[sdk]: Block storage service update volume status from reserved to available

Lee Yarwood lyarwood at redhat.com
Wed Sep 8 10:57:00 UTC 2021

On 08-09-21 10:35:48, Rajat Dhasmana wrote:
> Hi,
> On Wed, Sep 8, 2021 at 1:02 AM dmeng <dmeng at uvic.ca> wrote:
> > Hello there,
> >
> > Hope everything is going well.
> >
> > I would like to know if there is any method that could update
> > the volume status or reset its state. While we creating volumes use the
> > block storage service or creating vms with volumes use the compute service,
> > sometimes we got volumes stuck in creating/deleteing/attaching/reserved
> > state. And we couldn't remove them or reset their state without an admin
> > account even we own the project. The way we create volumes and boot vms is
> > like the following:
> >
> > # create the volume
> > cv = cinder.create_volume(name=bv_name, size=size, image_id=imageobj.id, volume_type=volume_type, is_bootable=True)
> > # format mapping data
> > bdm = [{'uuid': cv.id, 'source_type': 'volume', 'destination_type': 'volume', 'delete_on_termination': True, 'boot_index': 0}]
> > # create the vm
> > nova.create_server(name=hostname, image_id=imageobj.id, flavor_id=flavorl.id, key_name=key_name, block_device_mapping=bdm, user_data=format_userdata, networks=netid, security_groups=sec_groups)
> >
> > In the normal situation, everything works. But in the case of any
> > network issue, the volume might stuck in the process of creating,
> > deleting, or attaching. And if the volume got created successfully,
> > when it tries to create the vm, the volume will be reserved for it,
> > but if anything wrong happen when creating the vm or get timed out,
> > the volume will be in the reserved stage forever. We would like to
> > know if we could update the state or delete them without the admin
> > account?

You need to delete the ERROR'd out instance, this will delete the volume
attachment in Cinder moving the volume to available again. Nova does
this to ensure another instance doesn't take this volume as the user
might request that the instance is rebuilt to workaround whatever issue
they encountered during the initial build.

> > We tried the command line tools 'openstack volume set --state
> > available <volume_id>' and 'cinder reset-state --state available
> > <volume_id>', but both got: Policy doesn't allow
> > volume_extension:volume_admin_actions:reset_status to be performed.
> > (HTTP 403) (Request-ID: req-b50236c3-5662-4cc0-8602-19e79e80219d)
> > ERROR: Unable to reset the state for the specified entity(s).
> >
> For the specific case where a volume is stuck in reserved state and the
> above code uses cinder's new attachments API, you can use* ``cinder
> --os-volume-api-version 3.27 attachment-list`` *to list all attachments for
> volumes and if there exists attachment(s) for your volume, you can delete
> them via the command  *``cinder --os-volume-api-version 3.27
> attachment-delete <attachment-id>``*

As above we retain the volume attachment on build failures so either
rebuild the instance or delete it to free up the volume again.

Lee Yarwood                 A5D1 9385 88CB 7E5F BE64  6618 BCA6 6E33 F672 2D76
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 488 bytes
Desc: not available
URL: <http://lists.openstack.org/pipermail/openstack-discuss/attachments/20210908/af2e42d3/attachment.sig>

More information about the openstack-discuss mailing list