9 Defining Classic Replication Schemes

Considering failover and recovery scenarios

As you plan a replication scheme, consider every failover and recovery scenario. For example, subscriber failures generally have no impact on the applications connected to the master databases. Their recovery does not disrupt user service. If a failure occurs on a master database, you should have a means to redirect the application load to a subscriber and continue service with no or minimal interruption. This process is typically handled by a cluster manager or custom software designed to detect failures, redirect users or applications from the failed database to one of its subscribers, and manage recovery of the failed database. See Chapter 15, "Managing Database Failover and Recovery".

When planning failover strategies, consider which subscribers are to take on the role of the master and for which users or applications. Also, consider recovery factors. For example, a failed master must be able to recover its database from its most up-to-date subscriber, and any subscriber must be able to recover from its master. A bidirectional scheme that replicates the entire database can take advantage of automatic restoration of a failed master. See "Automatic catch-up of a failed master database".

Consider the failure scenario for the unidirectionally replicated database shown in Figure 9-1. In the case of a master failure, the application cannot access the database until it is recovered from the subscriber. You cannot switch the application connection or user load to the subscriber unless you use an ALTER REPLICATION statement to redefine the subscriber database as the master. See "Replacing a master database".

Figure 9-1 Recovering a master in a unidirectional scheme

Description of "Figure 9-1 Recovering a master in a unidirectional scheme"

Figure 9-2 shows a bidirectional distributed workload scheme in which the entire database is replicated. Failover in this type of replication scheme involves shifting the users of the application on the failed database to the application on the surviving database. Upon recovery, the workload can be redistributed to the application on the recovered database.

Figure 9-2 Recovering a master in a distributed workload scheme

Description of "Figure 9-2 Recovering a master in a distributed workload scheme"

Similarly, the users in a split workload scheme must be shifted from the failed database to the surviving database. Because replication in a split workload scheme is not at the database level, you must use an ALTER REPLICATION statement to set a new master database. See "Replacing a master database". Upon recovery, the users can be moved back to the recovered master database.

Propagation replication schemes also require the use of the ALTER REPLICATION statement to set a new master or a new propagator if the master or propagator fails. Higher availability is achieved if two propagators are defined in the replication scheme. See Figure 1-11 for an example of a propagation replication scheme with two propagators.

Making decisions about performance and recovery tradeoffs

When you design a replication scheme, weigh operational efficiencies against the complexities of failover and recovery. Factors that may complicate failover and recovery include the network topology that connects a master with its subscribers and the complexity of the replication scheme. For example, it is easier to recover a master that has been fully replicated to a single subscriber than recover a master that has selected elements replicated to different subscribers.

You can configure replication to work asynchronously (the default), "semi-synchronously" with return receipt service, or fully synchronously with return twosafe service. Selecting a return service provides greater confidence that your data is consistent on the master and subscriber databases. Your decision to use default asynchronous replication or to configure return receipt or return twosafe mode depends on the degree of confidence you require and the performance tradeoff you are willing to make in exchange.

Table 9-1 summarizes the performance and recover tradeoffs of asynchronous replication, return receipt service and return twosafe service.

In addition to the performance and recovery tradeoffs between the two return services, you should also consider the following:

• Return receipt can be used in more configurations, whereas return twosafe can only be used in a bidirectional configuration or an active standby pair.

• Return twosafe enables you to specify a "local action" to be taken on the master database in the event of a timeout or other error encountered when replicating a transaction to the subscriber database.

A transaction is classified as return receipt or return twosafe when the application updates a table that is configured for either return receipt or return twosafe. Once a transaction is classified as either return receipt or return twosafe, it remains so, even if the replication scheme is altered before the transaction completes.

For more information about return services, see "Using a return service in a classic replication scheme".

Distributing workloads

Consider configuring the databases to distribute application workloads and make the best use of a limited number of servers. For example, it may be efficient and economical to configure the databases in a bidirectional distributed workload replication scheme so that each serves as both master and subscriber, rather than as separate master and subscriber databases. However, a distributed workload scheme works best with applications that primarily read from the databases. Implementing a distributed workload scheme for applications that frequently write to the same elements in a database may diminish performance and require that you implement a solution to prevent or manage update conflicts, as described in Chapter 13, "Resolving Replication Conflicts".

