Oracle Corner

When you use OEM to create a standby on RAC, it will create a single instance on one standby node, which you later need to convert to a clustered database.

In my example that instance is called PPEN (playpen).

I converted PPEN to a clustered database, and I could see 2 instances: PPEN1 and PPEN2. Previous "PPEN" instance has been automatically removed from the cluster node, and I removed it manually from OEM. The problem is that it wasn't removed within Data Guard configuration.

Symptoms

When you go to OEM, and chose 'Availability'/'Verify Database Data Guard Configuration', you will get warnings:

Initializing
Connected to instance host2:PPEN2
Starting alert log monitor...
Updating Data Guard link on database homepage...
WARNING: Cluster database instance PPEN is not discovered.
WARNING: Switchover or failover may not succeed as a result.
Skipping verification of fast-start failover static services check.

Data Protection Settings:
  Protection mode : Maximum Performance
  Redo Transport Mode settings:
    PPEN: ASYNC
    PPENDR_A: ASYNC
... etc...

Solution

Fix is to delete the initial PPEN instance from Data Guard configuration.

DGMGRL>  show database verbose "PPENDR_A";
Database - PPENDR_A

  Role:               PHYSICAL STANDBY
  Intended State:     APPLY-ON
  Transport Lag:      0 seconds (computed 0 seconds ago)
  Apply Lag:          0 seconds (computed 0 seconds ago)
  Average Apply Rate: 4.00 KByte/s
  Active Apply Rate:  781.00 KByte/s
  Maximum Apply Rate: 1.01 MByte/s
  Real Time Query:    OFF
  Instance(s):
    PPEN  -- this instance shouldn't be here
    PPEN1 (apply instance)
    PPEN2

DGMGRL> remove instance "PPEN" on database "PPENDR_A";

This looks like a bug, but search through Oracle support didn't show it, and google-ing didn't help me find this warning, so hopefully this helps other DBAs.

Labels: Verify Database Data Guard Configuration, WARNING: Cluster database instance is not discovered., WARNING: Switchover or failover may not succeed as a result.