Ceph Cluster调优日志

在调优测试之前,先交代下Ceph集群环境:CentOS 7, Ceph Hammer 0.94.7, 3个节点,每个节点挂4块SAS 560G盘。

修改I/O Scheduler

CentOS 7默认的I/O Schduler是deadline, 所以不用修改。 使用命令查看:

[ceph@sh-cloud-test-ceph-4 ~]$ cat /sys/block/sda/queue/scheduler 
noop [deadline] cfq 

filestore_op_thread

[ceph@sh-cloud-test-ceph-4 ~]$ sudo ceph --show-config |grep filestore_op_thread
filestore_op_threads = 2

filestore_max_sync_interval/filestore_min_sync_interval

[ceph@sh-cloud-test-ceph-4 ~]$ sudo ceph --show-config |grep filestore_max_sync_interval
filestore_max_sync_interval = 5
[ceph@sh-cloud-test-ceph-4 ~]$ sudo ceph --show-config |grep filestore_min_sync_interval
filestore_min_sync_interval = 0.01

filestore_queue_max_ops/filestore_queue_max_bytes

[ceph@sh-cloud-test-ceph-4 ~]$ sudo ceph --show-config |grep filestore_queue_max_ops
filestore_queue_max_ops = 50
[ceph@sh-cloud-test-ceph-4 ~]$ sudo ceph --show-config |grep filestore_queue_max_bytes
filestore_queue_max_bytes = 104857600

filestore_queue_committing_max_ops/filestore_queue_committing_max_bytes

[ceph@sh-cloud-test-ceph-4 ~]$ sudo ceph --show-config |grep filestore_queue_commi
filestore_queue_committing_max_ops = 500
filestore_queue_committing_max_bytes = 104857600

journal_max_write_bytes/journal_max_write_entries

[ceph@sh-cloud-test-ceph-5 ~]$ sudo ceph --show-config|grep journal_max_write
journal_max_write_bytes = 10485760
journal_max_write_entries = 100

journal_queue_max_ops/journal_queue_max_bytes

[ceph@sh-cloud-test-ceph-5 ~]$ sudo ceph --show-config|grep journal_queue_max
journal_queue_max_ops = 300
journal_queue_max_bytes = 33554432

read_ahead_kb

sudo su -l root
echo "8192" > /sys/block/sda/queue/read_ahead_kb

参考

Ceph 性能调优

Ceph 参数性能调优

Table of Contents