This section introduces concepts, terms, and behavior that you need to be familiar with before attempting to back up and restore Yellowbrick databases.

Backward Compatibility for Backup and Restore

A backup taken in Version 3.2.x or 3.3.x prior to a 4.0 upgrade can be restored in Version 4.0.

After a 4.0 upgrade or installation, the previous version of the backup and restore tools is compatible with Version 4.0. The Version 4.0 ybtools package installs two sets of backup and restore tools: legacy tools and new 4.0 tools.

The legacy tools are deprecated and will be removed in a future release, along with the ability to restore backups taken in Version 3.x releases. There are significant differences in the functionality and client tool options between the legacy tools and the new tools. The tool sets are distinguished by their names:

ybtools Version Backup Restore List/Manage Backups
New in Version 4.0 ybbackup ybrestore ybbackupctl
Legacy in Version 4.0 ybbackup-old ybrestore-old ybbackup-list-old
Version 3.x ybbackup ybrestore ybbackup-list
To use the legacy tools in the 4.0 package, be sure to use the correct tool name in the command. For example, to return help for the legacy version of ybbackup:
$ ybbackup-old --help
Yellowbrick Backup and Restore version 4.0.0-21507
Important: To use backup and restore in Version 4.0, you must follow these steps:
  1. Use the legacy version of ybtools to restore your 3.x backups into 4.0 databases.
  2. Take new full backups of your 4.0 databases with the new 4.0 ybtools.
  3. Stop using the legacy ybtools for backup and restore purposes. Use only the new tools.

To support the legacy tools and the new tools, two sets of backup and restore system views are supported in Version 4.0. When Version 4.0 is installed, it is possible to run legacy backup and restore operations and new database replication operations on the same Yellowbrick appliance. However, you cannot monitor the legacy operations by using the SMC; you must use the legacy system views.

Use Cases

The backup and restore feature primarily supports disaster recovery (DR) scenarios: cases where a Yellowbrick system is irreparably damaged or lost, or a specific database on a system is compromised. You can resolve these problems by restoring databases from cold storage to the same appliance or a dedicated DR system. To keep a DR system in sync with minimal overhead, you can use incremental restore operations, which apply changes to databases after they have been fully restored.

You can back up and restore a database to the same appliance where the original database was created:

Alternatively, you can restore a database to a completely separate Yellowbrick system.

To facilitate database access on different systems, options are available to restore with or without shared security objects and attributes: users, roles, role memberships, and grants. You can also take backups that exclude specific schemas and their objects. See Scope of Backups.

Another use case is the ability to make a copy of a database. You may want to copy a database under a different name, then use the new database as a separate development environment: for example, to test a specific workload or diagnose a problem without disrupting the source database. You can use either backup and restore operations or replication to "clone" databases.

Database replication is a related but separate feature, which uses database backups (or backup snapshots) to maintain database replicas on a specific target system. The fundamental differences between backup/restore operations and replication are as follows:
  • Replication requires no intermediate storage system for backups. Data is streamed directly from the source to the target.
  • Replication typically implies the use of a second Yellowbrick appliance that is used as the target system. However, local "loopback" replication is also supported and can be used to clone databases on the same system.
For details, see Database Replication.

Online Backups, Hot-Standby Restores

Yellowbrick supports online backups. Backup operations do not interfere with user queries or other database activities. Note that changes being introduced by running transactions that have not yet committed are not captured in a concurrently running backup. As soon as a transaction that changes the database commits, a subsequently started backup picks up those changes.

When you do an initial full restore of a database, the restored database is placed in "hot-standby" mode, allowing read queries and subsequent incremental restores but blocking other write operations. See HOT_STANDBY and READONLY Modes.

Backup Types

Yellowbrick supports backups at three different levels:
  • Full: backs up the entire database from scratch.
  • Cumulative: backs up all changes to the database since the last full or cumulative backup.
  • Incremental: backs up all changes to the database since the last full, cumulative, or incremental backup.
Important: Yellowbrick supports two specific strategies for scheduling a sequence of backups:
  • Full, followed by a series of cumulatives (no incrementals), and so on. For example:

  • Full, followed by a series of incrementals, then a cumulative, then another series of incrementals, and so on. For example:

Although other combinations of backups may work, Yellowbrick recommends these two strategies only. For more details and examples, see Backup Strategies.

Scope of Backups

A single ybbackup command backs up a single named database. Within a full backup, you can exclude one or more schemas and all of the objects they contain. You cannot exclude tables or other objects independently.

A database backup includes the following items (or changes to the following items):
  • All schemas (except where one or more schemas are explicitly excluded from full backups)
  • Tables and views (temporary tables are not backed up)
  • Data in tables
  • Stored procedures
  • Sequences
  • System catalog
  • Users and roles (optional)
  • Ownership and permissions on database objects, including granted privileges and role memberships (optional)
Note: Other shared (system-level) artifacts, such as WLM profiles, external table files, and NFS mounts for external tables, are not backed up. Table statistics are not backed up; they are recalculated and re-created when a database is restored.

Encryption keys for SQL access, as created with the CREATE KEY command, are also not backed up.

All backup files are saved in a proprietary format, and the data is not encrypted. Users cannot access these files directly, and restore operations work only when the target is a Yellowbrick appliance. If you need to encrypt backup data at rest, implement encryption within your storage system.

Backup Chains

A backup chain is a sequence of backups that are stored in the same backup location and belong to the same, single database. A backup location is a path to a target directory on a storage system. A chain of backups in a specific location is also known as a backup set or a backup bundle.

A full backup occurs by default the first time you back up a database under a new backup chain. When you take a new full backup of a given database in a given location, you break the existing backup chain and start a new one.

You can drop backup chains with a SQL command when they are no longer needed. You can also use the ybbackupctl tool to delete backups.

Backup Points or Snapshots

A backup point or snapshot represents a specific point in time for a database, with metadata that defines the last transaction that committed before the snapshot was created. This point in time serves as a reliable record of the database state at that time: which tables and views were present, how many rows they contained, and so on. When you back up a database with the ybbackup tool, a snapshot record is implicitly created in the system and is visible in the sys.backup_snapshot view.

Snapshots are also used as the basis for database replication.

WLM Resources for Backup and Restore Operations

Backup and restore operations are assigned to the large pool in the Default profile. In most cases, this pool provides adequate resources for intensive write operations. In general, Yellowbrick Data recommends that you do not attempt to redirect backup and restore operations to other resource pools. However, you can create your own pools and WLM rules to run these operations if the Default profile cannot manage them efficiently.