后端ceph以太网切换为IB网

openstack集群已经对接后端ceph了, 但是使用的是以太网IP, 想要切换到IB IP.

# vim /etc/hosts
10.121.2.125 compute1
10.5.138.125 compute1

这中间涉及到:

• ceph集群更改ip并重启
• openstack这边需要更改配置

ceph集群更改后, ceph.conf文件肯定会变, 其它的比如client的key不变

openstack修改如下

修改/etc/kolla/global.yml

storage_interface: "ib0"

修改/etc/kolla/config/下的配置, 其实只需要修改ceph.conf即可

# ll /etc/kolla/config/*/ceph.conf
-rwxr-xr-x. 1 root root 267 3月  25 20:30 /etc/kolla/config/cinder/ceph.conf
-rwxr-xr-x. 1 root root 267 3月  25 20:30 /etc/kolla/config/glance/ceph.conf
-rwxr-xr-x. 1 root root 267 3月  25 20:30 /etc/kolla/config/nova/ceph.conf

# cat /etc/kolla/config/*/ceph.conf
[global]
fsid = f09aeb2a-0927-4bd1-8682-c43a1cddf826
mon_initial_members = SH-IDC1-10-5-8-126, SH-IDC1-10-5-8-127, SH-IDC1-10-5-8-128
mon_host = 10.5.8.126,10.5.8.127,10.5.8.128
auth_cluster_required = cephx
auth_service_required = cephx
auth_client_required = cephx
[global]
fsid = f09aeb2a-0927-4bd1-8682-c43a1cddf826
mon_initial_members = SH-IDC1-10-5-8-126, SH-IDC1-10-5-8-127, SH-IDC1-10-5-8-128
mon_host = 10.5.8.126,10.5.8.127,10.5.8.128
auth_cluster_required = cephx
auth_service_required = cephx
auth_client_required = cephx
[global]
fsid = f09aeb2a-0927-4bd1-8682-c43a1cddf826
mon_initial_members = SH-IDC1-10-5-8-126, SH-IDC1-10-5-8-127, SH-IDC1-10-5-8-128
mon_host = 10.5.8.126,10.5.8.127,10.5.8.128
auth_cluster_required = cephx
auth_service_required = cephx
auth_client_required = cephx

然后

修改/etc/kolla/config/需要执行

# ./kolla-ansible -i ../../multinode reconfigure 

reconfigure用来reconfigure OpenStack service.

修改/etc/kolla/global.yml

# ./kolla-ansible -i ../../multinode upgrade
# ./kolla-ansible -i ../../multinode post-deploy

upgrade用来upgrade现有openstack环境

创建虚拟机失败

2017年社区有人提过这个bug, 链接 https://bugs.launchpad.net/kolla-ansible/+bug/1678013 , 但是后面60天过期了, 就没有然后了.....

重启nova-compute的container就好了

定位:

可以很明显看到, neutron没有发送'network-vif-plugged'事件给nova-api

但是我们重启nova-compute的container就正常了

目前规避办法:

修改kolla-ansible/ansible/roles/nova/tasks/upgrade.yml文件, 在最下面添加

- name: Restart nova-compute container
  command: docker restart nova_compute
  when: inventory_hostname in groups['compute']

在中间涉及到mariadb的错误, 需要恢复mariadb, 执行

./kolla-ansible -i ../../multinode mariadb_recovery