edit the user certificate. Insert a new line after -----BEGIN CERTIFICATE----- and a new line before -----END CERTIFICATE-----.

5.5 Removing Metrics for Wait Classes Removes Them Permanently

Do not remove the key values for the wait class metrics. Doing so removes them permanently and currently there is no easy way to recover them.

This issue is tracked with Oracle bug 4602952.

5.6 Preventing Loss of Quorum

If a storage area network (SAN) device is used to provide access to a shared storage and IO Multi-pathing (MPxIO) is enabled, then you must install the following patches on all the nodes of the cluster.

• 119374-13

• 119715-10

• 119375-13

• 119716-10

Without these patches, a node can lose access to the shared storage being accessed through the physical link that gets disconnected or fails.

5.7 Solaris the SMF Boot Archive Service

In Solaris 10 update 1, the boot archive service is introduced. The main purpose of this service is to reduce the boot time and simplify the boot loader. Due to an aggressive shutdown, the boot archive verification service may require manual intervention during boot.

During a patch installation or software upgrade, if you do not shutdown the computer properly, then kernel modules are reported to be out of sync. In such a case, start the computer to the fail-safe menu entry, follow the prompts to re-sync the archive, and then restart. If you do not want the manual intervention, you can disable the check by running the following commands:

# svcprop -p start/exec boot-archive:true
# svcadm refresh boot-archive

However, if the kernel modules are not reported to be out of sync, then you can clear the check by logging into the single user shell and running the following command:

# svcadm clear boot-archive

It is possible to disable the boot archive check on production computer to avoid manual intervention during aggressive restarts. To do so, the patch application and installation of driver or kernel software must be followed by a boot archive update. The kernel software includes the non-Solaris kernel modules.

The boot archive update can be achieved by applying a graceful restart or an explicit re-sync of the boot archive by running the following command:

bootadm update-archive

Oracle recommends that you remove the node from the cluster before patch installation and software installation or upgrade. When using Oracle Clusterware this can be done by completing the following steps:

1. Stop the Oracle Clusterware service by using the following command:

   $CRS_HOME/bin/crsctl stop crs
   

2. Disable the Oracle Clusterware service by using the following command:

   $CRS_HOME/bin/crsctl disable crs
   

3. Install or upgrade the required patch or software

4. Restart or re-sync the boot archive

5. Enable the Oracle Clusterware service by using the following command:

   $CRS_HOME/bin/crsctl enable crs
   

6. Start the Oracle Clusterware service by using the following command:

   $CRS_HOME/bin/crsctl start crs
   

Complete the following steps to remove the node from the cluster if you are using using Sun Cluster:

1. Restart the node in the maintenance mode

   See Also:

   Sun Cluster documentation for details on restarting a node in maintenance mode

2. Install or upgrade the required patch or software

3. Restart the node in the cluster mode

A fix for the boot archive service will be integrated in a future Solaris 10 update. After this fix is in place there will be no need to disable the check or to provide manual intervention during a restart. The boot archive check can be re-enabled by running the following commands:

# svccfg -s boot-archive setprop start/exec=/lib/svc/method/boot-archive
# svcadm refresh boot-archive

5.8 Increasing the CSS misscount Parameter

When the Solaris fiber channel port driver senses that a link is down, it gives two minutes timeout period before offlining the LUN path associated with the port. The purpose of this delay is to prevent a premature fail over in as a result of a transient link failure. If there are alternate active paths to the LUNs and SCSI reservations are not active, then within this timeout MPxIO automatically reconfigures to use the alternate path.

However, during this timeout period the port failure is not perceived by Oracle Clusterware because the path is not offlined. If the node is evicted during this timeout period, the Oracle Clusterware daemons fail to restart the node and services will not fail over to other nodes. In this case, the node will restart only until the link is up again. After the restart, all services belonging to the node will be up.

To avoid this problem the misscount parameter must be set to a value greater than 120 seconds. As a result, the eviction and restart process is not affected by the Solaris timeout period. The node is evicted, it restarts and its services fail over to other nodes as expected.

After Oracle Clusterware installation is complete, you can change the misscount parameter by completing the following steps:

1. On any node run the $CRS_HOME/bin/crsctl set css misscount 130 command.

2. On all nodes run the $CRS_HOME/bin/crsctl stop crs command.

3. On all nodes run the $CRS_HOME/bin/crsctl start crs command.

The default value for the misscount parameter is 27. Increasing the value of this parameter increases the time the node takes to fail over. Therefore, the service level of the cluster reduces for a longer period of time. It is up to the user to decide if a longer fail over time is acceptable. Solaris has a mechanism in place to allow applications to be notified immediately of a link down.

In this release, Oracle is not using this mechanism but work is in progress to make use of this mechanism. In a future Oracle release this workaround will no longer be needed.

5.9 Node Names do not Reflect in the Specify Cluster Nodes to Add to Installation Dialog Box

In the Specify Cluster Nodes to Add to Installation dialog box, an intermittent timing issue is observed. This dialog box prompts you to enter Public Node Names, Private Node Names, and Virtual Node Names for the new nodes to be added. Even after entering the node names, the configuration specified does not reflect in the install area. To avoid this issue, you need to press Tab after the last input to this dialog box.

5.10 Invalid Link to Monitor in Memory Access Mode Feature

Do not click the link to the Monitor in Memory Access Mode feature in the database screen. This feature is not available in Enterprise Manager Database Control 10.2.0.2 release. Clicking this link may stop an agent from responding.

This issue is tracked with Oracle bug 4866231.

5.11 Oracle Clusterware Private IP Addresses with Sun Cluster

If you are using a Sun Cluster, then do not enter the private interconnect in the /etc/hosts file, but instead use clusternodeX-priv to indicate the private interconnect for Oracle Clusterware and Oracle RAC.

This issue is tracked with bug 6238217.

5.12 Vendor Clusterware Restrictions for Node Names

If you use a vendor clusterware with Oracle Clusterware and Oracle Real Application Clusters, then you must use the node names and host names registered with that vendor clusterware you have installed.