当OpenStack计算和存储一体节点强制重启后，会导致虚拟机无法启动，卡在I/O error上，究其原因是由于底层rbd设备的exclusive-lock造成的。有两种解决方案，一种是临时性的解除exclusive-lock client，一种则是关闭rbd的exclusive-lock。
解除exclusive-lock client
查看exclusive-lock client(pool的名字为volumes)
1
2
3
4
| [root@compute201 ~]# rbd lock ls volumes/volume-cad6425c-4453-4489-b2d1-45050b062bd2
There is 1 exclusive lock on this image.
Locker ID Address
client.14440803 auto 93942189655296 10.0.100.202:0/1431194814
|
尝试删除client
1
| rbd lock remove volumes/volume-cad6425c-4453-4489-b2d1-45050b062bd2 "auto 93942189655296" client.14440803
|
最后尝试重启相应的虚拟机，此时应该可以正常进入系统。
关闭exclusive-lock功能
通过查看rbd info，获取当前卷是否开启了exclusive-lock
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
| [root@compute201 ~]# rbd info volumes/volume-cad6425c-4453-4489-b2d1-45050b062bd2
rbd image 'volume-cad6425c-4453-4489-b2d1-45050b062bd2':
size 50 GiB in 12800 objects
order 22 (4 MiB objects)
snapshot_count: 0
id: dc59383d97f124
block_name_prefix: rbd_data.dc59383d97f124
format: 2
features: layering, exclusive-lock, object-map, fast-diff, deep-flatten
op_features:
flags:
create_timestamp: Fri Jun 18 16:28:44 2021
access_timestamp: Mon Jun 21 15:17:38 2021
modify_timestamp: Mon Aug 2 23:07:58 2021
parent: images/9e8a401b-d407-4946-b94c-f24b94ce0f02@snap
overlap: 8 GiB
|
关闭exclusive-lock功能，这里必须要和object-map或journaling一起关闭才能生效
1
| rbd feature disable volumes/volume-cad6425c-4453-4489-b2d1-45050b062bd2 exclusive-lock,object-map
|
再次查看功能列表，features只剩下layering和deep-flattern
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
| [root@compute201 ~]# rbd info volumes/volume-cad6425c-4453-4489-b2d1-45050b062bd2
rbd image 'volume-cad6425c-4453-4489-b2d1-45050b062bd2':
size 50 GiB in 12800 objects
order 22 (4 MiB objects)
snapshot_count: 0
id: dc59383d97f124
block_name_prefix: rbd_data.dc59383d97f124
format: 2
features: layering, deep-flatten
op_features:
flags:
create_timestamp: Fri Jun 18 16:28:44 2021
access_timestamp: Mon Jun 21 15:17:38 2021
modify_timestamp: Mon Aug 2 23:35:33 2021
parent: images/9e8a401b-d407-4946-b94c-f24b94ce0f02@snap
overlap: 8 GiB
|
批量关闭所有卷的exclusive-lock
1
| for i in $(rbd ls -p volumes); do rbd feature disable volumes/$i exclusive-lock,object-map; done
|
关闭新创建卷的exclusive-lock功能
如果是用Ceph Deploy进行的部署，则需要将配置文件分发到各个节点，在ceph.conf中增加rbd_default_features，写明你创建的卷需要哪几种属性。
1
2
3
4
5
6
7
8
9
| [global]
fsid = 74fcb885-c0f9-4c31-a8fe-89a7f0122434
public_network = 10.0.100.0/24
mon_initial_members = compute201
mon_host = 10.0.100.201
auth_cluster_required = cephx
auth_service_required = cephx
auth_client_required = cephx
rbd_default_features = "layering, deep-flatten"
|
1
| ceph-deploy --overwrite-conf config push compute201 compute202 compute203
|
此时再使用rbd create创建的卷默认就是这两种属性了
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
| [root@compute201 ~]# rbd create -p volumes test --size 10G
[root@compute201 ~]# rbd info volumes/test
rbd image 'test':
size 10 GiB in 2560 objects
order 22 (4 MiB objects)
snapshot_count: 0
id: 16ef5c2dacd2da
block_name_prefix: rbd_data.16ef5c2dacd2da
format: 2
features: layering, deep-flatten
op_features:
flags:
create_timestamp: Mon Aug 2 23:57:23 2021
access_timestamp: Mon Aug 2 23:57:23 2021
modify_timestamp: Mon Aug 2 23:57:23 2021
|
最后还需要对OpenStack各个模块的ceph进行更新，如果使用kolla更新，则替换所有配置文件，并重启容器。
1
2
3
4
| cp /etc/ceph/ceph.conf /etc/kolla/config/nova/ceph.conf
cp /etc/ceph/ceph.conf /etc/kolla/config/cinder/ceph.conf
kolla-ansible -i multinode reconfigure
|
老孙正经胡说