Background The overall read and write performance of ceph has declined. After checking the ceph osd perf, there is a piece of osd with a large delay of more than 200 ms. After the decision to remove the overall performance recovery, it means that a node problem in osd sometimes affects the performance of the overall ceph.
ceph --admin-daemon /var/run/ceph/ceph-osd.135.asok perf dump | more
"WBThrottle": {
"bytes_dirtied": 13333504,
"bytes_wb": 0,
"ios_dirtied": 86,
"ios_wb" : 0,
"inodes_dirtied": 27,
"inodes_wb": 0
}, the
whole should be 0.
You can check with MegaCli to see if there are problems caused by bad sectors. Don't rush to restore the disk
长时间的数据积累对磁盘的性能和使用周期是有影响的 也可以定时清理磁盘碎片
查看磁盘碎片
xfs_db -c frag -r /dev/sdd1
整理碎片
xfs_fsr /dev/sdd1