A cluster is a group of computers, called nodes, that work together to run a set of applications. Clients connecting to applications running on a cluster treat the cluster as a single system. If a node fails, other nodes in the cluster can automatically take over the services provided by the failed node. Clients may see a slight disruption in availability (the time it takes to resume the services on the remaining nodes), but are otherwise unaware that the node has failed.
When you use clustering with SQL Anywhere, any uncommitted transactions are lost when a database or database server fails over to another node in the cluster, and clients must reconnect to the database after failover occurs.
SQL Anywhere supports a variety of cluster environments where the cluster software can make any application into a generic resource subject to automatic failover so that high availability can be provided. However, only the database server process can be failed over, and the monitoring and control processes are limited.
For more information, see http://www.ianywhere.com/developer/technotes/asa_cluster_db_service.html.
Most cluster software provides an API for creating custom resources tailored to a specific application. SQL Anywhere includes two custom failover resources for Veritas Cluster Server: SAServer and SADatabase. The SAServer agent is responsible for database server failover, while the SADatabase agent is responsible for the failover of a specific database file. You can use one or both agents, depending on your application.
Your systems must be set up as follows to use the SQL Anywhere Veritas Cluster Server agents:
You must use Veritas Cluster Server 4.1 or later.
SQL Anywhere must be installed identically on each system node within the cluster.
Database files must be stored on a shared storage device that is accessible to all systems within the cluster.
The utility database password must be the same for all systems within the cluster.
The SADatabase agent uses the utility database to start and stop specific database files. All systems participating in the cluster must have the same utility database password. You can set the utility database password by specifying the -su server option when starting the database server or by adding the following line to the install-dir\win32\util_db.ini file on each node of the cluster:
pwd = [password-of-your-choice]
On Unix, the VCS agent is installed in install-dir/vcsagent/saserver.
There are three ways to configure and add a new agent to Veritas Cluster Server:
Using the Cluster Manager.
Using command line utilities.
Using a text editor and editing the main.cf configuration files.
The instructions in the following sections use the Cluster Manager.
For information about the available utilities, see Veritas Cluster Server Administration Guide
If you want to configure main.cf manually using a text editor, you must stop all Veritas Cluster Server services before editing the main.cf file. Otherwise, the changes will not take effect.
Configuring the SAServer agent
Configuring the SADatabase agent