7 Using Oracle Clusterware to Manage Active Standby Pairs

Active standby configurations

Use Oracle Clusterware to manage only these configurations:

• Active standby pair with or without read-only subscribers

• Active standby pair (with or without read-only subscribers) with AWT cache groups, read-only cache groups and global cache groups

Required privileges

See "ttCWAdmin" in Oracle TimesTen In-Memory Database Reference for information about the privileges required to execute ttCWAdmin commands.

Hardware and software requirements

Oracle Clusterware release 11.2.0.2.x is supported with TimesTen active standby pair replication, beginning with release 11.2.0.2.0. See Oracle Clusterware Administration and Deployment Guide for network and storage requirements and information about Oracle Clusterware configuration files.

Oracle Clusterware and TimesTen should be installed in the same location on all nodes.

The TimesTen instance administrator must belong to the same UNIX or Linux primary group as the Oracle Clusterware installation owner.

Note that the /tmp directory contains essential TimesTen Oracle Clusterware directories. Their names have the prefix crsTT. Do not delete them.

All machines should use Network Time Protocol (NTP) or a similar system so that clocks on the machines remain within 250 milliseconds of each other.

Restricted commands and SQL statements

When you use Oracle Clusterware with TimesTen, you cannot use these commands and SQL statements:

• CREATE ACTIVE STANDBY PAIR, ALTER ACTIVE STANDBY PAIR and DROP ACTIVE STANDBY PAIR SQL statements

• The -repStart and -repStop options of the ttAdmin utility

• The -cacheStart and -cacheStop options of the ttAdmin utility after the active standby pair has been created

• The -duplicate option of the ttRepAdmin utility

• The ttRepStart and ttRepStop built-in procedures

• Built-in procedures for managing a cache grid when the active standby pair in a cluster is a member of a grid

In addition, do not call ttDaemonAdmin -stop before calling ttCWAdmin -shutdown.

The TimesTen integration with Oracle Clusterware accomplishes these operations with the ttCWAdmin utility and the attributes in the cluster.oracle.ini file.

For more information about the built-ins and utilities, see Oracle TimesTen In-Memory Database Reference. For more information about the SQL statements, see Oracle TimesTen In-Memory Database SQL Reference.

Configuring basic availability

This example shows an active standby pair with no subscribers. The hosts for the active database and the standby database are host1 and host2. The list of hosts is delimited by commas. You can include spaces for readability if desired.

The ttCWAdmin utility is used to administer TimesTen active standby pairs in a cluster that is managed by Oracle Clusterware.

[basicDSN]
MasterHosts=host1,host2

This is an example of a cluster.oracle.ini file for an active standby pair with one subscriber on host3:

[basicSubscriberDSN]
MasterHosts=host1,host2
SubscriberHosts=host3

Configuring advanced availability

In this example, the hosts for the active database and the standby database are host1 and host2. host3 and host4 are extra nodes that can be used for failover. There are no subscriber nodes. MasterVIP specifies the virtual IP addresses defined for the master databases. VIPInterface is the name of the public network adaptor. VIPNetMask defines the netmask of the virtual IP addresses.

[advancedDSN]
MasterHosts=host1,host2,host3,host4
MasterVIP=192.168.1.1, 192.168.1.2
VIPInterface=eth0
VIPNetMask=255.255.255.0

This example has one subscriber on host4. There is one extra node that can be used for failing over the master databases and one extra node that can be used for the subscriber database. MasterVIP and SubscriberVIP specify the virtual IP addresses defined for the master and subscriber databases. VIPInterface is the name of the public network adaptor. VIPNetMask defines the netmask of the virtual IP addresses.

[advancedSubscriberDSN]
MasterHosts=host1,host2,host3
SubscriberHosts=host4,host5
MasterVIP=192.168.1.1, 192.168.1.2
SubscriberVIP=192.168.1.3
VIPInterface=eth0
VIPNetMask=255.255.255.0

Ensure that the extra nodes:

• Have TimesTen installed

• Have the direct-linked application installed if this is part of the configuration. See "Implementing application failover".

Including cache groups in the active standby pair

If the active standby pair replicates one or more AWT or read-only cache groups, set the CacheConnect attribute to y.

This example specifies an active standby pair with one subscriber in an advanced availability configuration. The active standby pair replicates one or more cache groups.

[advancedCacheDSN]
MasterHosts=host1,host2,host3
SubscriberHosts=host4, host5
MasterVIP=192.168.1.1, 192.168.1.2
SubscriberVIP=192.168.1.3
VIPInterface=eth0
VIPNetMask=255.255.255.0
CacheConnect=y

Including the active standby pair in a cache grid

If the active standby pair is a member of a cache grid, assign port numbers for the active and standby databases by setting the GridPort attribute.

This example specifies an active standby pair with no subscribers in an advanced availability configuration. The active standby pair is a member of a cache grid.

[advancedGridDSN]
MasterHosts=host1,host2,host3
MasterVIP=192.168.1.1, 192.168.1.2
VIPInterface=eth0
VIPNetMask=255.255.255.0
CacheConnect=y
GridPort=16101, 16102

For more information about using Oracle Clusterware with a cache grid, see "Using Oracle Clusterware with a TimesTen cache grid".

Implementing application failover

TimesTen integration with Oracle Clusterware can facilitate the failover of an application that is linked to any of the databases in the active standby pair. Both direct-linked and client/server applications that are on the same machine as Oracle Clusterware and TimesTen can be managed.

The required attributes in the cluster.oracle.ini file for failing over an application are:

• AppName - Name of the application to be managed by Oracle Clusterware

• AppStartCmd - Command line for starting the application

• AppStopCmd - Command line for stopping the application

• AppCheckCmd - Command line for executing an application that checks the status of the application specified by AppName

• AppType - Determines the database to which the application is linked. The possible values are Active, Standby, Subscriber (all) and Subscriber[index].

Optionally, you can also set AppFailureThreshold, DatabaseFailoverDelay, and AppScriptTimeout. These attributes have default values.

