GATHER_FULL_STATS_JOB oracle自动收集统计信息 options =＞ ‘gather‘

并行参数最好指定

Applies To

All Users

Summary

• High load is associated with stats jobs for each schema that run from the scheduler
• Job is running at the same time as the staggering of start times is not large enough
• Each schema has a job to connect full stats as follows:
  GATHER_FULL_STATS_JOB
  begin dbms_stats.gather_schema_stats (ownname => '',
  method_opt => 'for all columns size auto',
  estimate_percent => 100,
  cascade => true,
  degree=> DBMS_STATS.DEFAULT_DEGREE,
  options => 'gather'); end;
• There are many queries in AWR showing following:
  select /*+ parallel_index(t, "_IDX1", 384) dbms_stats cursor_sharing_exact use_weak_name_resl dynamic_sampling(0) no_monitoring xmlindex_sel_idx_tbl opt_param('optimizer_inmemory_aware' 'false') no_substrb_pad no_expand index_ffs(t, "_IDX1") */ count(*) as nrw, count(distinct sys_op_lbid(364661, 'L', t.rowid)) as nlb, count(distinct "SYS_NC00015$") as ndk, sys_op_countchg(substrb(t.rowid, 1, 15), 1) as clf from...

Solution

This is an expected feature.

By design, when using default_degree option, it is limited to current environment settings. This option is fully controlled by the number of CPUs and initialize parameters and the size of the object that are collecting statistics for.

The 384 value in general is dynamic, so the full value is usually not used.

Attachments :

Cause

Degree=> DBMS_STATS.DEFAULT_DEGREE translates inside the code to ===> GATHER_TABLE_STATS(... degree => 32768)