| DATABASE ADMINISTRATION |
REMOTE DBA
HOT-STANDBY
DISASTER RECOVERY
Bugchecking on Fast Systems
TECHNICAL
ARTICLES -> TROUBLESHOOTING
ARTICLES
Q: |
Since upgrading our
system to faster hardware, we have been experiencing periodic
bugchecks dumps while attempting to bind to the database. We are
running a very old version of DBMS (4.2a). This
happens infrequently, but we haven't been able to identify a reason.
|
A: |
The problem you are experiencing has to do with an algorithm that was used in older versions of Rdb and DBMS to generate the name of some data structures that are mapped into global sections during the bind process. Rdb and DBMS used the VMS system service, $GETTIM, to generate "unique" identifiers in older versions. However, on faster processors, this routine can be executed multiple times within the same VMS time interval (100 nanoseconds). Thus, this algorithm experienced periodic failures on the faster processors. The algorithm was changed in newer versions of Rdb and DBMS to avoid this problem. If you are running on the latest AlphaServer systems, SCI recommends upgrading to recent versions of Oracle Rdb and Oracle CODASYL DBMS – these have been thoroughly qualified to eliminate these "timing" issues. |