Backup Strategy and Method
The following must be considered before performing a backup:
- Selecting the data to be backed up
- Determine whether it is valid data worth being preserved.
- Determine whether to back up the entire database or only part of it.
- Check whether there are other files to be backed up along with the database.
- Choosing a backup method
- Choose the backup method from one of incremental and online backups. Also, specify whether to use compression backup, parallel backup, and mode.
- Prepare backup tools and devices available.
- Determining backup time
- Identify the time when the least usage in the database occur.
- Check the size of the archive logs.
- Check the number of clients using the database to be backed up.
An online backup (or a hot backup) is a method of backing up a currently running database. It provides a snapshot of the database image at a certain point in time. Because the backup target is a currently running database, it is likely that uncommitted data will be stored and the backup may affect the operation of other databases.
To perform an online backup, use the cubrid backupdb -C command.
An offline backup (or a cold backup) is a method of backing up a stopped database. It provides a snapshot of the database image at a certain point in time.
To perform an offline backup, use the cubrid backupdb -S command.
An incremental backup, which is dependent upon a full backup, is a method of only backing up data that have changed since the last backup. This type of backup has an advantage of requiring less volume and time than a full backup. CUBRID supports backup levels 0, 1 and 2. A higher level backup can be performed sequentially only after a lower lever backup is complete.
To perform an incremental backup, use the cubrid backupdb -l <level> command.
The following example shows incremental backup. Let's example backup levels in details.
- Full backup (backup level 0) : Backup level 0 is a full backup that includes all database pages.
- The level of a backup which is attempted first on the database naturally becomes a 0 level. DBA must perform full backups regularly to prepare for restore situations. In the example, full backups were performed on December 31st and January 5th.
- First incremental backup (backup level 1) : Backup level 1 is an incremental backup that only stores changes since the level 0 full backup, and is called a "first incremental backup."
- Note that the first incremental backups are attempted sequentially such as <1-1>, <1-2> and <1-3> in the example, but they are always performed based on the level 0 full backup.
- Suppose that backup files are created in the same directory. If the first incremental backup <1-1> is performed on January 1st and then the first incremental backup <1-2> is attempted again on January 2nd, the incremental backup file created in <1-1> is overwritten. The final incremental backup file is created on January 3rd because the first incremental backup is performed again on that day.
- Since there can be a possibility that the database needs to be restored the state of January 1st or January 2nd, it is recommended for DBA to store the incremental backup files <1-1> and <1-2> separately in storage media before overwriting with the final incremental file.
- Second incremental backup (backup level 2) : Backup level 2 is an incremental backup that only stores data that have changed since the first incremental backup, and is called a "second incremental backup."
- A second incremental backup can be performed only after the first incremental backup. Therefore, the second incremental backup attempted on January fourth succeeds; the one attempted on January sixth fails.
- Backup files created for backup levels 0, 1 and 2 may all be required for database restore. To restore the database to its state on January fourth, for example, you need the second incremental backup generated at <2-1>, the first incremental backup file generated at <1-3>, and the full backup file generated at <0-1>. That is, for a full restore, backup files from the most recent incremental backup file to the earliest created full backup file are required.
A compress backup is a method of backing up the database by compressing it. This type of backup reduces disk I/O costs and stores disk space because it requires less backup volume.
To perform a compress backup, use the cubrid backupdb -z|--compress command.
Parallel Backup Mode
A parallel or multi-thread backup is a method of performing as many backups as the number of threads specified. In this way, it reduces backup time significantly. Basically, threads are given as many as the number of CPUs in the system.
To perform a parallel backup, use the cubrid backupdb -t|--thread-count command.