pxe_ephemeral_format': u'ext4 left on nodes after deploys deleted

[Robert Collins]

At least, I think thats how its happening ;)

[James Polley (tchaypo)]

<tchaypo> I've had a quick look at https://bugs.launchpad.net/tripleo/+bug/1320090 and I have no idea what problem you're reporting or how to reproduce it. Could you add some detail?
<lifeless> run ironic node-show $somenode

[Ben Nemec (bnemec)]

Nothing in this bug tells me what problem this is causing (ie can nodes not be redeployed because of this?), and as such I can't triage it properly without more information.

[James Polley (tchaypo)]

Rob, I've assigned to you as no-one else is sure what this means. Please add some detail or close this off.

[Robert Collins (lifeless)]

Nodes can't be deployed (there is a separate bug for that), but the setting is ephemeral and shouldn't be left behind - its cruft.

[Adam Gandelman (gandelman-a)]

I'm unable to reproduce this. Setting default_ephemeral_format=ext4 in nova.conf, booting and deleting an instance. Node's driver_info fields get updated and cleaned appropriately:

driver_info post-boot:

{u'pxe_image_source': u'9f6345c6-e482-4805-839c-62d6b643e0b5',
u'pxe_ramdisk': u'fdb2c3a9-74f1-48b0-9a05-e678a00ad6c0', u'pxe_root_gb':
u'8', u'ssh_port': u'22', u'ssh_username': u'ubuntu', u'pxe_swap_mb':
u'0', u'pxe_ephemeral_format': u'ext4', u'pxe_deploy_kernel': u'003fa109
-dd9a-4c0a-b51d-64d87573f198', u'pxe_kernel': u'77393603-da4c-4bcc-
9c77-3217df20ddf1', u'pxe_deploy_key':
u'727BL418BCB4C7LDLNRFCYIRR020H9WJ', u'pxe_deploy_ramdisk': u'a6579774
-b88a-4f66-9703-13d9b8063ec1', u'ssh_key_filename':
u'/opt/stack/data/ironic/ssh_keys/ironic_key', u'ssh_address':
u'10.0.0.179', u'ssh_virt_type': u'virsh', u'pxe_ephemeral_gb': u'1'}

driver_info post-delete:

{u'ssh_port': u'22', u'ssh_username': u'ubuntu', u'ssh_virt_type':
u'virsh', u'ssh_address': u'10.0.0.179', u'ssh_key_filename':
u'/opt/stack/data/ironic/ssh_keys/ironic_key'}

[Robert Collins (lifeless)]

I was dealing with broken hardware at the time - I suspect its in the cleanup path that this happens.

[Dmitry Tantsur (divius)]

Hi Robert! We've approved a spec https://review.openstack.org/#/c/95701/ one part of which is moving all temporary variables from driver_info to instance_info, with the latter being cleaned unconditionally every time. https://review.openstack.org/#/c/94855/ is responsible for this. I'm assigning Lucas, as he's the one who works on that patch.

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/94855
Committed: https://git.openstack.org/cgit/openstack/ironic/commit/?id=58fdb1b6977fd9ffa39f78271152fbbec791294a
Submitter: Jenkins
Branch: master

commit 58fdb1b6977fd9ffa39f78271152fbbec791294a
Author: Lucas Alvares Gomes <email address hidden>
Date: Thu May 22 11:09:32 2014 +0100

    Move PXE instance level parameters to instance_info

    The commit 7735ad404bb2849f5be738c2f8a536b1f817c6b7 introduced a
    new instance_info field to the Nodes so that all the instance level
    information should live there and not mixed in the Node.driver_info
    field. This patch is moving all the PXE instance level parameters from
    the Node.driver_info to the Node.instance_info.

    As all the data in the Node.instance_info is related to the instance,
    this field will be completed erased as part of the node's tear down
    process. For that reason the _removes_internal_attrs() function from
    the PXE driver was removed.

    These fields have changed from driver_info to instance_info and
    their driver namespace ('pxe_') was removed because they are now
    instance-specific parameters:

      'pxe_image_source' -> 'image_source'
      'pxe_root_gb' -> 'root_gb'
      'pxe_swap_mb' -> 'swap_mb'
      'pxe_ephemeral_gb' -> 'ephemeral_gb'
      'pxe_ephemeral_format' -> 'ephemeral_format'
      'pxe_preserve_ephemeral' -> 'preserve_ephemeral'

    DocImpact

    Implements: blueprint add-node-instance-info
    Closes-Bug: #1320090
    Change-Id: I608d2cdc6170d0bed826e532dedd120dd32a3b7e