Oracle Database / Data Guard & High Availability
Check service placement when only one RAC endpoint fails
An application can work through one cluster member while a service or instance problem affects another. Collect database-side service and instance context.
Environment and inputs: Oracle Database 19c baseline · SQL*Plus-compatible SQL unless RMAN, DGMGRL, ADRCI or a host shell is specified. Check release, patch level, edition and container first.
Start with this diagnostic
Use the stated SQL, RMAN, DGMGRL or host shell and replace sample names with approved targets. DBA and V$ views need authorized catalog access in the correct container. Do not grant broad privileges for an example.
SELECT inst_id, name, con_id FROM gv$active_services ORDER BY name, inst_id;Read the result
The view shows active services visible across instances. It does not replace Clusterware configuration or prove every SCAN and network path is usable.
Resolve and verify
Compare active placement with the approved service policy, client descriptor and Clusterware status. Repair the managed service or network path through the cluster owner.
Operational precautions
Do not pin clients to a random instance address and bypass managed failover as a permanent fix.