The TimesTen application monitor process uses the user-supplied script or program specified by AppCheckCmd to monitor the application. The script that checks the status of the application must be written to return 0 for success and a nonzero number for failure. When Oracle Clusterware detects a nonzero value, it takes action to recover the failed application.

This example shows advanced availability configured for an active standby pair with with no subscribers. The reader application is an application that queries the data in the standby database. AppStartCmd, AppStopCmd and AppCheckCmd can include arguments such as start, stop and check commands. On UNIX and Linux, do not use quotes in the values for AppStartCmd, AppStopCmd and AppCheckCmd.

[appDSN]
MasterHosts=host1,host2,host3,host4
MasterVIP=192.168.1.1, 192.168.1.2
VIPInterface=eth0
VIPNetMask=255.255.255.0
AppName=reader
AppType=Standby
AppStartCmd=/mycluster/reader/app_start.sh start
AppStopCmd=/mycluster/reader/app_stop.sh stop
AppCheckCmd=/mycluster/reader/app_check.sh check

AppStartCmd, AppStopCmd and AppCheckCmd can include arguments. For example, consider a valid cluster.oracle.ini file on Windows. In this example the application is directly linked to the active database. The script for starting, stopping, and checking the application takes arguments for the DSN and the action to take (-start, -stop and -check).

Note the double quotes for the specified paths in AppStartCmd, AppStopCmd and AppCheckCmd. The quotes are needed because there are spaces in the path. Enclose only the path in quotes. Do not enclose the DSN or the action in quotes.

[appWinDSN]
MasterHosts=host1,host2,host3,host4
MasterVIP=192.168.1.1, 192.168.1.2
VIPInterface=Local Area Connection
VIPNetMask=255.255.255.0
AppName=UpdateApp
AppType=Active
AppStartCmd="C:\Program Files\UserApps\UpdateApp.exe" -dsn myDSN -start
AppStopCmd= "C:\Program Files\UserApps\UpdateApp.exe" -dsn myDSN -stop
AppCheckCmd="C:\Program Files\UserApps\UpdateApp.exe" -dsn myDSN -check

You can configure failover for more than one application. Use AppName to name the application and provide values for AppType, AppStartCmd, AppStopCmd and AppCheckCmd immediately following the AppName attribute. You can include blank lines for readability. For example:

[app2DSN]
MasterHosts=host1,host2,host3,host4
MasterVIP=192.168.1.1, 192.168.1.2
VIPInterface=eth0
VIPNetMask=255.255.255.0

AppName=reader
AppType=Standby
AppStartCmd=/mycluster/reader/app_start.sh
AppStopCmd=/mycluster/reader/app_stop.sh
AppCheckCmd=/mycluster/reader/app_check.sh

AppName=update
AppType=Active
AppStartCmd=/mycluster/update/app2_start.sh
AppStopCmd=/mycluster/update/app2_stop.sh
AppCheckCmd=/mycluster/update/app2_check.sh

The application is considered available if it has been running for 15 times the value of AppScriptTimeout attribute. The default value of AppScriptTimeout is 60 seconds, so the application's "uptime threshold" is 15 minutes by default. If the application fails after running for more than 15 minutes, it will be restarted on the same host. If the application fails within 15 minutes of being started, the failure is considered a failure to start properly, and the application will be restarted on another host. If you want to modify the application's uptime threshold after the application has started, use the crs_register -update command. See Oracle Clusterware Administration and Deployment Guide for information about the crs_register -update command.

Recovering from permanent failure of both master nodes

If both master nodes fail and then come back up, Oracle Clusterware can automatically recover the master databases. Automatic recovery of temporary dual failure requires:

• RETURN TWOSAFE is not specified for the active standby pair.

• AutoRecover is set to y.

• RepBackupDir specifies a directory on shared storage.

• RepBackupPeriod is set to a value greater than 0.

If both master nodes fail permanently, Oracle Clusterware can automatically recover the master databases to two new nodes if:

• Advanced availability is configured (virtual IP addresses and at least four hosts).

• The active standby pair does not replicate cache groups.

• A cache grid is not configured.

• RETURN TWOSAFE is not specified.

• AutoRecover is set to y.

• RepBackupDir specifies a directory on shared storage.

• RepBackupPeriod must be set to a value greater than 0.

TimesTen first performs a full backup of the active database and then performs incremental backups. You can specify the optional attribute RepfullbackupCycle to manage when TimesTen performs subsequent full backup. By default, TimesTen performs a full backup after every five incremental backups.

If RepBackupDir and RepBackupPeriod are configured for backups, TimesTen performs backups for any master database that becomes active. It does not delete backups that were performed for a database that used to be active and has become the standby unless the database becomes active again. Ensure that the shared storage has enough space for two complete database backups. ttCWAdmin -restore automatically chooses the correct backup files.

Incremental backups increase the amount of log records in the transaction log files. Ensure that the values of RepBackupPeriod and RepFullBackupCycle are small enough to prevent a large amount of log records in the transaction log file. For more information about the interaction between backups and log records, see the description of the ttBackup utility in Oracle TimesTen In-Memory Database Reference.

This example shows attribute settings for automatic recovery.

[autorecoveryDSN]
MasterHosts=host1,host2,host3,host4
MasterVIP=192.168.1.1, 192.168.1.2
VIPInterface=eth0
VIPNetMask=255.255.255.0
AutoRecover=y
RepBackupDir=/shared_drive/dsbackup
RepBackupPeriod=3600

If you have cache groups in the active standby pair or prefer to recover manually from failure of both master hosts, ensure that AutoRecover is set to n (the default). Manual recovery requires:

• RepBackupDir specifies a directory on shared storage

• RepBackupPeriod must be set to a value greater than 0

This example shows attribute settings for manual recovery. The default value for AutoRecover is n, so it is not included in the file.

[manrecoveryDSN]
MasterHosts=host1,host2,host3
MasterVIP=192.168.1.1, 192.168.1.2
VIPInterface=eth0
VIPNetMask=255.255.255.0
RepBackupDir=/shared_drive/dsbackup
RepBackupPeriod=3600

Using the RepDDL attribute

The RepDDL attribute represents the SQL statement that creates the active standby pair. The RepDDL attribute is optional. You can use it to exclude tables, cache groups and sequences from the active standby pair.

If you include RepDDL in the cluster.oracle.ini file, do not specify ReturnServiceAttribute, MasterStoreAttribute or SubscriberStoreAttribute in the cluster.oracle.ini file. Include those replication settings in the RepDDL attribute.

When you specify a value for RepDDL, use the <DSN> macro for the database file name prefix. Use the <MASTERHOST[1]> and <MASTERHOST[2]> macros to specify the master host names. TimesTen substitutes the correct values from the MasterHosts or MasterVIP attributes, depending on whether your configuration uses virtual IP addresses. Similarly, use the <SUBSCRIBERHOST[n]> macro to specify subscriber host names, where n is a number from 1 to the total number of SubscriberHosts attribute values or 1 to the total number of SubscriberVIP attribute values if virtual IP addresses are used.

Use the RepDDL attribute to exclude tables, cache groups and sequences from the active standby pair:

[excludeDSN]
MasterHosts=host1,host2,host3,host4
SubscriberHosts=host5,host6
MasterVIP=192.168.1.1, 192.168.1.2
SubscriberVIP=192.168.1.3
VIPInterface=eth0
VIPNetMask=255.255.255.0
RepDDL=CREATE ACTIVE STANDBY PAIR \
<DSN> ON <MASTERHOST[1]>, <DSN> ON <MASTERHOST[2]>
SUBSCRIBER <DSN> ON <SUBSCRIBERHOST[1]>\
EXCLUDE TABLE pat.salaries, \
EXCLUDE CACHE GROUP terry.salupdate, \
EXCLUDE SEQUENCE ttuser.empcount

The replication agent transmitter obtains route information as follows, in order of priority:

1. From the ROUTE clause in the RepDDL setting, if a ROUTE clause is specified. Do not specify a ROUTE clause if you are configuring advanced availability.

2. From Oracle Clusterware, which provides the private host names and public host names of the local and remote hosts as well as the remote daemon port number. The private host name is preferred over the public host name. The replication agent transmitter cannot connect to the IPC socket, it attempts to connect to the remote daemon, using information that Oracle Clusterware maintains about the replication scheme.

3. From the active and standby hosts. If they fail, then the replication agent chooses the connection method based on host name.

This is an example of specifying the ROUTE clause in RepDDL:

[routeDSN]
MasterHosts=host1,host2,host3,host4
RepDDL=CREATE ACTIVE STANDBY PAIR \
<DSN> ON <MASTERHOST[1]>, <DSN> ON <MASTERHOST[2]>\
ROUTE MASTER <DSN> ON <MASTERHOST[1]>  SUBSCRIBER <DSN> ON <MASTERHOST[2]>\
MASTERIP "192.168.1.2" PRIORITY 1\
SUBSCRIBERIP "192.168.1.3" PRIORITY 1\ 
MASTERIP "10.0.0.1" PRIORITY 2\
SUBSCRIBERIP "10.0.0.2" PRIORITY 2\
MASTERIP "140.87.11.203" PRIORITY 3\
SUBSCRIBERIP "140.87.11.204" PRIORITY 3\
ROUTE MASTER <DSN> ON <MASTERHOST[2]>  SUBSCRIBER <DSN> ON <MASTERHOST[1]>\
MASTERIP "192.168.1.3" PRIORITY 1\
SUBSCRIBERIP "192.168.1.2" PRIORITY 1\ 
MASTERIP "10.0.0.2" PRIORITY 2\
SUBSCRIBERIP "10.0.0.1" PRIORITY 2\
MASTERIP "140.87.11.204" PRIORITY 3\
SUBSCRIBERIP "140.87.11.203" PRIORITY 3\

Install Oracle Clusterware

Install Oracle Clusterware. By default, the installation occurs on all hosts concurrently. See Oracle Clusterware installation documentation for your platform.

Oracle Clusterware starts automatically after successful installation.

Install TimesTen on each host

Install TimesTen in the same location on each host in the cluster, including extra hosts. The instance name must be the same on each host. The user name of the instance administrator must be the same on all hosts. The TimesTen instance administrator must belong to the same UNIX or Linux primary group as the Oracle Clusterware installation owner.

On Linux and UNIX, the installer prompts you for values for:

• The TCP/IP port number associated with the TimesTen cluster agent. The port number can be different on each host. If you do not provide a port number, TimesTen uses the default TimesTen port.

• The Oracle Clusterware location. The location must be the same on each host.

• The hosts included in the cluster, including spare hosts, with host names separated by commas. This list must be the same on each host.

The installer uses these values to create the ttcrsagent.options file on Linux and UNIX platforms. See "TimesTen Installation" in Oracle TimesTen In-Memory Database Installation Guide for details. You can also use ttmodinstall -crs to create the file after installation. On Linux, you can use the -record and -batch options for setup.sh to perform identical installations on additional hosts if desired.

On Windows, execute ttmodinstall -crs on each node after installation to create the ttcrsagent.options file.

For more information about ttmodinstall, see "ttmodinstall" in Oracle TimesTen In-Memory Database Reference.

Register the TimesTen cluster information

TimesTen cluster information is stored in the Oracle Cluster Registry (OCR). As the root user on UNIX or Linux platforms, or as the instance administrator on Windows, enter this command:

ttCWAdmin -ocrConfig

As long as Oracle Clusterware and TimesTen are installed on the hosts, this step never needs to be repeated.

Start the TimesTen cluster agent

Start the TimesTen cluster agent by executing the ttCWAdmin -init command on one of the hosts. For example:

ttCWAdmin -init

This command starts the TimesTen cluster agent (ttCRSAgent) and the TimesTen daemon monitor (ttCRSDaemon). There is one TimesTen cluster agent and one TimesTen daemon monitor for the TimesTen installation. When the TimesTen cluster agent has started, Oracle Clusterware begins monitoring the TimesTen daemon and will restart it if it fails.

Create and populate a TimesTen database on one host

Create a database on the host where you intend the active database to reside. The DSN must be the same as the database file name.

Create schema objects such as tables, AWT cache groups and read-only cache groups. Do not load the cache groups.

