Oracle say this is due to network problems but, when installing weblogic and grid control on a rac cluster we frequently experienced this error when using the scan addresses as the connect string.

I believe, but can not prove, the issue is based around the use of SCAN addresses for an 11.2 grid and 11.2 db when using an 11.1 Grid Control.

To get around this issue we modified the tnsnames.ora file to use all 3 physical addresses for the scan name and rebooted the systems (I should have stopped OMS and DB and restarted).

The SCAN addresses balanced themselves across the cluster and everything is up.