control_file_record_keep_time参数的意义

星光酷表

zergduan 发表于 2018-2-6 10:30
1. 你这个跟我说的没任何关系，我是说controlfile和datafile不同，controlfile不能指定size，控制文件创 ...

分配给控制文件的裸设备大小是最大的大小，这个是一定的了，
控制文件使用的大小也有类似于表的高水位线（就是以前使用的block的最大值），当控制文件保存的内容超过这个"高水位线"如果设置了control_file_record_keep_time会继续扩展"高水位线",如果设置的较小，会覆盖之前的内容，是这么理解吧？
差不多，你可以这么理解，但是不准确。 并不存在HWM的概念，HWM是在一个已经allocate的空间内部，而不是在空间的边界上~

Yong Huang

Zergduan clearly explains that the size of the raw device where you place the controlfile is irrelevant; it just sets an upper limit to the growth of the controlfile. Nowadays, though, nobody uses raw devices directly. The most common practice in industry is to use ASM, which is essentially raw devices but has an excellent management layer.

Back to the real question. guocun09's message #8 is the best explanation. There's no high water mark inside the controlfile. The rule is simple. If a new record must be inserted into the controlfile and there's no space and the oldest record is not old enough compared to control_file_record_keep_time, then the controlfile is expanded. If none of the conditions is met, the file is not expanded. Note that the controlfile never shrinks. Suppose it's expanded for the above reasons, due to heavy activity of the database and therefore lots of archive logs and RMAN backups in a short time. Later, when the database becomes quiet for months, the oldest record in the controlfile will never be newer than control_file_record_keep_time and so will always be overwritten by new records. The controlfile becomes stable in size.

星光酷表

Yong Huang 发表于 2018-2-7 23:14
Zergduan clearly explains that the size of the raw device where you place the controlfile is irrelev ...

在RMAN恢复中，如果可以停机恢复，catalog start with是否可以完全取代control_file_record_keep_time，不需要在单独设置control_file_record_keep_time的天数