You
can manually recover the local portions of a distributed transaction, but this will usually result in
inconsistent data between the distributed databases. If a local recovery is performed, the remote
data will be out of sync.
To minimize the number of distributed recoveries necessary, you can influence the way that
the distributed transaction is processed. The transaction processing is influenced via the use of
commit point strength to tell the database how to structure the transaction.
Commit Point Strength
Each set of distributed transactions may reference multiple hosts and databases. Of those, one
host and database can normally be singled out as being the most reliable, or as owning the most
critical data. This database is known as the commit point site. If data is committed there, it should
be committed for all databases. If the transaction against the commit point site fails, the transactions
against the other nodes are rolled back. The commit point site also stores information about the
status of the distributed transaction.
The commit point site will be selected by Oracle based on each database??™s commit point
strength. This is set via the initialization file, as shown in the following example:
COMMIT_POINT_STRENGTH=100
632 Oracle Database 11g DBA Handbook
The values for the COMMIT_POINT_STRENGTH parameter are set on a scale relative to other
nodes participating in distributed transactions. In the preceding example, the value is set to 100
(the default is 1).
Pages:
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931