Create sys.odbc.ini files on other hosts

On all hosts that will be in the cluster, create sys.odbc.ini files. The DataStore attribute and the Data Source Name must be the same as the entry name for the cluster.oracle.ini file. See "The cluster.oracle.ini file" for information about the contents of the sys.odbc.ini files.

Create a cluster.oracle.ini file

Create a cluster.oracle.ini file on the host where you intend the active database to reside. See "The cluster.oracle.ini file" for details about its contents and location.

Create the virtual IP addresses (optional)

For advanced availability, execute the ttCWAdmin -createVIPs command on any host in the cluster. On UNIX, you must execute this command as the root user. For example:

ttCWAdmin -createVIPs -dsn myDSN

Create an active standby pair replication scheme

Create an active standby pair replication scheme by executing the ttCWAdmin -create command on any host. For example:

ttCWAdmin -create -dsn myDSN

The command prompts you to choose the host on which the active database will reside.

This command prompts for an encryption pass phrase that the user will not need again. The command also prompts for the user ID and password for an internal user with the ADMIN privilege if it does not find this information in the sys.odbc.ini file. This internal user will be used to create the active standby pair.

If CacheConnect is enabled, the command prompts for the user password for the Oracle database. The Oracle password is used to set the autorefresh states for cache groups.

Start the active standby pair

Start the active standby pair replication scheme by executing the ttCWAdmin -start command on any host. For example:

ttCWAdmin -start -dsn myDSN

This command starts the following processes for the active standby pair:

• ttCRSMaster

• ttCRSActiveService

• ttCRSStandbyService

• Monitor for application AppName

Load cache groups

If the active standby pair includes cache groups, use the LOAD CACHE GROUP statement to load the cache group tables from the Oracle tables.

Including more than one active standby pair in a cluster

If you want to use Oracle Clusterware to manage more than one active standby pair in a cluster, include additional configurations in the cluster.oracle.ini file. For example, this cluster.oracle.ini file contains configuration information for two active standby pair replication schemes:

[advancedSubscriberDSN]
MasterHosts=host1,host2,host3
SubscriberHosts=host4, host5
MasterVIP=192.168.1.1, 192.168.1.2
SubscriberVIP=192.168.1.3
VIPInterface=eth0
VIPNetMask=255.255.255.0

[advSub2DSN]
MasterHosts=host1,host2,host3
SubscriberHosts=host4, host5
MasterVIP=192.168.1.4, 192.168.1.5
SubscriberVIP=192.168.1.6
VIPInterface=eth0
VIPNetMask=255.255.255.0

Perform these tasks for additional replication schemes:

1. Create and populate the databases.

2. Create the virtual IP addresses. Use the ttCWAdmin -createVIPs command.

3. Create the active standby pair replication scheme. Use the ttCWAdmin -create command.

4. Start the active standby pair. Use the ttCWAdmin -start command.

Configuring an Oracle database as a disaster recovery subscriber

You can create an active standby pair on the primary site with an Oracle database as a remote disaster recovery subscriber. See "Using a disaster recovery subscriber in an active standby pair". Oracle Clusterware manages the active standby pair but does not manage the disaster recovery subscriber. The user must perform a switchover if the primary site fails.

To use Oracle Clusterware to manage an active standby pair that has a remote disaster recovery subscriber, perform these tasks:

1. Use the RepDDL or RemoteSubscriberHosts Clusterware attribute to provide information about the remote disaster recovery subscriber.For example:

   [advancedDRsubDSN]
   MasterHosts=host1,host2,host3
   SubscriberHosts=host4, host5
   RemoteSubscriberHosts=host6
   MasterVIP=192.168.1.1, 192.168.1.2
   SubscriberVIP=192.168.1.3
   VIPInterface=eth0
   VIPNetMask=255.255.255.0
   CacheConnect=y
   

2. Use ttCWAdmin -create to create the active standby pair replication scheme on the primary site. This does not create the disaster recovery subscriber.

3. Use ttCWAdmin -start to start the active standby pair replication scheme.

4. Load the cache groups that are replicated by the active standby pair.

5. Set up the disaster recovery subscriber using the procedure in "Rolling out a disaster recovery subscriber".

Configuring a read-only subscriber that is not managed by Oracle Clusterware

You can include a read-only TimesTen subscriber database that is not managed by Oracle Clusterware. Perform these tasks:

1. Include the RemoteSubscriberHosts Clusterware attribute in the cluster.oracle.ini file.For example:

   [advancedROsubDSN]
   MasterHosts=host1,host2,host3
   RemoteSubscriberHosts=host6
   MasterVIP=192.168.1.1, 192.168.1.2
   SubscriberVIP=192.168.1.3
   VIPInterface=eth0
   VIPNetMask=255.255.255.0
   

2. Use ttCWAdmin -create to create the active standby pair replication scheme on the primary site.

3. Use ttCWAdmin -start to start the active standby pair replication scheme. This does not create the read-only subscriber.

4. Use the ttRepStateGet procedure to verify that the state of the standby database is STANDBY.

5. On the subscriber host, use ttRepAdmin -duplicate option to duplicate the standby database to the read-only subscriber. See "Duplicating a database".

6. Start the replication agent on the subscriber host.

To add a read-only subscriber to an existing configuration, see "Adding a read-only subscriber not managed by Oracle Clusterware".

To rebuild a read-only subscriber, see "Rebuilding a read-only subscriber not managed by Oracle Clusterware".

Creating and initializing a cluster of cache grid members

See "Install TimesTen on each host" for installation requirements. In addition, each grid member must have a DSN that is unique within the cache grid.

Perform the tasks described in "Creating and initializing a cluster" for each grid member. Include the GridPort Clusterware attribute in the cluster.oracle.ini file as described in "Including the active standby pair in a cache grid". Ensure that the specified port numbers are not in use.

The ttCWAdmin -start command automatically attaches a grid member to the cache grid attach. The ttCWAdmin -stop command automatically detaches a grid member from the cache grid.

Failure and recovery of both nodes of an active standby pair grid member

Oracle Clusterware manages failover for active standby pair grid members. If both nodes of an active standby pair grid member fail, then the grid member fails. Oracle Clusterware evicts the failed grid member from the grid automatically.

