Host aggregate分区

需求：计算节点如今要拿出两台给单独的一个租户开虚机（虚机独占这两台计算节点），怎么作？

https://www.ibm.com/developerworks/cn/cloud/library/1604-openstack-host-aggregate/

问题描述：
生产区一期环境增长 4 台计算结点，但愿被大数据租户独占
1.  该大数据用户建立的虚拟机必须被调度到这 4 个计算结点
2.  其余租户建立虚拟机禁止调度到这 4 台机器

标黄的 须要自行根据状况修改
解决方案：
建立 ha 分区
1.  将 4 台计算结点置为一个 ha 分区
[root@controller-1 ~]# nova aggregate-create bigdata
+----+---------+-------------------+-------+----------+
| Id | Name | Availability Zone | Hosts | Metadata |
+----+---------+-------------------+-------+----------+
| 9 | bigdata | - | | |
+----+---------+-------------------+-------+----------+

1.  将 4 台计算结点加入这个 ha 分区
[root@controller-1 ~]# nova aggregate-add-host   9   compute-2
Host compute-3 has been successfully added for aggregate 9
+----+---------+-------------------+-------------+----------+
| Id | Name | Availability Zone | Hosts | Metadata |
+----+---------+-------------------+-------------+----------+
| 9 | bigdata | - | 'compute-2' | |
+----+---------+-------------------+-------------+----------+
（执行 4 次）

1.  增长 ha 区调度属性
[root@controller-1 ~]# nova aggregate-set-metadata   9   bigdata=yes
Metadata has been successfully updated for aggregate 9.
+----+---------+-------------------+-------------+---------------+
| Id | Name | Availability Zone | Hosts | Metadata |
+----+---------+-------------------+-------------+---------------+
| 9 | bigdata | - | 'compute-2' | 'bigdata=yes' |
+----+---------+-------------------+-------------+---------------+
bigdata=yes   是写死的，不能修改

4.   增长调度属性 2 ：
[root@controller-1 ~]# nova aggregate-set-metadata   9   filter_tenant_id= 7b597e8e774c41d9bcc92d494c0ad6ef
Metadata has been successfully updated for aggregate 9.
+----+---------+-------------------+-------------+--------------------------------------------------------------------+
| Id | Name | Availability Zone | Hosts | Metadata |
+----+---------+-------------------+-------------+--------------------------------------------------------------------+
| 9 | bigdata | - | 'compute-2' | 'bigdata=yes', 'filter_tenant_id=7b597e8e774c41d9bcc92d494c0ad6ef' |
+----+---------+-------------------+-------------+--------------------------------------------------------------------+
filter_tenant_id= 7b597e8e774c41d9bcc92d494c0ad6ef   等号后面的填写大数据租户的 tenant-id

此时， ha 分区建立完毕

修改调度策略
1.   修改三台控制结点的 nova.conf
修改：
scheduler_default_filters=RetryFilter,AvailabilityZoneFilter,RamFilter,CoreFilter,DiskFilter,ComputeFilter,ComputeCapabilitiesFilter,ImagePropertiesFilter,ServerGroupAntiAffinityFilter,ServerGroupAffinityFilter
为：
scheduler_default_filters=RetryFilter,AvailabilityZoneFilter,RamFilter,ComputeFilter, ImagePropertiesFilter,ServerGroupAntiAffinityFilter,ServerGroupAffinityFilter,AggregateMultiTenancyIsolation,AggregateInstanceExtraSpecsFilter

1.  重启 nova-scheduler
service openstack-nova-scheduler restart

建立大数据组独享的 flavor （模板）
1.   建立一个私有的模板（用 admin 租户的身份建立）
[root@controller-1 ~]# nova flavor-create   lihao1234 auto 1024 20 1   --is-public false
+--------------------------------------+-----------+-----------+------+-----------+------+-------+-------------+-----------+
| ID | Name | Memory_MB | Disk | Ephemeral | Swap | VCPUs | RXTX_Factor | Is_Public |
+--------------------------------------+-----------+-----------+------+-----------+------+-------+-------------+-----------+
| e5a55a80-015e-4aed-96cc-d6ca42574cf1 | lihao1234 | 1024 | 20 | 0 | | 1 | 1.0 | False |
+--------------------------------------+-----------+-----------+------+-----------+------+-------+-------------+-----------+

2.   将该模板给大数据租户使用
[root@controller-1 ~]# nova flavor-access-add   e5a55a80-015e-4aed-96cc-d6ca42574cf1   7b597e8e774c41d9bcc92d494c0ad6ef
+--------------------------------------+----------------------------------+
| Flavor_ID | Tenant_ID |
+--------------------------------------+----------------------------------+
| e5a55a80-015e-4aed-96cc-d6ca42574cf1 | 7b597e8e774c41d9bcc92d494c0ad6ef |
+--------------------------------------+----------------------------------+
前者是建立的 flavor 的 id
后者是大数据租户的 tenant-id

3.   将模板中置上调度参数
[root@controller-1 ~]# nova flavor-key   e5a55a80-015e-4aed-96cc-d6ca42574cf1   set bigdata=yes

效果：
1.   用大数据帐号，专用模板建立的虚拟机，必定且只会被调度到指定的 4 台主机上

2. 其余用户看不到特殊模板，且建立的虚拟机只能被调度到剩余的主机上，也就是说，原来的用户对该改造无感

来自为知笔记(Wiz)