|
|
WAS V6 has a new built in component called High Availability Manager (HAManager). This component is responsible for running key services on available servers. Prior to V6, these key services ran on dedicated servers and became single points of failure. Transaction Manager recovery is one of the key services that take advantage of the HAManager component to ensure that recovery of failed or in-doubt transactions can be done by other servers in the cluster.
The Transaction Manager supports three different HA policies to achieve the recovery of transaction logs in a highly available manner:
This is the default style of peer recovery initiation. If an appserver fails, the HAManager selects another server to perform peer recovery processing on behalf of the failed server.
This style of peer recovery must be explicitly configured. If an appserver fails, the operator can use the Administrative Console to select another server to perform the recovery processing.
This style of peer recovery must be explicitly configured. It indicates that external clustering software is monitoring the Transaction Manager and will failover to an appserver that is configured by the external clustering software to perform the recovery processing.
Chapter 6, WebSphere HAManager, contains information about Transaction Manager failover and specifically the effect of the One of N and Static policies. In addition, the white paper Transactional high availability and deployment considerations in WAS V6 provides an excellent description of the One of N and Static policies. You can find this paper at:
http://www.ibm.com/developerworks/websphere/techjournal/0504_beaven/ 0504_beaven.html
This section concentrates on how to use the No Operation (NoOP) policy for TM recovery.
The No Operation policy is used in situations where the customer requirements dictate great control over when and where transaction log recovery occurs. Some customers might want to store the WebSphere transaction logs on file systems that allow only one system to mount at a time, for example the Journal File System (JFS) of AIX. In such an environment, external clustering software is necessary to make the WebSphere solution highly available. The external HA software performs failure detection and controls the allocation of shared resources such as shared file systems and IP addresses.
