Recommendations for performing a backup

About this section

Some recommendations in this section are applicable to all kinds of backups; some recommendations depend on your priorities, such as your need to minimize backup time or to minimize tape handling.

General recommendations

Follow the recommendations in this section when you back up data from the filer.

Avoid backing up too much data in a single dump command

The reasons are listed below:

The dump command cannot be restarted. That is, if the dump command encounters an error, you cannot correct the error and proceed from the point where the command fails. You must start the command from the beginning.
The backup takes a long time to finish. This leaves you with a long time period during which changed data cannot be written to tape by an incremental backup.

Store incremental backups for the same dump path on the same tape

If you want to do a subtree restore, you must restore from all incremental backups to restore the most recent data. Having all the incremental backups in the same tape minimizes tape management during the restore process.

Write down qtree information before backing up qtrees

Record the information about qtrees if your backup contains qtrees. This is because when you restore data from tape, the filer does not automatically re-create the qtrees. You must re-create the qtrees into which data is to be restored. Having a record of all qtrees enables you to re-create the qtrees quickly.

Recommendations for minimizing backup time and data loss

The recommendations in this section enable you to minimize both the time required to perform a backup and the possibility of data loss. This is because the shorter the time for the dump command to finish, the more incremental backups you can perform. A large number of incremental backups minimize the amount of unrecoverable data.

Note: There is a disadvantage to having a large number of incremental backups: When you restore data, you must restore from all the incremental backup tapes, which might take a long time.

Use multiple local tape drives

Attach the maximum number of tape drives to the filer. The filer can write faster to a local tape drive than to a tape drive attached to another system.

Organize data to be backed up

The first two passes of the dump command run faster if the dump path is one of these:

A volume
A qtree
Data not belonging to any qtrees

Limit the amount of data in each backup

Limit the amount of data in a volume or qtree to be backed up to 200 GB.

Schedule the backups appropriately

Schedule backups when the load on the filer is light. Refer to "Example of backing up the entire filer" for an example of executing one full backup and several incremental backups a night.

Avoid using an exclude list

If you exclude certain files from the backup, the dump command takes longer because it must compare each file name to the strings in the exclude list to determine whether it should back up the file.

Recommendations for minimizing downtime during data recovery

The method used for restoring data has a greater effect on downtime than does the dump command. However, it helps to keep the data in the dump path to 100 GB or less because a full restore is more efficient than a partial restore. For example, it takes less time to restore an entire 100-GB qtree than to extract 100 GB of data from a 500-GB backup.

Recommendations for minimizing the number of tape drives required

Because the filer supports the rmt protocol, several filers can share the same tape drive. Attach the tape drive to the filer with the most data to back up. Follow these guidelines if multiple filers back up to the same tape drive:

Use a private network for the backup so that the traffic load on the network does not slow down the backup process.
Schedule the dump command on each filer so that it starts only when no other filers are using the tape drive.

Part No. 210-02334
File: dump7.htm
Last updated: 02/24/99 12:17:01