Add a RAC service with preferred/available instances (including Application Continuity)

Command

srvctl add service -db <db_unique_name> -service <service> \
  -preferred <inst1>,<inst2> -available <inst3> \
  -pdb <pdb> -role PRIMARY -policy AUTOMATIC \
  -failovertype TRANSACTION -commit_outcome TRUE \
  -failoverretry 30 -failoverdelay 10 \
  -clbgoal SHORT -rlbgoal SERVICE_TIME -notification TRUE \
  -drain_timeout 300 -stopoption IMMEDIATE
srvctl start service -db <db_unique_name> -service <service>
# Verify:
srvctl config service -db <db_unique_name> -service <service>
srvctl status service -db <db_unique_name> -service <service>

What it does

Defines a service for applications that normally runs on the preferred instances and moves to an available instance when they fail; with -role PRIMARY it starts only in the primary role under Data Guard.

When to use it

To connect applications to a service rather than an instance, to partition workloads across nodes, and to enable seamless failover (drain/failover) during planned maintenance.

Caution

Applications must connect with this service name (through SCAN), not with a SID or instance name. -failovertype TRANSACTION (Application Continuity) requires application/driver support; if unsure, start with -failovertype SELECT (TAF) or NONE. Do not use SYS$ services or the default service named after db_unique_name for applications.