This error has been appearing on a radom base at the alert.log file. There is no apparent reason for this to appear. And when I take a look at the detailes provided on the trace file the only thing I see is the same error meaningless pattern, the connection descriptor was not found.
In this particular case the contents at the alert.log file showed this information:
Tue Mar 09 15:45:43 2010
Errors in file ...\trace\orapro_reco_4892.trc:
ORA-02019: connection description for remote database not found
From the generated trace file name, it states that it is the RECO background process the one responsible for this trace. RECO is used in distributed transactions, and it happens that when a database goes down RECO tries to resend the pending transactions according to the Two Phase Commit procedure.In this case there it used to be a database attached to the main database, but it was gone long ago, however some non applied transactions remained in the transaction queue, making RECO to keep on retrying forever. The database was deconfigured and all the related files and file descriptors were removed.
By taking a look at the DBA_2PC_PENDING view there there were a couple of transactions that remained in the queue and which were never applied.
SELECT LOCAL_TRAN_ID,GLOBAL_TRAN_ID,STATE,MIXED,COMMIT#
FROM DBA_2PC_PENDING;
LOCAL_TRAN_ID GLOBAL_TRAN_ID STATE MIXED COMMIT#
-------------------------------------------------------------------------------------------------------------------------------------------------------------------
9.21.102540 ORCL.ee36125b.9.21.102540 collecting no 9462810864
8.32.86528 ORCL.ee36125b.8.32.86528 collecting no 9462814781
In this case, and since there is no target database to apply the transactions to, then it is just enough to get rid of them by means of the PURGE_LOST_DB_ENTRY procedure:
execute DBMS_TRANSACTION.PURGE_LOST_DB_ENTRY('8.32.86528');
execute DBMS_TRANSACTION.PURGE_LOST_DB_ENTRY('9.21.102540');
commit;
That's it, this way the ORA-02019 error is gone.
No comments:
Post a Comment