一、 GlusterFS卷维护

1.1 GlusterFS的常用操作

卷的常用操作 命令
创建卷 gluster volume create
启动卷 gluster volume start
停止卷 gluster volume stop
删除卷 gluster volume delete
添加节点 gluster peer probe
删除节点 gluster peer detach
列出集群中的所有卷 gluster volume list
查看集群中的卷信息 gluster volume info
查看集群中的卷状态 gluster volume statu

1.2 扩容卷

先添主机到主机池中

  1. gluster peer  probe server3
  2. gluster peer  probe server4

创建目录

在server3和server4上创建目录

mkdir /data/glusterfs -p

扩容节点并增加复制份数

  • data是卷名
  1.  gluster volume add-brick data replica 2 server3:/data/glusterfs server4:/data/glusterfs

查看

  1. [root@server1 ~]# gluster volume info data
  3. Volume Name: data
  4. Type: Distributed-Replicate
  5. Volume ID: a99e0ecc-132d-4778-8cc8-36b9a2c04ac4
  6. Status: Started
  7. Snapshot Count: 0
  8. Number of Bricks: 2 x 2 = 4
  9. Transport-type: tcp
  10. Bricks:
  11. Brick1: server1:/data/glusterfs
  12. Brick2: server2:/data/glusterfs
  13. Brick3: server3:/data/glusterfs
  14. Brick4: server4:/data/glusterfs
  15. Options Reconfigured:
  16. transport.address-family: inet
  17. nfs.disable: on

1.3 收缩节点

执行缩卷命令

  • data是卷名
  1. gluster volume remove-brick  data server3:/data/glusterfs server4:/data/glusterfs start

查看迁移状态

  1. [root@server1 ~]# gluster volume remove-brick data server3:/data/glusterfs server4:/data/glusterfs  status
  2.                                     Node Rebalanced-files          size       scanned      failures       skipped               status  run time in h:m:s
  3.                                ---------      -----------   -----------   -----------   -----------   -----------         ------------     --------------
  4.                                  server3                0        0Bytes             0             0             0            completed        0:00:00
  5.                                  server4                0        0Bytes             0             0             0            completed        0:00:00
  • 当status状态显示completed时就表示好了

提交

  1. gluster volume remove-brick data server3:/data/glusterfs server4:/data/glusterfs  commit

查看状态

  1. [root@server1 ~]# gluster volume info
  3. Volume Name: data
  4. Type: Replicate
  5. Volume ID: a99e0ecc-132d-4778-8cc8-36b9a2c04ac4
  6. Status: Started
  7. Snapshot Count: 0
  8. Number of Bricks: 1 x 2 = 2
  9. Transport-type: tcp
  10. Bricks:
  11. Brick1: server1:/data/glusterfs
  12. Brick2: server2:/data/glusterfs
  13. Options Reconfigured:
  14. transport.address-family: inet
  15. nfs.disable: on

1.4 GlusterFS目录迁移

创建迁移目录

mkdir /data/backup -p

迁移

gluster volume replace-brick data server1:/data/glusterfs server1:/data/backup commit force

查看结果

  1. [root@server1 ~]# gluster volume  info
  3. Volume Name: data
  4. Type: Distributed-Replicate
  5. Volume ID: a99e0ecc-132d-4778-8cc8-36b9a2c04ac4
  6. Status: Started
  7. Snapshot Count: 0
  8. Number of Bricks: 2 x 2 = 4
  9. Transport-type: tcp
  10. Bricks:
  11. Brick1: server1:/data/backup
  12. Brick2: server2:/data/glusterfs
  13. Brick3: server3:/data/glusterfs
  14. Brick4: server4:/data/glusterfs
  15. Options Reconfigured:
  16. nfs.disable: on
  17. transport.address-family: inet

二、 GlusterFS常用的优化

2.1 GlusterFS常用优化参数

配置选项 用途 默认值 合法值
network.ping-timeout 客户端等待检查服务器是否响应的持续时间，节点挂了数据不能写入 42 0-42
Performance.cache-refresh-timeout 缓存校验周期 1s 0-61
auth.allow 允许访问卷的客户端ip
auth.reject 拒接访问卷的客户端ip
performance.cache-size 读取缓存的大小 32 MB
performance.write-behind-window-size 能提高写性能单个文件后写缓冲区的大小 1MB
performance.io-thread-count IO操作的最大线程 16 0-65

官方参数

  • 修改语句的命令格式 gluster volume set <VOLNAME>

  • 例如

GlusterFS集群会有一个检测时间默认是42s，如果一个节点挂了或网络不通了，整个集群都会在这42秒内无法写入数据，所以通常会修改这个检测时间。

  • 把它改成10s
  1. [root@server1 ~]# gluster volume set data  network.ping-timeout 10
  2. volume set: success

原文: https://www.rainbond.com/docs/stable/operation-manual/storage/GlusterFS/maintain.html