When a new scan result is significantly different from the previous one, Discovery may create a new Host (see "Endpoint identity" in the documentation)

In the old Host node, the attribute "Endpoint Used To Be Identified As" will point to the new host.
In the new Host node, the attribute "Endpoint Now Identified As"  will point to the old host.

What happens in the CMDB in this case?

For a Spanish version of this KA see 000376313

A new BMC_ComputerSystem will be created in the Discovery CMDB dataset (i.e. BMC.ADDM) by synchronization to model the new Host. Other CIs will be attached to it (BMC_OperatingSystem, BMC_SoftwareServer, etc. )

If the new BMC_ComputerSystem CI matches an existing CI in BMC.ASSET (this depends on the CMDB Reconciliation Engine (RE) identification rule and the type of change between the old and the new CI):

• The next reconciliation job will try to identify the old and the new CI of BMC.ADDM with the same CI in BMC.ASSET.
• This will fail because BMC.ADDM already contains a CI identified with the one in BMC.ASSET (nothing will change). As a result, the new CI won't be merged into BMC.ASSET until the old CI in BMC.ADDM is deleted.

The expected RE errors are:

ARERR[120092] The dataset ID and Reconciliation Identity combination is not unique.
[...]
Found multiple matches (instances) for class <BMC.CORE:BMC_ComputerSystem>

If Discovery keep scanning the same ip, the old Host will start aging and when it will met the deletion condition (model maintenance):

• the old host will be deleted
• the matching CIs in BMC.ADDM will be marked as deleted
• The next reconciliation job will propagate the MarkAsDeleted flag into the BMC.ASSET CI and it will purge the Marked As Deleted CIs in the dataset BMC.ADDM.
• The next reconciliation job will allow the identification/merge of the new CI of BMC.ADDM with the one in BMC.ADDM.

 
If possible, it is recommended to let the Discovery deletion process and the reconciliation process delete the old CI and then promote the new one in BMC.ASSET.
 
If required, it is possible to accelerate the process of deletion of the "old CI" as follows:

1. Change the settings below in Administration > Model Maintenance

"Host/Network Device/Mainframe aging time limit"
"Host/Network Device/Mainframe aging access failures"

This settings are documented here.

Note these are global settings and reducing these values could lead to premature aging of hosts (before they are really gone)

OR

2. Identify, manually destroy, and synchronize the old Host nodes.
 

To identify them, use the report "Devices with Endpoint Identity Change" (see the Reports page in the UI).
 
Alternatively, list the hosts that are now identified as other hosts with the following query:

search Host where nodecount(traverse Previous:EndpointIdentity:Next:Host)

Select these Hosts and choose Actions > Destroy and then:

• • If case the CMDB Continuous Sync is not enabled then go to Actions > Sync with CMDB. This will cause the old CI to be marked as deleted in BMC.ADDM.
  • If it is enabled then it should immediately be updated in the CMDB

See also Discovery: How are CI's marked as deleted and removed in Discovery and CMDB ?