openstack中的延迟删除
glance镜像的延迟删除
在控制节点的glance-api.conf文件中设置延迟删除：
# Turn on/off delayed delete delayed_delete = False # Delayed delete time in seconds scrub_time = 43200
重启openstack-glance-api.service
systemctl start openstack-glance-api.service
测试：删除镜像fedora
[root@controller ~(keystone_admin)]# glance image-list +--------------------------------------+-----------+-------------+------------------+-------------+--------+ | ID | Name | Disk Format | Container Format | Size | Status | +--------------------------------------+-----------+-------------+------------------+-------------+--------+ | 42d61acd-1ab3-4a7e-a14e-68ad6ba75215 | centos | iso | bare | 358959104 | active | | a7f49865-0388-48f1-a547-f6f23066fb4f | centos_wc | raw | bare | 10737418240 | active | | f01af3f9-58c3-463d-bf66-fb58825466b4 | fedora_s2 | qcow2 | bare | 983629824 | active | +--------------------------------------+-----------+-------------+------------------+-------------+--------+ [root@controller ~(keystone_admin)]# glance image-delete fedora_s2 [root@controller ~(keystone_admin)]# glance image-list +--------------------------------------+-----------+-------------+------------------+-------------+--------+ | ID | Name | Disk Format | Container Format | Size | Status | +--------------------------------------+-----------+-------------+------------------+-------------+--------+ | 42d61acd-1ab3-4a7e-a14e-68ad6ba75215 | centos | iso | bare | 358959104 | active | | a7f49865-0388-48f1-a547-f6f23066fb4f | centos_wc | raw | bare | 10737418240 | active | +--------------------------------------+-----------+-------------+------------------+-------------+--
查看数据库信息, fedora_s2的状态为pending_delete。
+--------------------------------------+-----------+-----------+----------------+-----------+---------------------+---------------------+---------------------+---------+-------------+------------------+----------------------------------+----------------------------------+----------+---------+-----------+ | id | name | size | status | is_public | created_at | updated_at | deleted_at | deleted | disk_format | container_format | checksum | owner | min_disk | min_ram | protected | +--------------------------------------+-----------+-----------+----------------+-----------+---------------------+---------------------+---------------------+---------+-------------+------------------+----------------------------------+----------------------------------+----------+---------+-----------+ | f01af3f9-58c3-463d-bf66-fb58825466b4 | fedora_s2 | 983629824 | pending_delete | 0 | 2014-02-15 15:03:22 | 2014-04-10 17:51:53 | 2014-04-10 17:51:53 | 1 | qcow2 | bare | c5870838c1c85547d5b85084071db21a | 86196260e1694d0cbb5049cfba3883f8 | 20 | 2048 | 0 | +--------------------------------------+-----------+-----------+----------------+-----------+--------
备注：目前image找回功能的接口还在review中，目前仍不可用。
实现原理
实现原理很简单。当打开延迟删除开关后，对image的删除不会立刻触发动作，而只是记录的状态为pending_delete和删除时间。
另外glance有个scrubber的清理服务，会周期性检查pending_delete的image是否到期，到期则进行删除动作。
glance镜像的删除保护
设置镜像保护是个admin操作，操作方法见如下：
[root@controller ~(keystone_admin)]# glance image-update --is-protected True centos_wc +------------------+--------------------------------------+ | Property | Value | +------------------+--------------------------------------+ | checksum | 2f7476ac2fe077979d2f0cda7640d1a8 | | container_format | bare | | created_at | 2014-04-03T10:45:19 | | deleted | False | | deleted_at | None | | disk_format | raw | | id | a7f49865-0388-48f1-a547-f6f23066fb4f | | is_public | False | | min_disk | 0 | | min_ram | 0 | | name | centos_wc | | owner | 86196260e1694d0cbb5049cfba3883f8 | | protected | True | | size | 10737418240 | | status | active | | updated_at | 2014-04-10T19:00:29 | +------------------+--------------------------------------+ [root@controller ~(keystone_admin)]# glance image-delete centos_wc Request returned failure status. 403 Forbidden Image is protected (HTTP 403): Unable to delete image centos_wc
实现原理
通过image的属性“protected”进行控制。 删除前检查该属性，未保护则可以删除。保护则不可以删除。 需要先将属性“protected”修改为“False“才能删除。
这里需要注意的是：我没在生产环境测试过，所以不知道效果如何
nova虚拟机的软删除
通过nova.conf，设置回收已删除虚拟机的间隔。
# Interval in seconds for reclaiming deleted instances # (integer value) #reclaim_instance_interval=0
reclaim_instance_interval=86400 #延迟一天后删除
该值为0，则立刻删除。不为0，则为软删除。reclaim_instance_interval的单位是秒
在该时间到达前，可以通过API将虚拟机找回。
周期性任务_reclaim_queued_deletes会定期检查，真正删除已经到达时间的虚拟机
重启服务：
在控制节点设置后，需要重启openstack-nova-api.service
systemctl restart openstack-nova-api.service
在计算节点设置后，需要重启openstack-nova-compute.service
systemctl restart openstack-nova-compute.service
实现原理
实现原理比较简单，通过虚拟机数据库中的状态进行软删除控制。
通过定时任务进行虚拟机的删除。
注意：
1.控制节点和计算节点都要设置
2.如果只在控制节点设置reclaim_instance_interval=86400，则虚拟机会有SOFT_DELETED标志，但是机器不会真正删除
用命令行查看软删除的机器：
nova list --deleted --all-tenants | grep SOFT
软删除的虚拟机恢复
nova restore 机器的uuid
