Contents Index Adaptive Server Enterprise transaction log and backup management Stable queue recovery issues

SQL Remote User's Guide
  Administering SQL Remote for Adaptive Server Enterprise
    Adaptive Server Enterprise transaction log and backup management

Protecting against media failure on the transaction log


Media failure on the transaction log can cause committed transactions to be lost. If the transaction log has been scanned and these transactions have already been sent to subscriber databases, then the subscribing databases contain transactions that are lost from the publishing database, and the databases are in an inconsistent state.

Why the transaction log is needed 

The transaction log is needed, even after the entries have been scanned into the stable queue, to guard against media failure on the database file. If the database is lost, it must be recovered to a point that includes every transaction that may have been sent to remote databases.

This recovery is done by restoring a database dump and loading transaction dumps to bring the database up to date. The last transaction dump restored is the dump of the active transaction log at the time of the failure.

Protecting against transaction log loss 

There are two ways of protecting against inconsistency arising from media failure on the transaction log:

Mirroring the transaction log 

The only way to protect against media failure on the transaction log is by mirroring the transaction log.

Disk mirroring can provide nonstop recovery in the event of media failure. The disk mirror command causes an Adaptive Server Enterprise database device to be duplicated—that is, all writes to the device are copied to a separate physical device. If one of the devices fails, the other contains an up-to-date copy of all transactions.

For information on disk mirroring in Adaptive Server Enterprise, see the chapter "Mirroring Database Devices", in the Adaptive Server Enterprise System Administration Guide.

Replicating only backed-up transactions 

The Message Agent also provides a command-line option (-u) that only sends transactions that have been backup up. In Adaptive Server Enterprise, this means transactions complete before the latest dump database command or dump transaction command.

Choosing an approach 

The goal of the strategy is to reduce the possibility of requiring re-extraction of remote databases to an acceptable level. In large setups, the possibility must be as close to zero as possible, as the cost of re-extraction (in terms of down time) is very high.


Contents Index Adaptive Server Enterprise transaction log and backup management Stable queue recovery issues