Understanding and Designing Serviceguard Disaster Recovery Architectures

Terms and Concepts
Continentalclusters
Continentalclusters is configured using two or more Serviceguard clusters. The packages in different
Serviceguard clusters have a special primary-recovery relationship between them. This relation is
defined by the user in Continentalclusters configuration and is enforced by the Continentalclusters
software.
Primary and Recovery Clusters
A Serviceguard cluster that has a package in the recovery role is called the recovery cluster or the
monitoring cluster and a Serviceguard cluster that has a package in the primary role is called the
primary cluster or the monitored cluster. In case of a mutual recovery configuration, each cluster
acts as the monitoring cluster for the other.
Recovery Group
The packages that have the primary-recovery relationship is called a “Recovery Group“ in
Continentalclusters. A “Recovery Group” can optionally have a rehearsal package, a data sender
package or a data receiver package in addition to the mandatory primary and recovery packages.
The rehearsal package is used for disaster recovery rehearsal operations. The data sender and
data receiver packages are used if primary application data is replicated to the recovery data
center using software replication. A maximum of 75 Recovery Groups can be configured in a
Continentalclusters configuration. The primary or the recovery package can be a failover package
or an multi-node package or a site controller package managing a complex workload.
Monitoring Package
A cluster that has a package in a recovery role needs a monitoring package. The configuration
for the monitoring package is provided with the Continentalclusters software. The monitoring
package is responsible for periodically checking the health of the cluster running the primary
package.
Cluster Events and notifications
Continentalclusters provides Cluster Events and the notification feature to ensure information from
an independent source can be gathered before the recovery process starts. The cluster events and
their notifications are user configurable.
A cluster event occurs when the monitored cluster stays in a given status for a user defined period
of time. After the occurrence of a cluster event, the set of notifications associated with that cluster
event is issued by Continentalclusters.
For example, if the primary cluster is unreachable for 10 minutes it can be considered as a Cluster
Event. If an email notification is configured, the administrator receives an email when the primary
cluster is unreachable for 10 minutes.
Notifications Methods
Notifications are user configurable and configured for each cluster event separately. The following
types of notifications are supported:
• Email messages
• SNMP traps
• TCP/UDP port message delivery
• Text log files
• OPC messages to OpenView IT/Operations
28 Metrocluster and Continentalclusters