Particular attention must be given to the operation of PRIMECLUSTER Wizard for Oracle.
Prerequisites and requirements
PRIMECLUSTER Wizard for Oracle has specific system requirements and component prerequisites that need to be understood prior to operation.
More than one Oracle Database product which version/release is different can be installed on the same node. However, the version and release should be same on an operating node and standby nodes. For the Oracle Database products supported, refer to "Readme" and "Handbook".
A user name, user ID, group name, and group ID should be identical on all the nodes to install Oracle Database.
If multiple Oracle homes are configured on one server, a different user name must be assigned for each Oracle home.
An Oracle database must be configured on the shared disk.
When the database using the multitenant architecture is created, it is necessary to allocate the database of CDB and PDB on the shared disk.
Setting Oracle database log output on the shared disk is not recommended. When a disk device fails, log might not be output.
Note of location of archived redo logs.
Located on local disks
If Oracle database recovery is required, the archived redo logs on both nodes should be put together.
Located on shared disks
Note that there is sufficient space to archive for archived redo log files. If there is insufficient space, and data updating processing through monitoring SQL hangs, an oracle resource might fail.
Afterwards, the service might stop finally on both operating node and standby node, because the space of shared disk is insufficient and the failover of userApplication fails on the standby node.
Note that there is sufficient space in the disk of the directory specified by AUDIT_FILE_DEST for the database using the multitenant architecture. The log file in the AUDIT_FILE_DEST directory increases by about 1KB at monitoring interval of the Oracle instance resource(Default: 30 seconds) when the PDB is monitored. Capacity and frequency of the log file are different according to the environment. Check the capacity of the AUDIT_FILE_DEST directory, and backup/delete it regularly.
The Oracle "TWO_TASK" environment variable cannot be used. For details, refer to "2.2.6 Oracle Database Creation and Setting".
The setting to connect to an Oracle database must not be described in login.sql of an Oracle user. For details, refer to "2.2.6 Oracle Database Creation and Setting".
PRIMECLUSTER Wizard for Oracle connects to Oracle instance with SYSDBA system privilege to start up or stop Oracle instance/database. In this case, local connection in operating system authentication is used. Therefore, the following requirements should be satisfied. (If not, startup and stop of Oracle instance/database might fail.)
Oracle user that is set for Oracle instance resource in "2.2.7.1 Oracle Resource Creation and Registration" should belong to OSDBA group.
SQLNET.AUTHENTICATION_SERVICES parameter should not define in sqlnet.ora file.
The other connections (ex. user own connection for maintenance) can use the password file authentication.
A command (e.g. script) that has interactive mode should not be described in Oracle user's profile.
(e.g. /etc/profile, .bash_profile, .cshrc, .profile)
It may cause the following failures.
Failures of userApplication startup, shutdown or monitoring of Oracle resources.
clgetoralog command and clorainfo command hang up and they do not return to command line.
Environment that Oracle user's home directory and profile are located on the shared disk is not supported. Oracle user's home directory and profile must be located on the local disk of each node.
Operating system authentication of Listener should be enabled. For details, refer to "2.2.6 Oracle Database Creation and Setting".
Oracle user should have access privileges for /tmp directory.
The programs of PRIMECLUSTER Wizard for Oracle working with Oracle user use /tmp directory. If Oracle user does not have access privileges for /tmp directory, the programs do not work correctly and startup or shutdown of Oracle database fails.
In order to use Oracle ASM, it is necessary to install Oracle Grid Infrastructure. When installing Oracle Grid Infrastructure, the composition which separates the installation user of Oracle Database and the installation user of Oracle Grid Infrastructure is supported. The composition which creating one user is not supported.
In order to use Oracle ASM, a different directory must be assigned respectively for an ORACLE_BASE of the installation user of Oracle Database and an ORACLE_BASE of the installation user of Oracle Grid Infrastructure.
Upgrade to local disk if the Oracle Database software is located on local disk.
Upgrade to shared disk if the Oracle Database software is located on shared disk.
Changing the placement of the Oracle Database software at the upgrade does not support.
Features and functions
Below are guidelines and precautions to be taken in using the features and functions provided by PRIMECLUSTER Wizard for Oracle.
The detector connects itself to the database using SYSTEM user access privileges to monitor Oracle instances. It also monitors Oracle database operation by creating a monitoring table on the SYSTEM user default table space. The size of the table is only a few bytes. The table data is periodically updated, so the REDO log is updated, and the archive log data is output.
Stopping an Oracle instance in the "immediate" mode fails while Oracle Online backup is running.
In this case, the Oracle instance will be stopped in "abort" mode.
The time required for Oracle instance startup and stop varies depending on the database size. The default startup timeout is 10 minutes. The default stop timeout is 3 minutes.
The RMS command "hvshut" has its own timeout. Even though hvshut times out, userApplication stop processing will be continued along with RMS stop.
If a database failure is detected during Oracle instance startup, the database will automatically be recovered (END BACKUP, MEDIA RECOVERY). If the backup mode of database is detected during the startup of Oracle instances, it will automatically be canceled. If media failure is detected, they will automatically be recovered.
MonitorOnly can be enabled in the following cases. For details, see “2.2.7.1 Oracle Resource Creation and Registration.
One resource of “Instance” type and one resource of “Listener” type exist, and StartPriority is set to “Same”. In this case, MonitorOnly only for the one of them can be enabled.
More than one resources of “Instance” type exist. In this case, MonitorOnly for the one of them cannot be enabled, but all of the others can be enabled.
More than one resources of “Listener” type exist. In this case, MonitorOnly for the one of them cannot be enabled, but all of the others can be enabled.
Configuration
Below are guidelines and precautions to be taken in configuring an environment that allows for PRIMECLUSTER Wizard for Oracle operations.
Be aware of the exit code and timeout duration when the following scripts are set:
PreOnlineScript, PostOnlineScript, PreOfflineScript, PostOfflineScript, FaultScript.
For details, refer to "2.3.1 Oracle resource Online/Offline Script".
Do not create multiple resources for the same an Oracle instance, an Oracle ASM instance or a Listener.
In "2.2.4.1 userApplication creation userApplication creation", if you use Oracle resources in combination with PRIMECLUSTER Wizard for NAS, refer to "PRIMECLUSTER Wizard for NAS Configuration and Administration Guide".
Operations
The following items are notes about the operation of PRIMECLUSTER Wizard for Oracle.
For information on how to change the Oracle SYSTEM user password while a cluster system is running, refer to "3.3 Oracle SYSTEM user Password Change ". Register the changed Oracle SYSTEM user password to PRIMECLUSTER Wizard for Oracle. When the passwords of Oracle user, Grid user or Oracle SYS user are changed, it is not necessary to register the passwords to PRIMELCUSTER Wizard for Oracle.
When UsePDB of Oracle instance resource is set to yes, the monitoring of PDB is executed. When fault of PDB is detected, the message is output to syslog. Refer to "Chapter 6 Message" for the content of the messages. The restart and the failover due to fault of PDBs is not executed.
Others
Users need to take the following special precaution when operating PRIMECLUSTER Wizard for Oracle:
If an Oracle instance is stopped in the ABORT mode, it will automatically be recovered at the next instance startup.
