To maximize the reliability and availability of your servers, create a master backup plan that describes in detail what, when, and how backups will be performed. In addition to this master plan, be sure to create backup plans for each type or class of server. These plans should reflect the various applications and data stored on each server.
A good backup plan includes the following information:
Following is a more detailed explanation of the three types of backups and when they should be used:
Differential backup: Copies files created or changed since the last normal or incremental backup; does not mark files as having been backed up. Differential backups are usually much faster than full backups, but take somewhat longer to restore. To restore when you are performing only normal and differential backups, you need the full backup and the last differential backup.
Incremental backup: Copies files created or changed since the last normal or incremental backup, and, unlike differential backups, marks files as having been backed up. These are typically the fastest backups, but take the longest time to restore. To restore from a combination of normal and incremental backups, you need the full backup and each incremental backup in order.
Normal backup (full backup): Copies all selected files. Typically, a full backup is considered the most reliable, but might take too long to perform frequently.
Online: Online backups usually require the backup software to cooperate with the application through which normal database access is provided. The application must have extensions that allow it to retrieve the data and send it to the backup media. Online backups are typically much slower than offline backups and can generate a significant load on the server. If possible, perform online backups during slack times.
Offline: With an offline backup, it is assumed that user applications are not writing to the data files while the backup is running. This is not a problem for servers that are usually read-only, like Web servers or file servers. For these servers, the few files that are open during the backup are typically retried later, and if they are missed, they are logged. Offline backups are much faster than online backups and do not require interaction with the application software.
Offline backups are useful for backing up database applications before repairing or replacing disk systems. During scheduled down time, the applications are stopped, the data is backed up, the drives are repaired or replaced, and if necessary, the data is restored. It is recommended that you perform an offline backup before major hardware or software changes. If something fails, offline backups are the fastest way to recover.
If a failure occurs, it is recommended that you immediately start restoring the backup to a stand-by system. If the failed system can be repaired, nothing is lost. If the failed system cannot be repaired, however, your total repair time is reduced because you have already restored the backup to your standby system.
To maximize the reliability and availability of your servers, practice backup and restoring procedures routinely, and budget time for this practice. It is recommended that you practice backup routinely once a quarter for each backup plan. This is particularly important for operations staffs with a high rate of employee turnover.
When you practicing your backup, use a test computer, a stand-by computer, or a new computer that does not have software installed on it. Your practice computer should be similar to the one that is used in production, and preferably the type of computer that you use most often. During the practice backup, remove the practice computer from the network to prevent address or name collisions with the actual production computer.
It is recommended that you simulate a variety of types component failures and practice restoring them. This will give you experience performing different kinds of restoration and recovery. This is important because, for example, the approaches you use to restore a data drive can be quite different from those you use to restore a drive that contains the operating system.
When the practice backup is complete, review and analyze the recovery process. Identify the approaches that succeeded and those that did not. Determine which procedures need changing, and how the operation could be done faster. Then use this information to improve the recovery plan procedures.
In addition to conducting a review and analysis after a practice back up and recovery, you should be sure to do so after a real backup and recovery.