Protecting PureData for Analytics data with Backup / Restore

Answer

One of the most important best practices for the PDA appliance is to ensure a good backup strategy is in place. Having a current backup will prevent a long outage in the unlikely event recovery is needed. In planning a backup and restore strategy, there are several factors to consider such as size of the data set, business critical availability of the data, and backup target resources available.

There are several options available for backing up PDA data include such as PDA backup utilities, PDA replication, duel ETL feeds, and the continuous backup and restore method. This article will focus on the PDA backup and restore utilities.

The PDA appliance has two different types of data that need to be backed up with each type having its own utilities for backing up and restoring data:

• database catalog, configuration, and log files located on the PDA host drives
• user data which is located on data or S-blade drives.

Catalog/host backup: catalog and all associated files on the host are backed up via the nzhostbackup which makes a copy of the of /nz/data directory. This backup is relatively short depending on the catalog size.

In a rare situation when the host data is corrupted or lost, but the user data is intact, you can restore the host data in a very short time without the having to restore all the databases.

It is recommended to perform a manual vacuum prior to an nzhostbackup. This will decrease the time and size of the backup.

nzhostbackup will pause the database during the backup. The utility will create a file with a name and location you provide. Once the backup is complete, you should date and archive this file off the server. The nzhostbackup should be taken regularly in order to capture schema and system files changes.



User data backup and restore utilities are nzbackup and nzrestore. You can configure the utilities to run using multiple streams. This will speed up the operations by reading and writing to multiple devices in parallel. nzbackup and nzrestore support multiple external backup and recovery solutions such as: filesystem, Veritas, Tivoli (IBM), and EMC Networker. PDA supplies a connector for each solution. Specific connector types are called as a command line argument that directs the data streams to pass data to the backup client software. See the PDA admin guide for detailed information on how to configure each connector type.

User data backups includes all the databases on the PDA appliance. It is vital this data is backed up on a regular basis. There are three modes available:

• Full backup will back up all databases, or a specific database specified on the command line. A full backup is important because it is the base to start when a full restore is needed. The full backup also frees up outdated and deleted rows to be removed via the groom command.
• Differential backup will back up only data that has changed since the last backup of any kind. This will generally be the smallest and shortest backup option. To restore, you would restore the last full backup and apply each subsequent differential.

• Cumulative backup consolidates all previous differential backups. This is useful to run between full backups after several differentials. This will simplify and shorten the restore by allowing you to use the cumulative backup in place of all the differentials.

An example of using all three backup could be:
1. Differential backups daily
2. Cumulative backups weekly
3. Full backups monthly

When backing up at the database level with nzbackup (by using the –db option) only user, group, and privileges saved in the specified database will be backed up. If you wish to include all users and groups, use the –global flag when backing up a database. It is also important to note that PDA does not provide a way to backup anything outside of the PDA elements. This means anything you may have installed or created on the host will not be backed up with PDA’s utilities.

While the nzbackup utilities offer a database level backup option, nzrestore allows table level restore. This is useful if you need to restore a specific table that may have been inadvertently dropped. If you need to restore multiple tables, you can list them in a file and specify the file name at nzrestore command line with the -tablefile option.

Details of each backup and restore are stored in the system database and viewed by the nzbackup -history and nzrestore –history commands. This will show the dates, what was backed up or restored, type, and location of the log files.