An example of a database that encountered in work: Operating system: Sun Solaris 8 Duplex Software: Sun Cluster 3.0 Database: Oracle 8.1.7.3 Fault Phenomenon: Database cannot start or CLUSTER system cannot switch. Check database log file $ ORACLE_BASE / admin / ORCL / bdump / alert_ORCL.log have similar ORA-01599: failed to acquire rollback segment (49), cache space is full (currently has (48) entries) Error 1599 happened during db open SHUTTING DOWN DATABASE. Fault Analysis: ORE has an initialization parameter MAX_ROLLBACK_SEGMENTS which represents the number of returns that can be online. Its default is Max (30, Transactions / Transactions_Per_rollback_segment). When the initialization parameter is specified (via the rollback_segments parameter) or the number of rollback segments of the default Online exceeds max_rollback_segments, the database cannot be started after this parameter. This error is often due to the user after adding the rollback segment but does not modify the max_rollback_segments parameter. Alternatively, the user adds the max_rollback_segments parameter in the initialization parameter. It will also cause this issue. Troubleshooting: You can change the max_rollback_segments in the Oracle Initialization Parameters to greater than the total number of all current returns. If you don't know how many rollbacks are currently now, you can set this value to 100 to restart the database. If the same error is also reported, this value is increased until success. When the database starts, you can get the exact rollback segment by select count (*) from DBA_ROLLBACK_SEGS, then modify the initialization parameters. Daily maintenance: Please pay attention to the interaction between max_rollback_segments and rollback_segments when adding a rollover segment, it is best to see it on both nodes.
