This condition may occur when one of the nodes stays down for a significant time period, in which configuration changes are made to the active node. Obviously, these changes are not available to the failed node. Now, if the primary node shuts down for some reason and the previously downed node is brought up alone in the cluster with an old database revision, there is a likely chance of configuration corruption. This condition is known as partition in time.

To resolve this split brain condition, each Fabric Interconnect writes a database version number to its section of EEPROM. Now when a Fabric Interconnect is brought alone in the cluster, it can still read the EEPROM database version and compare it to its own information. If the version number is the same or higher, the FI can become the active member and if the version number is lower, the FI does not become the active member. This mechanism thus protects against applying an older version of the UCS configuration.