If the active standby pair grid member is in an asynchronous replication scheme, the grid member is recovered automatically and reattached to the grid. If the active standby pair grid member is in a replication scheme with RETURN TWOSAFE configured, perform these steps to recover the grid member and reattach it to the grid:

1. Stop the replication agent and the cache agent and disconnect the application from both databases. This step detaches the grid member from the grid.

   ttCWAdmin -stop advancedGridDSN
   

2. Drop the active standby pair.

   ttCWAdmin -drop advancedGridDSN
   

3. Create the active standby pair replication scheme.

   ttCWAdmin -create advancedGridDSN
   

4. Start the active standby pair replication scheme. This step attaches the grid member to the grid.

   ttCWAdmin -start advancedGridDSN
   

Making schema changes to active standby pairs in a grid

You can add, drop or change a cache group while the active database is attached to the grid.

Use the ttCWAdmin -beginAlterSchema command to make these schema changes. This command stops replication but allows the active database to remain attached to the grid. The ttCWAdmin -endAlterSchema command duplicates the changes to the standby database, registers the altered replication scheme and starts replication.

To add a table and include it in the active standby pair, see "Making DDL changes in an active standby pair". See the same section for information about dropping a replicated table.

How TimesTen performs recovery when Oracle Clusterware is configured

The TimesTen database monitor (ttCRSmaster process) performs recovery. It attempts to connect to the failed database without using the forceconnect option. If the connection fails with error 994 (Data store connection terminated), the database monitor tries to connect 10 times. If the connection fails with error 707 (Attempt to connect to a data store that has been manually unloaded from RAM), the database monitor changes the RAM policy and tries to connect again. If the database monitor cannot connect, it returns connection failure.

If the database monitor can connect to the database, then it performs these tasks:

• It queries the CHECKSUM column in the TTREP.REPLICATIONS replication table.

• If the value in the CHECKSUM column matches the checksum stored in the Oracle Cluster Registry, then the database monitor verifies the role of the database. If the role is 'ACTIVE', then recovery is complete.

  If the role is not 'ACTIVE', then the database monitor queries the replication Commit Ticket Number (CTN) in the local database and the CTN in the active database to find out whether there are transactions that have not been replicated. If all transactions have been replicated, then recovery is complete.

• If the checksum does not match or if some transactions have not been replicated, then the database monitor performs a duplicate operation from the remote database to re-create the local database.

If the database monitor fails to connect with the database because of error 8110 or 8111 (master catchup required or in progress), then it uses the forceconnect=1 option to connect and starts master catchup. Recovery is complete when master catchup has been completed. If master catchup fails with error 8112 (Operation not permitted), then the database monitor performs a duplicate operation from the remote database. For more information about master catchup, see"Automatic catch-up of a failed master database".

If the connection fails because of other errors, then the database monitor tries to perform a duplicate operation from the remote database.

The duplicate operation verifies that:

• The remote database is available.

• The replication agent is running.

• The remote database has the correct role. The role must be 'ACTIVE' when the duplicate operation is attempted for creation of a standby database. The role must be 'STANDBY' or 'ACTIVE' when the duplicate operation is attempted for creation of a read-only subscriber.

When the conditions for the duplicate operation are satisfied, the existing failed database is destroyed and the duplicate operation starts.

When an active database or its host fails

If there is a failure on the node where the active database resides, Oracle Clusterware automatically changes the state of the standby database to 'ACTIVE'. If application failover is configured, then the application begins updating the new active database.

Figure 7-2 shows that the state of the standby database has changed to 'ACTIVE' and that the application is updating the new active database.

Figure 7-2 Standby database becomes active

Description of "Figure 7-2 Standby database becomes active"

Oracle Clusterware tries to restart the database or host where the failure occurred. If it is successful, then that database becomes the standby database.

Figure 7-3 shows a cluster where the former active node becomes the standby node.

Figure 7-3 Standby database starts on former active host

Description of "Figure 7-3 Standby database starts on former active host"

If the failure of the former active node is permanent and advanced availability is configured, Oracle Clusterware starts a standby database on one of the extra nodes.

Figure 7-4 shows a cluster in which the standby database is started on one of the extra nodes.

Figure 7-4 Standby database starts on extra host

Description of "Figure 7-4 Standby database starts on extra host"

When a standby database or its host fails

If there is a failure on the standby node, Oracle Clusterware first tries to restart the database or host. If it cannot restart the standby database on the same host and advanced availability is configured, Oracle Clusterware starts the standby database on an extra node.

Figure 7-5 shows a cluster in which the standby database is started on one of the extra nodes.

Figure 7-5 Standby database on new host

Description of "Figure 7-5 Standby database on new host"

When read-only subscribers or their hosts fail

If there is a failure on a subscriber node, Oracle Clusterware first tries to restart the database or host. If it cannot restart the database on the same host and advanced availability is configured, Oracle Clusterware starts the subscriber database on an extra node.

When failures occur on both master nodes

This section includes these topics:

• Automatic recovery

• Manual recovery for advanced availability

• Manual recovery for basic availability

• Manual recovery to the same master nodes when databases are corrupt

• Manual recovery when RETURN TWOSAFE is configured

[basicTwosafeDSN]
MasterHosts=host1,host2
ReturnServiceAttribute=RETURN TWOSAFE
RepBackupDir=/shared_drive/dsbackup
RepBackupPeriod=3600

When more than two master hosts fail

Approach a failure of more than two master hosts as a more extreme case of dual host failure. Use these guidelines:

• Address the root cause of the failure if it is something like a power outage or network failure.

• Identify or obtain at least two healthy hosts for the active and standby databases.

• Update the MasterHosts and SubscriberHosts entries in the cluster.oracle.ini file.

• See "Manual recovery for advanced availability" and "Manual recovery for basic availability" for guidelines on subsequent actions to take.

Changing the schema

To include or exclude a table, see "Making DDL changes in an active standby pair".

To include or exclude a cache group, see "Making schema changes to active standby pairs in a grid".

