Matakuliah Tahun Versi : M0184 / Pengolahan Data Distribusi : 2005 : Session – 16 RECOVERY CONTROL - 1 OBJECTIVE • The importance role of Log file for recovery plan • Checkpoint Log File • Function : recorded all operation on the DB that carried out by all transaction. • One log file for each DBMS which is shared by all transaction under the control of DBMS • In DDB environment, each site will have its own separate log. • Another purpose of log file is for performance monitoring and auditing Log File contains information : • Transaction identifier • Type of log record, that is, which of the list of DB actions listed above it is recording • Identifier of data object affected by the DB action (insert, delete and update) • Before-image of the data object, that is the value before update • After-image of data object, that is its value after update • Log management and information such as pointer to previous log record for that transaction Log File • Log files were traditionally stored on magnetic tape, because tape was a more reliable form of stable storage than magnetic disk and was cheaper • Now, the log file can be stored on-line on a fast direct access storage device to be able to recover quickly from minor failures Local Recovery Manager Local Transaction Manger Local Scheduler Local Recovery Manager DB Buffer Manager Database Buffer DB Log Buffer Log File Main memory Secondary storage Checkpoint • How far back in the log to go in order to identify transaction which might have redone ? • To limit its search, the recovery manager takes periodic checkpoint and on recovery it only has to go back as far as the last checkpoint • Two Approach : synchronous and asynchronous Synchronous Checkpoint • System stops accepting any new transaction until all executing transaction have finished. TC5 TC4 TC1 Synchronous checkpoint Failure Asynchronous Checkpoint • System processing is allowed to continue uninterrupted by the checkpoint TC5 TC4 TC3 TC2 TC1 Asynchronous checkpoint Failure Asynchronous Checkpoint The following action are carried out at an synchronous checkpoint : • A list of all transaction currently active in the system is written to the log • The address of the checkpoint record in the log is written to special file known as the restart file • All buffers are forcewritten to both the log and the BD