ceph ( requests are blocked ) 异常解决方法

最近在执行 ceph 扩容
注: 如果有条件, 建议不要扩容, 直接创建新 POOL, 这样会避免很多异常与影响

扩容每天大约 2T 空间, 扩容过程中, pg 会对数据执行自动迁移, 但在迁移过程中会遇到下面异常错误

[root@hh-yun-puppet-129021 ~]# ceph -s
    cluster dc4f91c1-8792-4948-b68f-2fcea75f53b9
     health HEALTH_WARN 1 requests are blocked > 32 sec
     monmap e3: 5 mons at {hh-yun-ceph-cinder015-128055=240.30.128.55:6789/0,hh-yun-ceph-cinder017-128057=240.30.128.57:6789/0,hh-yun-ceph-cinder024-128074=240.30.128.74:6789/0,hh-yun-ceph-cinder025-128075=240.30.128.75:6789/0,hh-yun-ceph-cinder026-128076=240.30.128.76:6789/0}, election epoch 216, quorum 0,1,2,3,4 hh-yun-ceph-cinder015-128055,hh-yun-ceph-cinder017-128057,hh-yun-ceph-cinder024-128074,hh-yun-ceph-cinder025-128075,hh-yun-ceph-cinder026-128076
     osdmap e97975: 190 osds: 190 up, 190 in
      pgmap v13666786: 20544 pgs, 2 pools, 77479 GB data, 19508 kobjects
            228 TB used, 426 TB / 654 TB avail
               20542 active+clean
                   2 active+clean+scrubbing+deep
  client io 47657 kB/s rd, 164 MB/s wr, 5406 op/s

注意: 1 requests are blocked > 32 sec 有可能是在数据迁移过程中, 用户正在对该数据块进行访问, 但访问还没有完成, 数据就迁移到别的 OSD 中, 那么就会导致有请求被 block, 对用户也是有影响的

解决方法
寻找 block 的请求

[root@hh-yun-puppet-129021 ~]# ceph health detail
HEALTH_WARN 1 requests are blocked > 32 sec; 1 osds have slow requests
1 ops are blocked > 33554.4 sec
1 ops are blocked > 33554.4 sec on osd.16
1 osds have slow requests

可以看到 osd.16 具有一个操作 block
解决方法
查询 osd 对应主机

[root@hh-yun-puppet-129021 ~]# ceph osd tree
# id weight type name  up/down reweight
-1  598 root default
-2  40     host hh-yun-ceph-cinder015-128055
0   4      osd.0 up 1
1   4      osd.1 up 1
2   4      osd.2 up 1
3   4      osd.3 up 1
4   4      osd.4 up 1
5   4      osd.5 up 1
6   4      osd.6 up 1
7   4      osd.7 up 1
8   4      osd.8 up 1
9   4      osd.9 up 1
-3  40     host hh-yun-ceph-cinder016-128056
10  4      osd.10 up   1
11  4      osd.11 up   1
12  4      osd.12 up   1
13  4      osd.13 up   1
14  4      osd.14 up   1
15  4      osd.15 up   1
16  4      osd.16 up   1
17  4      osd.17 up   1
下面省略

重启 osd

[root@hh-yun-ceph-cinder016-128056 ~]# /etc/init.d/ceph stop osd.16
=== osd.16 ===
Stopping Ceph osd.16 on hh-yun-ceph-cinder016-128056...kill 2799859...kill 2799859...done
[root@hh-yun-ceph-cinder016-128056 ~]# /etc/init.d/ceph start osd.16
=== osd.16 ===
create-or-move updated item name 'osd.16' weight 3.64 at location {host=hh-yun-ceph-cinder016-128056,root=default} to crush map
Starting Ceph osd.16 on hh-yun-ceph-cinder016-128056...
Running as unit run-3126361.service.

系统会对该 osd 执行 recovery 操作, recovery 过程中, 会断开 block request, 那么这个 request 将会重新请求 mon 节点, 并重新获得新的 pg map, 得到最新的数据访问位置, 从而解决上述问题

参考恢复后的状态

[root@hh-yun-puppet-129021 ~]# ceph -s
    cluster dc4f91c1-8792-4948-b68f-2fcea75f53b9
     health HEALTH_OK
     monmap e3: 5 mons at {hh-yun-ceph-cinder015-128055=240.30.128.55:6789/0,hh-yun-ceph-cinder017-128057=240.30.128.57:6789/0,hh-yun-ceph-cinder024-128074=240.30.128.74:6789/0,hh-yun-ceph-cinder025-128075=240.30.128.75:6789/0,hh-yun-ceph-cinder026-128076=240.30.128.76:6789/0}, election epoch 216, quorum 0,1,2,3,4 hh-yun-ceph-cinder015-128055,hh-yun-ceph-cinder017-128057,hh-yun-ceph-cinder024-128074,hh-yun-ceph-cinder025-128075,hh-yun-ceph-cinder026-128076
     osdmap e97981: 190 osds: 190 up, 190 in
      pgmap v13669826: 20544 pgs, 2 pools, 77488 GB data, 19510 kobjects
            228 TB used, 426 TB / 654 TB avail
               20541 active+clean
                   3 active+clean+scrubbing+deep
  client io 21801 kB/s rd, 66461 kB/s wr, 2328 op/s