To create PL/SQL procedures, sequences, materialized views and indexes on tables with data, perform these tasks:

1. Enable the addition of the object to the active standby pair.

   ttCWAdmin -beginAlterSchema advancedDSN
   

2. Create the object.

3. If the object is a sequence and you want to include it in the active standby pair replication scheme, alter the active standby pair.

   ALTER ACTIVE STANDBY PAIR INCLUDE samplesequence;
   

4. Duplicate the change to the standby database.

   ttCWQAdmin -endAlterSchema advancedDSN
   

To add or drop a subscriber database or alter database attributes, perform the following tasks:

1. Stop the replication agents on the databases in the active standby pair. These commands use the advancedCacheDSN as an example.

   ttCWAdmin -stop -dsn advancedCacheDSN
   

2. Drop the active standby pair.

   ttCWAdmin -drop  -dsn advancedCacheDSN
   

3. Modify the schema as desired.

4. Re-create the active standby pair replication scheme.

   ttCWAdmin -create -dsn advancedCacheDSN
   

5. Start the active standby pair replication scheme.

   ttCWAdmin -start -dsn advancedCacheDSN
   

Performing a rolling upgrade of Oracle Clusterware software

See Oracle Clusterware Administration and Deployment Guide.

Upgrading TimesTen

If you are upgrading from a TimesTen release 11.2.1.6.6 or later to a later release of TimesTen 11.2.1, see Chapter 9, "Upgrading TimesTen When Using Oracle Clusterware". Otherwise, perform the tasks described in this section:

1. Stop the replication agents on the databases in the active standby pair.

   ttCWAdmin -stop -dsn advancedDSN
   

2. Drop the active standby pair.

   ttCWAdmin -drop -dsn advancedDSN
   

3. Stop the TimesTen cluster agent on the host. This removes the host from the cluster and stops the TimesTen daemon.

   ttCWAdmin -shutdown -hosts localhost
   

4. Upgrade TimesTen on the desired hosts. Every node in the cluster must have TimesTen from the same major release. If you are upgrading between major releases, use the ttMigrate utility. See "Database Upgrades" in Oracle TimesTen In-Memory Database Installation Guide.

5. Start the TimesTen cluster agent. This includes the host in the cluster and starts the TimesTen daemon.

   ttCWAdmin -init
   

6. Create the active standby pair replication scheme.

   ttCWAdmin -create -dsn advancedDSN
   

7. Start the active standby pair replication scheme.

   ttCWAdmin -start -dsn advancedDSN
   

Adding a read-only subscriber to an active standby pair

To add a read-only subscriber to an active standby pair replication scheme managed by Oracle Clusterware, perform these steps:

1. Stop the replication agents on all databases. This example uses the advancedSubscriberDSN, which already has a subscriber and is configured for advanced availability.

   ttCWAdmin -stop -dsn advancedSubscriberDSN
   

2. Drop the active standby pair.

   ttCWAdmin -drop -dsn advancedSubscriberDSN
   

3. Modify the cluster.oracle.ini file.

   • Add the subscriber to the SubscriberHosts attribute.

   • If the cluster is configured for advanced availability, add a virtual IP address to the SubscriberVIP attribute.

   See "Configuring advanced availability" for an example using these attributes.

4. Create the active standby pair replication scheme.

   ttCWAdmin -create -dsn advancedSubscriberDSN
   

5. Start the active standby pair replication scheme.

   ttCWAdmin -start -dsn advancedSubscriberDSN
   

Removing a read-only subscriber from an active standby pair

To remove a read-only subscriber from an active standby pair, perform these steps:

1. Stop the replication agents on all databases. This example uses the advancedSubscriberDSN, which has a subscriber and is configured for advanced availability.

   ttCWAdmin -stop -dsn advancedSubscriberDSN
   

2. Drop the active standby pair.

   ttCWAdmin -drop -dsn advancedSubscriberDSN
   

3. Modify the cluster.oracle.ini file.

   • Remove the subscriber from the SubscriberHosts attribute or remove the attribute altogether if there are no subscribers left in the active standby pair.

   • Remove a virtual IP from the SubscriberVIP attribute or remove the attribute altogether if there are no subscribers left in the active standby pair.

4. Create the active standby pair replication scheme.

   ttCWAdmin -create -dsn advancedSubscriberDSN
   

5. Start the active standby pair replication scheme.

   ttCWAdmin -start -dsn advancedSubscriberDSN
   

Adding an active standby pair to a cluster

To add an active standby pair (with or without subscribers) to a cluster that is already managing an active standby pair, perform these tasks:

1. Create and populate a database on the host where you intend the active database to reside initially. See "Create and populate a TimesTen database on one host".

2. Modify the cluster.oracle.ini file. This example adds advSub2DSN to the cluster.oracle.ini file that already contains the configuration for advancedSubscriberDSN. The new active standby pair is on different hosts from the original active standby pair.

   [advancedSubscriberDSN]
   MasterHosts=host1,host2,host3
   SubscriberHosts=host4, host5
   MasterVIP=192.168.1.1, 192.168.1.2
   SubscriberVIP=192.168.1.3
   VIPInterface=eth0
   VIPNetMask=255.255.255.0
   
   [advSub2DSN]
   MasterHosts=host6,host7,host8
   SubscriberHosts=host9, host10
   MasterVIP=192.168.1.4, 192.168.1.5
   SubscriberVIP=192.168.1.6
   VIPInterface=eth0
   VIPNetMask=255.255.255.0
   

3. Create new virtual IP addresses. On Linux and UNIX, the user must be root to do this.

   ttCWAdmin -createVIPs -dsn advSub2DSN
   

4. Create the new active standby pair replication scheme.

   ttCWAdmin -create -dsn advSub2DSN
   

5. Start the new active standby pair replication scheme.

   ttCWAdmin -start -dsn advSub2DSN
   

Adding a read-only subscriber not managed by Oracle Clusterware

You can add a read-only subscriber that is not managed by Oracle Clusterware to an existing active standby pair replication scheme that is managed by Oracle Clusterware. Using the ttCWAdmin -beginAlterSchema command enables you to add the subscriber without dropping and recreating the replication scheme. Oracle Clusterware does not manage the subscriber because it is not part of the configuration that was set up for Oracle Clusterware management.

