Nova compute process crash doesn't reset task state 'migrating'

[John Warren]

When a live migration (task state = migrating) task is running, and the compute service goes down, the instance remains in migrating state when the compute service restarts. The compute/manager.py _init_instance method seems to handle all states except migrating. It handles inactive migration (resize_migrating), but not live migration.

[Matt Riedemann (mriedem)]

Bug 1276214 sounds related but the difference is when the failure happens.

[Loganathan Parthipan (parthipan)]

Bug 1319797 is also related.

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/107435

[Jay Pipes (jaypipes)]

Setting to medium, as there is a workaround of updating the database manually... will review the patch shortly.

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/107435
Committed: https://git.openstack.org/cgit/openstack/nova/commit/?id=fc821e10f259ec8747c2c3b495ce8a4c05c8349c
Submitter: Jenkins
Branch: master

commit fc821e10f259ec8747c2c3b495ce8a4c05c8349c
Author: John Warren <email address hidden>
Date: Wed Jul 16 16:00:32 2014 +0000

    Reset task state 'migrating' on nova compute restart

    This change has the compute manager's init_instance method check
    for a task state set to 'migrating' and resets the task state to
    None, so that live migrations that were pending when the compute
    service went down do not result in instances that are perpetually
    stuck in 'migrating' task state.

    Change-Id: If23ea983cc048f317daae594f420a643f1bfe6ac
    Closes-Bug: #1339823