Failover and Data Recovery in the Core Database

Core Database Failover in a Single Data Center HA Configuration

Refer to Configuring the Core Database for HA to learn more about a single data center HA configuration.

Core Database Failover in a Multiple Data Center HA Configuration

Disaster Recovery (DR) architecture varies greatly from data center to data center. In a strategy where all systems are fully replicated to a DR facility, you may be able to architect the Jive Platform as you would with a Single Data Center HA Configuration. In many cases, however, the DR strategy is more manual and requires a Multiple Data Center HA Configuration.

The following represents two possible DR strategies.

Create Backup at Remote Location (Simplest DR Strategy)

If a recent full backup of the database is available at a remote location, it is possible to recover the system to the point in time of the available backup. Upon declaration of a disaster, perform the following:

1. Set up a new cluster of application servers and point them to a new empty database at the recovery facility.
2. After completion of initial setup, save specific property values from the new database's jiveProperty table somewhere or copy them to a backup table. For more on which properties need to persist, see Persistent Properties for a Database Recovery.
3. Restore the database backup over the newly created database.
4. Apply the properties that should persist (as determined in Step 2) from the new database to the restored database's jiveProperty table. For more on this, see Persistent Properties for a Database Recovery.
5. Restart the application server.

Note: Data loss in the event of a disaster can be minimized by pushing incremental backups, and/or transaction \ write-ahead logs to the remote facility. Point-in-time recovery may be performed up to the point of the disaster if backups are available.

Use Cold/Warm State Servers

Streaming replication, or Write-Ahead Logging (WAL)/log shipping, maintains a recent copy of the database at the remote facility. You need to make sure a cluster of application servers are already set up and attached to an empty database. Upon declaration of disaster, the replicated database should replace the empty database with specific values in jiveProperty persisted to reflect the DR environment. For more on this, see Persistent Properties for a Database Recovery.

It is important to consider which properties should be replaced with values from the original production site, and which values should reflect values of the new facility. The persisted values depend on system configuration, as well as any customizations that could impact the jiveProperty table. Review, validation, and live testing of a system failover will eliminate any potential issues that could arise during an actual disaster.