All datasources within a dataservice have a specific role within the
dataservice. The Primary (master) role is
one that provides a source of replication information, and a Replica
(slave) one that receives that information.
| Role | Supplies Replication Data | Receives Replication Data | Load Balancing | Failover |
|---|---|---|---|---|
master
| Yes | No | Yes | Yes |
slave
| No | Yes | Yes | Yes |
relay
| Yes | Yes | Yes | Yes |
standby
| No | Yes | No | Yes |
archive
| No | Yes | Yes | No |
More detailed information for each role:
master (Primary)
A datasource in a master role
is providing a source for replication information to other datasources
in the dataservice and is able to provide both read and write
connections for applications.
slave (Replica)
A slave datasource is
receiving data from a master
and having that replicated data applied by Tungsten Cluster. Replicas are
used for read-only operations by applications.
A relay datasource receives
replication data, from upstream primary nodes in remote clusters, and
serves the downstream replicas within the same cluster.
This role is specific to the passive cluster within an Composite Active/Passive topology, or within the cross-site sub-services within Composite Active/Active and Composite Dynamic Active/Active topologies.
In the event of a failover between the active and passive cluster within
a Composite Active/Passive configuration, the relay
node will be promoted to the master role.
During a switch, the old master will become the
relay.
A standby datasource receives
replication data, but is never chosen by the connector to act as a
read source by application clients. Standby datasources are therefore
kept up to date with replication, but not used for load balancing.
When a failover occurs, a
standby datasource can be
enabled as a standard slave
and included in load-balanced operations.
An archive datasource can be
used to provide an active (up to date) copy of the data, without the
datasource being used in the event of a failover. This can be useful
for providing backup support, offline querying outside of the normal
dataservice operations, or auditing purposes.