Perform these steps:

1. Enter the ttCWAdmin -beginAlterSchema command to stop the replication agent on the active and standby databases.

2. Using ttIsql to connect to the active database, add the subscriber to the replication scheme by using an ALTER ACTIVE STANDBY PAIR statement.

   ALTER ACTIVE STANDBY PAIR ADD SUBSCRIBER ROsubDSN ON host6;
   

3. Enter the ttCWAdmin -endAlterSchema command to duplicate the standby database, register the altered replication scheme and start replication.

4. Enter the ttIsql repschemes command to verify that the read-only subscriber has been added to the replication scheme.

5. Use the ttRepStateGet procedure to verify that the state of the standby database is STANDBY.

6. On the subscriber host, use ttRepAdmin -duplicate to duplicate the standby database to the read-only subscriber. See "Duplicating a database".

7. Start the replication agent on the subscriber host.

Rebuilding a read-only subscriber not managed by Oracle Clusterware

You can destroy and rebuild a read-only subscriber that is not managed by Oracle Clusterware. Perform these tasks:

1. Stop the replication agent on the subscriber host.

2. Use the ttDestroy utility to destroy the subscriber database.

3. On the subscriber host, use ttRepAdmin -duplicate to duplicate the standby database to the read-only subscriber. See "Duplicating a database".

Removing an active standby pair from a cluster

To remove an active standby pair (with or without subscribers) from a cluster, perform these tasks:

1. Stop the replication agents on all databases in the active standby pair. This example uses advSub2DSN, which was added in "Adding an active standby pair to a cluster".

   ttCWAdmin -stop -dsn advSub2DSN
   

2. Drop the active standby replication scheme.

   ttCWAdmin -drop -dsn advSub2DSN
   

3. Drop the virtual IP addresses for the active standby pair.

   ttCWAdmin -dropVIPs -dsn advSub2DSN
   

4. Modify the cluster.oracle.ini file (optional). Remove the entries for advSub2DSN.

5. If you want to destroy the databases, log onto each host that was included in the configuration for this active standby pair and use the ttDestroy utility.

   ttDestroy advSub2DSN
   

   For more information about ttDestroy, see "ttDestroy" in Oracle TimesTen In-Memory Database Reference.

Adding a host to the cluster

Adding a host requires that the cluster be configured for advanced availability. The examples in this section use the advancedSubscriberDSN.

To add two spare master hosts to a cluster, enter a command similar to the following:

ttCWAdmin -addMasterHosts -hosts "host8,host9" -dsn advancedSubscriberDSN

To add a spare subscriber host to a cluster, enter a command similar to the following:

ttCWAdmin -addSubscriberHosts -hosts "subhost1" -dsn advancedSubscriberDSN

Removing a host from the cluster

Removing a host from the cluster requires that the cluster be configured for advanced availability. MasterHosts must list more than two hosts if one of the master hosts is to be removed. SubscriberHosts must list at least one more host than the number of subscriber databases if one of the subscriber hosts is to be removed.

The examples in this section use the advancedSubscriberDSN.

To remove two spare master host from the cluster, enter a command similar to the following:

ttCWAdmin -delMasterHosts "host8,host9" -dsn advancedSubscriberDSN

To remove a spare subscriber hosts from the cluster, enter a command similar to the following:

ttCWAdmin -delSubscriberHosts "subhost1" -dsn advancedSubscriberDSN

Reversing the roles of the master databases

After a failover, the active and standby databases are on different hosts than they were before the failover. You can use the -switch option of the ttCWAdmin utility to restore the original configuration.

For example:

ttCWAdmin -switch -dsn basicDSN

Ensure that there are no open transactions before using the -switch option. If there are open transactions, the command fails.

Figure 7-6 shows the hosts for an active standby pair. The active database resides on host A, and the standby database resides on host B.

Figure 7-6 Hosts for an active standby pair

Description of "Figure 7-6 Hosts for an active standby pair"

The ttCWAdmin -switch command performs these tasks:

• Deactivates the TimesTen cluster agent (ttCRSAgent) on host A (the active node)

• Disables the database monitor (ttCRSmaster) on host A

• Calls the ttRepSubscriberWait, ttRepStop and ttRepDeactivate built-in procedures on host A

• Stops the active service (ttCRSActiveService) on host A and reports a failure event to the Oracle Clusterware CRSD process

• Enables monitoring on host A and moves the active service to host B

• Starts the replication agent on host A, stops the standby service (ttCRSStandbyService) on host B and reports a failure event to the Oracle Clusterware CRSD process on host B

• Starts the standby service (ttCRSStandbyService) on host A

Moving a database to a different host

When a cluster is configured for advanced availability, you can use the -relocate option of the ttCWAdmin utility to move a database from the local host to the next available spare host specified in the MasterHosts attribute in the cluster.oracle.ini file. If the database on the local host has the active role, the -relocate option first reverses the roles. Thus the newly migrated active database becomes the standby and the standby becomes the active.

The -relocate option is useful for relocating a database if you need to take the host offline. Ensure that there are no open transactions before you use the command.

For example:

ttCWAdmin -relocate -dsn advancedDSN

Performing host or network maintenance

If you need to upgrade the operating system or hardware for a host or perform network maintenance, shut down Oracle Clusterware and disable automatic startup. Execute these Oracle Clusterware commands as root or OS administrator:

# crsctl stop crs

# crsctl disable crs

Shut down TimesTen. See "Shutting down a TimesTen application" in Oracle TimesTen In-Memory Database Operations Guide.

Perform the host maintenance. Then enable automatic startup and start Oracle Clusterware:

# crsctl enable crs

# crsctl start crs

See Oracle Clusterware Administration and Deployment Guide for more information about these commands.

Performing maintenance on the entire cluster

When all of the hosts in the cluster need to be brought down, stop Oracle Clusterware on each host individually. Execute these Oracle Clusterware commands as root or OS administrator:

# crsctl stop crs

# crsctl disable crs

Shut down TimesTen. See "Shutting down a TimesTen application" in Oracle TimesTen In-Memory Database Operations Guide.

