Cannot delete vm instance if send duplicate delete requests
| Affects | Status | Importance | Assigned to | Milestone | |
|---|---|---|---|---|---|
| OpenStack Compute (nova) |
Confirmed
|
Low
|
Unassigned | ||
Bug Description
I deployed openstack with icehouse rc1 and booted 100 vms on my nodes. After my testing, i tried to delete my vms at the same time. Then i fount all of my vms` status change to deleting but cannot be deleted. I checked my openstack, the rabbitmq-server crashed . Then i restart rabbitmq-server and my openstack nova services, sended the delete requests again and again, the vms still cannot be deleted. While , in havana, the vms can be deleted if received duplicate delete requests .
I think icehouse should handle duplicate delete requests like havana .
Note:
This bug is already reported in launchpad [1] but the fix [2] proposed to resolve it was reverted back as it was breaking the cells.
[1] https:/
[2] https:/
| Changed in nova: | |
| assignee: | nobody → Abhishek Kekane (abhishek-kekane) |
| status: | New → In Progress |
| Changed in nova: | |
| importance: | Undecided → Low |
| Abhishek Kekane (abhishek-kekane) wrote | #1 |
| Changed in nova: | |
| assignee: | Abhishek Kekane (abhishek-kekane) → Rajesh Tailor (rajesh-tailor) |
| tags: |
added: compute removed: ntt |
| OpenStack Infra (hudson-openstack) wrote Change abandoned on nova (master) | #2 |
| Changed in nova: | |
| assignee: | Rajesh Tailor (rajesh-tailor) → nobody |
| status: | In Progress → Confirmed |
| OpenStack Infra (hudson-openstack) wrote | #3 |
| Sean Dague (sdague) wrote | #4 |
| tags: | added: openstack-version.icehouse |
Hi,
In Normal case, user will try to delete instance using command [1]. If user is not able to delete the instance
using command [1], then only he/she will try to force-delete it using command [2].
IMO the sole purpose of 'force-delete' is to delete an instance if it gets stuck in unrecoverable state.
In current bug, instance gets stuck in deleting task_state and since user can't reset the instance state, there is no way
user can delete that instance.
The proposed patch has a side effect, if user tries to force-delete the same instance multiple times. In that case,
nova-api server will be sending lots of rabbit messages to compute node to force-delete single instance.
Can we add new task_state 'force_deleting' which is set if instance is deleted using force-delete command [2] ?
If yes, we can add a check to reject subsequent force-delete requests.
[1] $ nova delete <instance_id>
[2] $ nova force-delete <instance_id>
Please give your opinion for the same.