Owner of the replication scheme and replicated objects

The replication scheme and the replicated objects must be owned by the same user on every database in a replication scheme. To ensure that there is a common owner across all databases, you should explicitly specify the user and replication scheme in the CREATE REPLICATION statement.

For example, create a replication scheme named repscheme owned by user repl. The first line of the CREATE REPLICATION statement for repscheme is:

CREATE REPLICATION rep1.repscheme

Database names

These are the roles of the databases in a replication scheme:

• Master: Applications update the master database. The master sends the updates to the propagator or to the subscribers directly.

• Propagator: The propagator database receives updates from the master database and sends them to subscriber databases.

• Subscriber: Subscribers receive updates from the propagator or the master.

Before you define the replication scheme, you need to define the data source names (DSNs) for the databases in the replication scheme. On UNIX platforms, create an odbc.ini file. On Windows, use the ODBC Administrator to name the databases and set connection attributes. See "Step 1: Create the DSNs for the master and the subscriber" for an example.

Each database "name" specified in a replication scheme must match the prefix of the database file name without the path specified for the DataStore data store attribute in the DSN definition. Use the same name for both the DataStore and Data Source Name data store attributes in each DSN definition. If the database path is directory/subdirectory/foo.ds0, then foo is the database name that you should use. For example, this entry in an odbc.ini file shows a Data Source Name (DSN) of masterds, while the DataStore value shows the path for masterds:

[masterds]
DataStore=/tmp/masterds
DatabaseCharacterSet=AL32UTF8
ConnectionCharacterSet=AL32UTF8

Defining the DATASTORE element

To replicate the entire contents of the master database (masterds) to the subscriber database (subscriberds), the ELEMENT description (named ds1) might look like the following:

ELEMENT ds1 DATASTORE
  MASTER masterds ON "system1"
  SUBSCRIBER subscriberds ON "system2"

Identify a database host using the host name returned by the hostname operating system command. It is good practice to surround a host name with double quotes.

You can choose to exclude certain tables and sequences from the DATASTORE element by using the EXCLUDE TABLE and EXCLUDE SEQUENCE clauses of the CREATE REPLICATION statement. When you use the EXCLUDE clauses, the entire database is replicated to all subscribers in the element except for the objects that are specified in the EXCLUDE clauses. Use only one EXCLUDE TABLE and one EXCLUDE SEQUENCE clause in an element description. For example, this element description excludes two tables and one sequence:

ELEMENT ds1 DATASTORE
  MASTER masterds ON "system1"
  SUBSCRIBER subscriberds ON "system2"
  EXCLUDE TABLE ttuser.tab1, ttuser.tab2
  EXCLUDE SEQUENCE ttuser.seq1

You can choose to include only certain tables and sequences in the database by using the INCLUDE TABLE and INCLUDE SEQUENCE clauses of the CREATE REPLICATION statement. When you use the INCLUDE clauses, only the objects that are specified in the INCLUDE clauses are replicated to each subscriber in the element. Use only one INCLUDE TABLE and one INCLUDE SEQUENCE clause in an element description. For example, this element description includes one table and two sequences:

ELEMENT ds1 DATASTORE
  MASTER masterds ON "system1"
  SUBSCRIBER subscriberds ON "system2"
  INCLUDE TABLE ttuser.tab3
  INCLUDE SEQUENCE ttuser.seq2, ttuser.seq3

Defining table elements

To replicate the ttuser.tab1 and ttuser.tab2 tables from a master database (named masterds and located on a host named system1) to a subscriber database (named subscriberds on a host named system2), the ELEMENT descriptions (named a and b) might look like the following:

ELEMENT a TABLE ttuser.tab1
  MASTER masterds ON "system1"
 SUBSCRIBER subscriberds ON "system2"
ELEMENT b TABLE ttuser.tab2
  MASTER masterds ON "system1"
  SUBSCRIBER subscriberds ON "system2"

For requirements for tables in classic replication schemes, see "Table requirements and restrictions for classic replication schemes".

