|
|
最初由 wlidflower 发布
[B]"验证并不难,两个10046 ,level 12 的trace + tkprof 基本差不多了。
如果不嫌麻烦,做个statspack,对比一下,也很容易看出差别。
总而言之,因为archive log产生的额外的IO应该就是本例的罪魁祸首。"
当然是ARCHIVE搞的鬼
可能我忽略了一个细节,我的是RAC环境,但是我的操作在一个节点上做
不知道是不是RAC环境对这个也有影响,大家不知道发现没有,在RAC环境下,仅仅在一个节点进行操作,另外一个节点的REDO也会发生切换,同时也会生产归档日志,如果节点2没开,那么这些归档日志会写在1这个节点上
我已经发了相关的帖子, WING HONG斑竹也给了点已经,大家可以去关注一下
我的是RAID5,做完RMAN以后拷贝出来进行的备份 [/B]
Archive logs for an instance may be archived by another active instance in the
following situations:
1. The 'alter system archive log all' command was issued manually without
specifying the thread parameter. Manual archiving is performed by the user
process that issues the archiving command; it is not performed by the
instance's ARCH process. The redo log files are archived to the destination
of the instance to which the user is connected.
2. Parallel server can force a log switch for an instance that fails to
archive its online redo log files for some period of time, either because
the instance has not generated many redo entries or because the instance
has shutdown. This prevents a thread from remaining unarchived for too
long. After an instance has shutdown, another instance can force a log
switch for that instance so that its current redo log can be archived.
Either situation can result in archive logs being written to the destination
of the active instance. |
|