> -----Urspr�ngliche Nachricht----- > Von: Filip Sergeys [mailto:[EMAIL PROTECTED] > Gesendet: Mittwoch, 17. M�rz 2004 13:29 > An: [EMAIL PROTECTED] > Betreff: log reader, redo task and database backup question. > > > Hello, > > I want to setup a database HA like described in the SAP DB Ha howto by > Yves trudeau (sapdb.2scale.net/moin.cgi/SapDbHaHowto), > but I have a question about the database consistency in this setup. > > During a recover, the redo task reads the redolist. > For commits: > it uses the generate redo log files to roll forward the database > > For rollbacks: > if the rollback started before the savepoint -> the redo task > evaluates > the UNDO LOG FILE of the transaction, in order to set the > transaction to > the state before the savepoint. > (http://www.mysql.com/documentation/maxdb/62/79bb3ee5bf7173e10 > 000000a114084/content.htm) > => My question: the main database crashed, making the data area > unavailable. The backupdatabase uses the databackups and logbackups to > be rebuild, but the "last night" databackups do not contain these latest > undo log files (undo log files are stored in the data area) and the > logbackups only contain the redo log entries. Does this mean the > database in this case can not be reset to a consistent state? Or are the > undo log files also regenerated by the redo task?
Hi, the short answer is: You do not need to undo a change which never had happened. The longer answer: When reading the redo-log, all logentries are associated to their transaction by writing them into a redofile specific to their transaction. At the end of the transaction either a commit or a rollback will be found on the log. In case of a reollback the redo-information is thrown away, because these changes never had been made persistent. regards, Martin -- MaxDB Discussion Mailing List For list archives: http://lists.mysql.com/maxdb To unsubscribe: http://lists.mysql.com/[EMAIL PROTECTED]