Replicating tables with foreign key relationships in a classic replication scheme

In a classic replication scheme, you may choose to replicate all or a subset of tables that have foreign key relationships with one another. To do so, create the tables and the foreign key relationship on each master and subscriber. Then, add the tables to the replication scheme with the ALTER REPLICATION ADD ELEMENT statement on each master and subscriber.

However, if the foreign key relationships have been configured with ON DELETE CASCADE, then you must create all of the tables before the replication scheme is created. Then, configure the replication scheme with the CREATE REPLICATION statment to include all tables with either the DATASTORE element (that does not exclude any of the tables) or the TABLE element for every table that is involved in the relationship.

You cannot add a table with a foreign key relationship configured with ON DELETE CASCADE to the replication scheme after the replication scheme is created with the ALTER REPLICATION statement. Instead, you must drop the replication scheme, create the new table with the foreign key relationship with ON DELETE CASCADE, and then create a new replication scheme that includes all of the related tables.

If a table with a foreign key configured with ON DELETE CASCADE is replicated, then the matching foreign key on the subscriber must also be configured with ON DELETE CASCADE. In addition, you must replicate any other table with a foreign key relationship to that table. This requirement prevents foreign key conflicts from occurring on subscriber tables when a cascade deletion occurs on the master database.

TimesTen replicates a cascade deletion as a single operation, rather than replicating to the subscriber each individual row deletion which occurs on the child table when a row is deleted on the parent. As a result, any row on the child table on the subscriber database, which contains the foreign key value that was deleted on the parent table, is also deleted, even if that row did not exist on the child table on the master database.

Replicating sequences

Sequences are replicated unless you exclude them from the replication scheme or unless they have the CYCLE attribute. Replication of sequences is optimized by reserving a range of sequence numbers on the standby database each time a sequence is updated on the active database. Reserving a range of sequence numbers reduces the number of updates to the transaction log. The range of sequence numbers is called a cache. Sequence updates on the active database are replicated only when they are followed by or used in replicated transactions.

Consider a sequence my.seq with a MINVALUE of 1, an INCREMENT of 1 and the default Cache of 20. The very first time that you use my.seq.NEXTVAL, the current value of the sequence on the master database is changed to 2, and a new current value of 21 (20+1) is replicated to the subscriber. The next 19 references to my.seq.NEXTVAL on the master database result in no new current value being replicated, because the current value of 21 on the subscriber database is still ahead of the current value on the master. On the twenty-first reference to my.seq.NEXTVAL, a new current value of 41 (21+20) is transmitted to the subscriber database because the subscriber's previous current value of 21 is now behind the value of 22 on the master.

Sequence replication has these restrictions:

• Sequences with the CYCLE attribute cannot be replicated.

• The definition of the replicated sequence on each peer database must be identical.

• No conflict checking is performed on sequences. If you make updates to sequences in both databases in a bidirectional replication configuration without using the RETURN TWOSAFE service, it is possible for both sequences to return the identical NEXTVAL.

If you need to use sequences in a bidirectional replication scheme where updates may occur on either peer, you may instead use a nonreplicated sequence with different MINVALUE and MAXVALUE attributes on each database to avoid conflicts. For example, you may create sequence my.seq on database DS1 with a MINVALUE of 1 and a MAXVALUE of 100, and the same sequence on DS2 with a MINVALUE of 101 and a MAXVALUE of 200. Then, if you configure DS1 and DS2 with a bidirectional replication scheme, you can make updates to either database using the sequence my.seq with the guarantee that the sequence values never conflict. Be aware that if you are planning to use ttRepAdmin -duplicate to recover from a failure in this configuration, you must drop and then re-create the sequence with a new MINVALUE and MAXVALUE after you have performed the duplicate operation.

Operations on sequences such as SELECT my.seq.NEXTVAL FROM sys.dual, while incrementing the sequence value, are not replicated until they are followed by transactions on replicated tables. A side effect of this behavior is that these sequence updates are not purged from the log until followed by transactions on replicated tables. This causes ttRepSubscriberWait and ttRepAdmin -wait to fail when only these sequence updates are present at the end of the log.

