Last participant support enables the use of a single one-phase commit capable resource with any number of two-phase commit capable resources in the same global transaction.
At transaction commit, the two-phase commit resources are prepared first using the two-phase commit protocol, and if this is successful the one-phase commit-resource is then called to commit(one_phase). The two-phase commit resources are then committed or rolled back depending on the response of the one-phase commit resource.
Note: If the global transaction is distributed across multiple application servers that are all running at WebSphere Application Server version 5.1 or later then you can exploit last participant support to coordinate a one-phase commit capable resource and any number of two-phase commit capable resources within the same transaction, in a limited number of scenarios.
In this scenario, last participant support can coordinate a one-phase commit capable resource and any number of two-phase commit capable resources within the same transaction.
In this scenario, the transaction typically cannot be committed. To be able to commit (as part of a global transaction) a one-phase commit resource enlisted on a transaction subordinate server, the transaction service must delegate coordination responsibility from the transaction root to the subordinate server. This occurs only if no other resources were registered with the transaction root server.
Note: If the global transaction
is distributed across multiple application servers that are not running
at WebSphere Application Server version 5.1 or later, you cannot coordinate
access to one-phase and two-phase commit capable resources within the same
transaction.
Last participant support introduces an increased risk of an heuristic outcome to the transaction. That is, the transaction manager cannot be sure that all resources were completed in the same direction (either committed or rolled back). For this reason, to enable an application to coordinate access to one-phase and two-phase commit capable resources within the same transaction, you configure the application to accept the increased risk of an heuristic outcome.
An heuristic outcome occurs if the transaction service (JTS) receives no response from the commit one-phase flow on the one-phase commit resource. In this situation the transaction service cannot determine whether changes for the one-phase commit resource were committed or rolled back, so cannot drive reliably the correct outcome of the global transaction on the other two-phase commit resources.
You can configure the transaction service for an application server to indicate whether or not to log that it is about to commit the one-phase commit resource. This does not reduce the heuristic hazard, but ensures that any failure, and subsequent recovery, of the application server during the one-phase commit phase occurs with knowledge of whether or not the one-phase commit resource was asked to commit: