I have a RAC cluster which had 4 nodes, and another RAC cluster with 4 nodes in a remote data center for DR purposes, and Data Guard is in use for replication between these clusters.
There are multiple databases involved, The Grid layer is 11i as are the databases.
A given database may be primary on either cluster at a given time, and the databases do not all switch over together normally (different applications).
The 4th node in the RAC cluster crashed and became unavailable. Services are relocated and database instances are running on available nodes appropriately, but we won't be getting that 4th node back. (server is hosed and we are in a process of migrating off of this cluster so we won't be reprovisioning it).
I see that this leaves the instance in the Grid/RAC definitions and that it should be removed. I see how to remove a node from specific service configurations but I am unclear about how to remove the 4th RAC instance from each database when the instance is down, the node is down and they can't be made available. Procedures I found using dbca require the node/instance to be up, what would be the right process when the instance/node are down and can't be run? What is the impact on Data Guard (if any) for the unavailable instance which is associated with a thread#?
So far this has not caused an issue in the cluster or for the database itself, but it prevents operation of a GoldenGate replication where one of these RAC databases is the source. I can fix that if the RAC database has the right number of threads but this is production so I don't want to just guess.
Thanks
Sandra