To replicate the ttuser.seq sequence from a master database (named masterds and located on a host named system1) to a subscriber database (named subscriberds on a host named system2), the element description (named a) might look like the following:

ELEMENT a SEQUENCE ttuser.seq
  MASTER masterds ON "system1"
  SUBSCRIBER subscriberds ON "system2"

Views and materialized views in a replicated database

A materialized view is a summary of data selected from one or more TimesTen tables, called detail tables. Although you cannot replicate materialized views directly, you can replicate their underlying detail tables in the same manner as you would replicate regular TimesTen tables.

The detail tables on the master and subscriber databases can be referenced by materialized views. However, TimesTen replication verifies only that the replicated detail tables have the same structure on both the master and subscriber. It does not enforce that the materialized views are the same on each database.

If you replicate an entire database containing a materialized or non-materialized view as a DATASTORE element, only the detail tables associated with the view are replicated. The view itself is not replicated. A matching view can be defined on the subscriber database, but is not required. If detail tables are replicated, TimesTen automatically updates the corresponding view.

Materialized views defined on replicated tables may result in replication failures or inconsistencies if the materialized view is specified so that overflow or underflow conditions occur when the materialized view is updated.

Single classic subscriber schemes

The classic replication scheme shown in Example 9-2 is a single master and subscriber unidirectional replication scheme. The two databases are located on separate hosts, system1 and system2. We use the RETURN RECEIPT service to confirm that all transactions committed on the ttuser.tab table in the master database are received by the subscriber.

CREATE REPLICATION repscheme
ELEMENT e TABLE ttuser.tab
    MASTER masterds ON "system1"
    SUBSCRIBER subscriberds ON "system2"
      RETURN RECEIPT;

CREATE REPLICATION repscheme
ELEMENT e DATASTORE
    MASTER masterds ON "server1"
    SUBSCRIBER subscriberds ON "server2";

Multiple subscriber classic replication schemes with return services and a log failure threshold

You can create a classic replication scheme that includes up to 128 subscriber databases. If you are configuring propagator databases, you can configure up to 128 propagators. Each propagator can have up to 128 subscriber databases. See "Propagation scheme" for an example of a classic replication scheme with propagator databases.

CREATE REPLICATION twosubscribers
ELEMENT e TABLE ttuser.tab
    MASTER masterds ON "server1"
    SUBSCRIBER subscriber1ds ON "server2",
               subscriber2ds ON "server3";

CREATE REPLICATION twosubscribers
ELEMENT e TABLE ttuser.tab
  MASTER masterds ON "server1"
  SUBSCRIBER subscriber1ds ON "server2",
             subscriber2ds ON "server3"
  RETURN RECEIPT
STORE masterds FAILTHRESHOLD 10;

CREATE REPLICATION twosubscribers
ELEMENT e TABLE ttuser.tab
    MASTER masterds ON "server1"
    SUBSCRIBER subscriber1ds ON "server2"
    SUBSCRIBER subscriber2ds ON "server3" RETURN RECEIPT
STORE masterds FAILTHRESHOLD 10;

CREATE REPLICATION twosubscribers
ELEMENT e TABLE ttuser.tab
    MASTER masterds ON "server1"
    SUBSCRIBER subscriberds1 ON "server2" RETURN RECEIPT BY REQUEST
    SUBSCRIBER subscriber2ds ON "server3" RETURN RECEIPT
STORE masterds FAILTHRESHOLD 10;

Replicating tables to different subscribers

The classic replication scheme shown in Example 9-8 establishes a master database, named centralds, that replicates four tables. ttuser.tab1 and ttuser.tab2 are replicated to the subscriber backup1ds. ttuser.tab3 and ttuser.tab4 are replicated to backup2ds. The master database is located on the finance server. Both subscribers are located on the backupsystem server.

