High Availability of Oracle Flashback

Introduction

The Flashback database is a point-in-time (PIT) database restoration method. This Incomplete recovery policy can be used to restore a database that is logically damaged due to human errors. After being introduced in 10 Gbit/s, it is designed to achieve maximum availability by reducing the recovery time.

Traditional recovery vs. Reproduction of databases

The first cause of downtime is logical damage caused by human errors, which has been widely recognized. Examples of logical corruption can be seen everywhere, from incorrect data update and table truncation by the user to incorrect batch processing task running twice or disordered order. The results are the same-the database is corrupted and the results are wide and hard to identify. Oracle uses two policies to return the database to a previous point in time: traditional restoration and reproduction of the database.

Incomplete recovery refers to the restoration from the database to a previous state. This process involves two steps: Re-store the data and restore the transaction activity to a certain time you want. The main difference between traditional recovery and database reproduction is that the traditional recovery starts from re-storing all the data files before restoring to a desired recovery time, the reproduction of the database is performed backward by re-storing the changed block after the database is damaged. From this perspective, let's think about a 10 TB database with 1 MB of data corruption. Traditional restoration starts from re-storing 10 TB of application data, while recreating the database is to retrieve the 1 MB of application data to reach the point before the damage. Now let's look at each policy separately.

Traditional recovery

Before Oracle 10 Gb, the only option to restore a database that is faulty due to human error to a previous point in time is traditional recovery. This policy includes extracting from the backup and re-storing all database data files, and then executing the restore to a desired time point. Media recovery can be based on servers (RMAN) or users (operating system tools ). Recovery is completed by re-storing all data files from the backup and rolling back the redo log to the desired time point. The recovery time is proportional to the database size, rather than the number of changes to be restored. This means that the recovery time (MTTR) actually increases with the growth of the database scale.

Reproduce Database

In Oracle 10 Gb, a new reproduction technology, known as Flashback Database, is introduced as a substitute for traditional restoration. Recreating a database allows you to quickly restore the entire database to a previous point in time without having to re-store the database from the backup. It is often described as an inverted button in the database. It only restores the modified data blocks to the desired recovery time. Then apply the Redo change record to reach the desired recovery time point. The modified data block is called Log reproduction.

Recreating a database provides significant advantages over traditional databases. Analyticdb does not have such obvious advantages. In a data warehouse, block operations are generally performed in a mode that does not record logs. In recreating a database, as long as the database runs in the document log mode, it can return to a status before the block operation, because the modified block can be restored and the operation is canceled.

1 2 3 4 5 6 7 8 Next Page