A, 11g database version initial setup parameters to be adjusted:
parameter | Setting the standard | default | There is a risk |
db_files | 1000 | 200 | The maximum allowed number of files in the database, the initial value is 200, the number can not be increased too small will result in the data file |
processes | 1500 | Database maximum number of processes, the number of databases can effectively prevent the internal processes enough to affect the normal operation of the business | |
sessions | 2272 | ||
undo_retention | 3600 | 900 | undo recovery period of time, in seconds. Achieve recovery time period is about to undo the rollback set to expire, preventing undo segment can not be recycled |
_undo_autotune | FALSE | TRUE | Whether to enable automatic optimization of the database table spaces undo function, belonging oracle implicit parameter is not enabled automatic optimization function to prevent undo automatic database recovery, affecting the normal operation of the business |
event | 28401 | Using the wrong password attempts will lead to a high Library Cache Locks or row cache lock | |
audit_trail | db,extended | Control audit database storage location, stored in the file system directory, in order to maintain the daily operation and maintenance. Prevent audit directories full. | |
_use_adaptive_log_file_sync | FALSE | TRUE | Implicit parameter is set by _use_adaptive_log_file_sync, disable the automatic switching mode when the value is FALSE. |
_serial_direct_read | NEVER | AUTO | Can be significantly reduced direct_path_read |
_ktb_debug_flags | 8 | 0 | Avoid BUG, datagaurd 11g of the bug, during switching, it can cause bad block index, the need to set parameters or patch 22241601 |
_gby_hash_aggregation_enabled | FALSE | TRUE | Avoid BUG, if the hash table data as large as a threshold, there will be serious upgrade table space [bug] |
job_queue_processes | 20 | 1000 | The default value is too high, it can lead to the problem of high CPU load, set this parameter limits the number of simultaneously launched the largest JOB |
session_cached_cursors | 200 | 50 | A single session may be cached in the number of the cursor, the cursor cache appropriate, to enhance the resolving power soft |
_smu_debug_mode | 134217728 | 0 | There will be some performance problems and BUG (Note: killMMOM process does not terminate example, the main process AWR, then kill a new MMON process will automatically start using _smu_debug_mode = 134217728) |
_clusterwide_global_transactions | FALSE | TRUE | When _clusterwide_global_transactions = false when, Oracle will be the local transaction as a single transaction commits coordinated by a multi-stage process |
_PX_use_large_pool | TRUE | It may interchange data and information with the parallel implementation dependent processes, a fixed allocation of memory from largepool | |
_ges_direct_free_res_type | CTARAHDXBB | Share_pool prevent high memory overhead | |
_drop_stat_segment | 1 | Truncate improve efficiency | |
sql92_security | TRUE | FALSE | SELECT privilege checks if the current user is provided with a corresponding table is sql92_security When set to TRUE, performs UPDATE / DELETE operations on the table |
enable_ddl_logging | TRUE | FALSE | Parameter is set to TRUE, the following DDL statements can be recorded in the alert log |
deferred_segment_creation | FALSE | TRUE | Avoid delays create section |
_resource_manager_always_on | FALSE | TRUE | Disabling Oracle resource scheduling enabled by default, to avoid possible resmgr: cpu quantum waiting circumstances of the incident |
_resource_manager_always_off | TRUE | FALSE | Disabling Oracle resource scheduling enabled by default, to avoid possible resmgr: cpu quantum waiting circumstances of the incident |
_mv_refresh_use_stats | TRUE | FALSE | Materialized view fast refresh performance issues may sometimes occur |
_memory_imm_mode_without_autosga | FALSE | TRUE | AMM use, is not provided ; AMM if closed, set to FALSE, to avoid ORA-4031 |
_bloom_filter_enabled | FALSE | TRUE | 11R2会遇到一个BLOOM过滤器导致的BUG 9124206和BUG 8361126,出现ORA-00060 ORA-10387错误, _bloom_pruning_enabled、_bloom_filter_enabled均设为FALSE避免BUG |
_optimizer_use_feedback | FALSE | TRUE | 关闭Cardinality Feedback新特性动能 |
_cleanup_rollback_entries | 2000 | 100 | 加大该参数来达到加快串行事务恢复的效果,(同时设置并行恢复fast_start_parallel_rollback为high) |
_datafile_write_errors_crash_instance | FALSE | TRUE | 11.2.0.2/3数据文件(sysytem以外表空间)I/O读写错误被发现时,实例down |
_gc_defer_time | 3 | 0 | 用于确定服务器在将频繁使用的块写入磁盘之前要等待的时间长度 (以 1/1000 秒为单位),以减少进程对热块的争用 |
_gc_policy_time | 0 | 10 | DRM在11G中不稳定,存在众多BUG |
_gc_read_mostly_locking | FALSE | TRUE | DRM在11G中不稳定,存在众多BUG |
_gc_undo_affinity | FALSE | TRUE | DRM在11G中不稳定,存在众多BUG |
disk_asynch_io | TRUE | TRUE | 使用文件系统存放数据文件时,建议开启异步I/O |
filesystemio_options | asynch | none | 使用文件系统存放数据文件时,建议开启异步I/O(使用ASM存储,默认开启异步I/O) |
_optimizer_adaptive_cursor_sharing | FALSE | TRUE | 关闭ACS自适应游标共享,11.2.0.4可不关闭 |
_optimizer_extended_cursor_sharing | none | UDO | 关闭ACS自适应游标共享,11.2.0.4可不关闭 |
_optimizer_extended_cursor_sharing_rel | none | SIMPLE | 关闭ACS自适应游标共享,11.2.0.4可不关闭 |
二、12c版本初始安装数据库需要调整的参数:
参数 | 建议设置标准 | 默认 | 存在风险 |
_optimizer_aggr_groupby_elim | FALSE | 在进行某些聚合运算的时候,由于bug 21826068的存在,导致出现错误结果(Doc ID 21826068.8) for 12.1.0.2 | |
_optimizer_unnest_scalar_sq | FALSE | 在进行某些标量子查询的时候,报错ORA-600[kkqcsfixfro:1--frooutj] (Doc ID 19894622.8) for 12.1.0.2 | |
_optimizer_reduce_groupby_key | FALSE | 带绑定变量的外连接,进行group by运算的时候,导致错误结果 12.1.0.2 bug 20634449 | |
_kks_obsolete_dump_threshold | 0 | 并发session的状态dump在12c中得到增强,在dump的时候,会影响性能(Doc ID 1955319.1) for 12.1.0.2 | |
_sql_plan_directive_mgmt_control | 0 | 贴近12.2的优化器设置,防止过度的动态采样(Doc ID 2187449.1) for 12.1.0.2 | |
_optimizer_dsdir_usage_control | 0 | 贴近12.2的优化器设置,防止过度的动态采样(Doc ID 2187449.1) for 12.1.0.2 | |
_optimizer_cbqt_or_expansion | off | Delete 语句在 12.2.0.1 上 Hang住Bug 26693987 | |
_cursor_obsolete_threshold | 1024 | 8192 | 高版本数(>1024)的SQL语句在升级到12.2及更高版本后会导致数据库性能下降 11.2.0.3: _cursor_obsolete_threshold=100 11.2.0.4: _cursor_obsolete_threshold=1024 12.1: _cursor_obsolete_threshold=1024 12.2: _cursor_obsolete_threshold=8192 |