This error is caused by a bug recognized by Oracle that can be easily found in Metalink [ID 4483084.8], the error indicates that you have tried to stop the database without the Library Cache was empty.
The proposed solution is officially migrate to 188.8.131.52 and further indicates that it does not produce any kind of corruption in the datafiles, suggesting that we can get used to seeing on the alert without worrying.
What they do not say is that this error may generate a typical shutdown immediate, may fail and be the BD in nomount state, this has happened to me in HP-UX systems. If it happens during a cold weekend back, the database will not start because it is, but of course in nomunt state.
To solve this problem and have a clean alert can add a trigger system:
CREATE OR REPLACE TRIGGER flush_shared_pool BEFORE SHUTDOWN ON DATABASE BEGIN execute immediate 'ALTER SYSTEM FLUSH SHARED_POOL'; EXCEPTION WHEN OTHERS THEN RAISE_APPLICATION_ERROR (num => -20000, msg => 'Error flushing pool'); END;
To me it has worked in BD installed on HP-UX, Linux and Windows.