Perform the maintenance. Then enable automatic startup and start Oracle Clusterware:

# crsctl enable crs

# crsctl start crs

See Oracle Clusterware Administration and Deployment Guide for more information about these commands.

Changing user names or passwords

When you create the active standby pair replication scheme with the ttCWAdmin -create command, Oracle Clusterware prompts for the user name and password of the internal user. If there are cache groups in the active standby pair, Oracle Clusterware also stores the cache administration user name and password. To change the user name or password for the internal user or the cache administration user, you must re-create the cluster.

To change the user name or password of the internal user that created the active standby pair replication or to change the cache administration user name or password, perform these tasks:

1. Stop the replication agents on the databases in the active standby pair. These commands use the advancedCacheDSN as an example.

   ttCWAdmin -stop -dsn advancedCacheDSN
   

2. Drop the active standby pair.

   ttCWAdmin -drop -dsn advancedCacheDSN
   

3. Change the appropriate user name or password:

   • Change the internal user name or password by using the CREATE USER or ALTER USER statements. See "Creating or identifying users to the database" in Oracle TimesTen In-Memory Database Operations Guide.

   • Change the cache administration user name or password by using the ttCacheUidPwdSet built-in procedure. See "Setting the cache administration user name and password" in Oracle In-Memory Database Cache User's Guide.

4. Re-create the active standby pair replication scheme.

   ttCWAdmin -create -dsn advancedCacheDSN
   

5. Start the active standby pair replication scheme.

   ttCWAdmin -start -dsn advancedCacheDSN
   

Obtaining cluster status

Using the -status option of the ttCWAdmin utility reports information about all of the active standby pairs in an instance that are managed by the same instance administrator. If you specify the DSN, the utility reports information for the active standby pair with that DSN.

$ ttCWAdmin -status
TimesTen Cluster status report as of Thu Nov 11 13:54:35 2010
 
====================================================================
TimesTen daemon monitors:
Host:HOST1 Status: online
Host:HOST2 Status: online
 
====================================================================
====================================================================
TimesTen Cluster agents
Host:HOST1 Status: online
Host:HOST2 Status: online
 
====================================================================
 
Status of Cluster related to DSN MYDSN:
====================================================================
1. Status of Cluster monitoring components:
Monitor Process for Active datastore:NOT RUNNING
Monitor Process for Standby datastore:NOT RUNNING
Monitor Process for Master Datastore 1 on Host host1: NOT RUNNING
Monitor Process for Master Datastore 2 on Host host2: NOT RUNNING
 
2.Status of  Datastores comprising the cluster
Master Datastore 1:
Host:host1
Status:AVAILABLE
State:ACTIVE
Grid:NO GRID
Master Datastore 2:
Host:host2
Status:UNAVAILABLE
State:UNKNOWN
Grid:UNKNOWN
====================================================================
The cluster containing the replicated DSN is offline

$ ttcwadmin -status
TimesTen Cluster status report as of Thu Nov 11 13:58:25 2010
 
====================================================================
TimesTen daemon monitors:
Host:HOST1 Status: online
Host:HOST2 Status: online
 
====================================================================
====================================================================
TimesTen Cluster agents
Host:HOST1 Status: online
Host:HOST2 Status: online
 
====================================================================
 
Status of Cluster related to DSN MYDSN:
====================================================================
1. Status of Cluster monitoring components:
Monitor Process for Active datastore:RUNNING on Host host1
Monitor Process for Standby datastore:RUNNING on Host host1
Monitor Process for Master Datastore 1 on Host host1: RUNNING
Monitor Process for Master Datastore 2 on Host host2: RUNNING
 
2.Status of  Datastores comprising the cluster
Master Datastore 1:
Host:host1
Status:AVAILABLE
State:ACTIVE
Grid:NO GRID
Master Datastore 2:
Host:host2
Status:AVAILABLE
State:IDLE
Grid:NO GRID
====================================================================
The cluster containing the replicated DSN is online

2.Status of  Datastores comprising the cluster
Master Datastore 1:
Host:host1
Status:AVAILABLE
State:ACTIVE
Grid:AVAILABLE
Master Datastore 2:
Host:host2
Status:AVAILABLE
State:IDLE
Grid:NO GRID

$ ttcwadmin -status
TimesTen Cluster status report as of Thu Nov 11 13:59:20 2010
 
====================================================================
TimesTen daemon monitors:
Host:HOST1 Status: online
Host:HOST2 Status: online
 
====================================================================
====================================================================
TimesTen Cluster agents
Host:HOST1 Status: online
Host:HOST2 Status: online
 
====================================================================
 
Status of Cluster related to DSN MYDSN:
====================================================================
1. Status of Cluster monitoring components:
Monitor Process for Active datastore:RUNNING on Host host1
Monitor Process for Standby datastore:RUNNING on Host host2
Monitor Process for Master Datastore 1 on Host host1: RUNNING
Monitor Process for Master Datastore 2 on Host host2: RUNNING
 
2.Status of  Datastores comprising the cluster
Master Datastore 1:
Host:host1
Status:AVAILABLE
State:ACTIVE
Grid:NO GRID
Master Datastore 2:
Host:host2
Status:AVAILABLE
State:STANDBY
Grid:NO GRID
====================================================================
The cluster containing the replicated DSN is online

2.Status of  Datastores comprising the cluster
Master Datastore 1:
Host:host1
Status:AVAILABLE
State:ACTIVE
Grid:AVAILABLE
Master Datastore 2:
Host:host2
Status:AVAILABLE
State:STANDBY
Grid:AVAILABLE

Message log files

The monitor processes report events and errors to the ttcwerrors.log and ttcwmsg.log files. The files are located in the daemon_home/info directory. The default size of these files is the same as the default maximum size of the user log. The maximum number of log files is the same as the default number of files for the user log. When the maximum number of files has been written, additional errors and messages overwrite the files, beginning with the oldest file.

For the default values for number of log files and log file size, see "Modifying informational messages" in Oracle TimesTen In-Memory Database Operations Guide.