我的书签
添加书签
移除书签rt_merge_maxiosize
rt_merge_maxiosize
A maximum size of an I/O operation that the RT chunks merge thread is allowed to start. Optional, default is 0 (no limit). Added in 2.1.1-beta.
This directive lets you throttle down the I/O impact arising from
the OPTIMIZE statements. I/Os bigger than this limit will be
broken down into 2 or more I/Os, which will then be accounted as separate I/Os
with regards to the rt_merge_iops
limit. Thus, it is guaranteed that all the optimization activity will not
generate more than (rt_merge_iops * rt_merge_maxiosize) bytes of disk I/O
per second.
Example:
rt_merge_maxiosize = 1M
