默认情况下，cephfs文件系统只配置一个活跃的mds进程。在大型系统中，为了扩展元数据性能，可以配置多个活跃的mds进程，此时他们会共同承担元数据负载。

要配置mds多活，只需要修改cephfs系统的max_mds参数即可。以下是未配置之前的集群状态

  1. # ceph -s
  2.   cluster:
  3.     id:     94e1228c-caba-4eb5-af86-259876a44c28
  4.     health: HEALTH_OK
  6.   services:
  7.     mon: 3 daemons, quorum test1,test2,test3
  8.     mgr: test1(active), standbys: test3, test2
  9.     mds: cephfs-2/2/1 up  {0=test2=up:active,1=test3=up:active}, 1 up:standby
  10.     osd: 18 osds: 18 up, 18 in
  11.     rgw: 3 daemons active
  13.   data:
  14.     pools:   8 pools, 400 pgs
  15.     objects: 305 objects, 3.04MiB
  16.     usage:   18.4GiB used, 7.84TiB / 7.86TiB avail
  17.     pgs:     400 active+clean
1、配置多活
  1. # ceph mds set max_mds 2
  2. # ceph -s
  3.   cluster:
  4.     id:     94e1228c-caba-4eb5-af86-259876a44c28
  5.     health: HEALTH_OK
  7.   services:
  8.     mon: 3 daemons, quorum test1,test2,test3
  9.     mgr: test1(active), standbys: test3, test2
  10.     mds: cephfs-2/2/2 up  {0=test2=up:active,1=test3=up:active}, 1 up:standby
  11.     osd: 18 osds: 18 up, 18 in
  12.     rgw: 3 daemons active
  14.   data:
  15.     pools:   8 pools, 400 pgs
  16.     objects: 305 objects, 3.04MiB
  17.     usage:   18.4GiB used, 7.84TiB / 7.86TiB avail
  18.     pgs:     400 active+clean
2、恢复单活mds
  1. # ceph mds set max_mds 1
  2. # ceph mds deactivate 1
  3. # ceph -s
  4.   cluster:
  5.     id:     94e1228c-caba-4eb5-af86-259876a44c28
  6.     health: HEALTH_OK
  8.   services:
  9.     mon: 3 daemons, quorum test1,test2,test3
  10.     mgr: test1(active), standbys: test3, test2
  11.     mds: cephfs-1/1/1 up  {0=test2=up:active}, 2 up:standby
  12.     osd: 18 osds: 18 up, 18 in
  13.     rgw: 3 daemons active
  15.   data:
  16.     pools:   8 pools, 400 pgs
  17.     objects: 305 objects, 3.04MiB
  18.     usage:   18.4GiB used, 7.84TiB / 7.86TiB avail
  19.     pgs:     400 active+clean
  21.   io:
  22.     client:   31.7KiB/s rd, 170B/s wr, 31op/s rd, 21op/s wr