2014-10-14 18:32:43 |
Auston McReynolds |
description |
on instance-delete, the database should be stopped before deleting the nova compute.
why: if volumes are being used, delete_on_terminate today defaults to true (meaning if the compute is terminated, the volume is deleted). the phrase "the volume is deleted" means different things depending on the vendor you're using. for example, solidfire supports the ability to "undelete" a volume if done so within a specified window after deletion. if you're using such a vendor, you want the database properly stopped before deleting the server+volume to ensure that the data is a-ok for an undelete situation. |
on instance-delete, the database should be stopped before deleting the nova compute.
why: if volumes are being used, delete_on_terminate today defaults to true (meaning if the compute is terminated, the volume is deleted). the phrase "the volume is deleted" means different things depending on the vendor you're using. for example, solidfire supports the ability to "undelete" a volume if done so within a specified window after deletion. if you're using such a vendor, you want the database properly stopped before deleting the server+volume to ensure that the data is a-ok for an undelete situation.
relevant code section: https://github.com/openstack/trove/blob/0007a5076aa795fc354b122ae591980c9413f054/trove/taskmanager/models.py#L846-L857 |
|