|
|
|
|
|
|
by jeffbee
1917 days ago
|
|
|
Yeah, you _really_ have to watch out for this. I once spent months chasing down a serious but rare performance problem in a large-scale mysql 5.6 deployment, which was eventually root-caused to kernel slab reclaim pressure, caused by very high XFS metadata mutation rate, caused by MySQL creating an on-disk temporary ISAM file for every query with a union, which was most of the traffic. |
|
|