CREATE REPLICATION twobackups
ELEMENT a TABLE ttuser.tab1
  MASTER centralds ON "finance"
  SUBSCRIBER backup1ds ON "backupsystem"
ELEMENT b TABLE ttuser.tab2
  MASTER centralds ON "finance"
  SUBSCRIBER backup1ds ON "backupsystem"
ELEMENT d TABLE ttuser.tab3
  MASTER centralds ON "finance"
  SUBSCRIBER backup2ds ON "backupsystem"
ELEMENT d TABLE ttuser.tab4
  MASTER centralds ON "finance"
  SUBSCRIBER backup2ds ON "backupsystem";

Propagation scheme

In Example 9-9, the master database sends updates on a table to a propagator that forwards the changes to two subscribers. The master database is centralds on the finance host. The propagator database is propds on the nethandler host. The subscribers are backup1ds on backupsystem1 and backup2ds on backupsystem2.

The classic replication scheme has two elements. For element a, the changes to the tab table on centralds are replicated to the propds propagator database. For element b, the changes to the tab table received by propds are replicated to the two subscribers, backup1ds and backup2ds.

CREATE REPLICATION propagator
ELEMENT a TABLE ttuser.tab
  MASTER centralds ON "finance"
  SUBSCRIBER propds ON "nethandler"
ELEMENT b TABLE ttuser.tab
  PROPAGATOR propds ON "nethandler"
  SUBSCRIBER backup1ds ON "backupsystem1",
             backup2ds ON "backupsystem2";

Bidirectional split workload schemes

In Example 9-10, there are two databases, westds on the westcoast host and eastds on the eastcoast host. Customers are represented in two tables: waccounts contains data for customers in the Western region and eaccounts has data for customers from the Eastern region. The westds database updates the waccounts table and replicates it to the eastds database. The eaccounts table is owned by the eastds database and is replicated to the westds database. The RETURN RECEIPT attribute enables the return receipt service to guarantee that transactions on either master table are received by their subscriber.

CREATE REPLICATION r1
ELEMENT elem_waccounts TABLE ttuser.waccounts
  MASTER westds ON "westcoast"
  SUBSCRIBER eastds ON "eastcoast" RETURN RECEIPT
ELEMENT elem_eaccounts TABLE ttuser.eaccounts
  MASTER eastds ON "eastcoast"
  SUBSCRIBER westds ON "westcoast" RETURN RECEIPT;

Bidirectional distributed workload scheme

Example 9-11 shows a bidirectional general workload classic replication scheme in which the ttuser.accounts table can be updated on either the eastds or westds database. Each database is both a master and a subscriber for the accounts table.

CREATE REPLICATION r1
ELEMENT elem_accounts_1 TABLE ttuser.accounts
  MASTER westds ON "westcoast"
  SUBSCRIBER eastds ON "eastcoast"
ELEMENT elem_accounts_2 TABLE ttuser.accounts
  MASTER eastds ON "eastcoast"
  SUBSCRIBER westds ON "westcoast";

When elements are replicated in this manner, the applications should write to each database in a coordinated manner to avoid simultaneous updates on the same data. To manage update conflicts, include a timestamp column of type BINARY(8) in the replicated table and enable timestamp comparison by including the CHECK CONFLICTS clause in the CREATE REPLICATION statement. See Chapter 13, "Resolving Replication Conflicts" for a complete discussion on how to manage update conflicts.

Example 9-12 shows that the tstamp timestamp column is included in the ttuser.accounts table. The CREATE REPLICATION statement has been modified to include the CHECK CONFLICTS clause.

CREATE TABLE ttuser.accounts (custname VARCHAR2(30) NOT NULL,
                       address VARCHAR2(80),
                       curbalance DEC(15,2),
                       tstamp BINARY(8),
                       PRIMARY KEY (custname));

CREATE REPLICATION r1
ELEMENT elem_accounts_1 TABLE ttuser.accounts
  CHECK CONFLICTS BY ROW TIMESTAMP
    COLUMN tstamp
    UPDATE BY SYSTEM
    ON EXCEPTION ROLLBACK WORK
  MASTER westds ON "westcoast"
  SUBSCRIBER eastds ON "eastcoast"
ELEMENT elem_accounts_2 TABLE ttuser.accounts
  CHECK CONFLICTS BY ROW TIMESTAMP
    COLUMN tstamp
    UPDATE BY SYSTEM
    ON EXCEPTION ROLLBACK WORK
  MASTER eastds ON "eastcoast"
  SUBSCRIBER westds ON "westcoast";