The bug #1341420 can cause the driver to try to deploy onto a node which
already has an instance associated with it, the way Ironic reserve a
node to a specific instance is associating the UUID of that instance
with the node via the instance_uuid field. This patch just makes the
request to associate that field non-retriable when the api returns HTTP
409 (Conflict) (that's the HTTP error indicating that the node is
already associated with another instance), so it can fail fast in
case the node is already pick and the RetryFilter could then pick
another node for that instance.
Reviewed: https:/ /review. openstack. org/226235 /git.openstack. org/cgit/ openstack/ nova/commit/ ?id=608af875412 40793425760c16e 528d9b3ec02fee
Committed: https:/
Submitter: Jenkins
Branch: master
commit 608af8754124079 3425760c16e528d 9b3ec02fee
Author: Lucas Alvares Gomes <email address hidden>
Date: Mon Sep 21 16:36:04 2015 +0100
Ironic: Workaround to mitigate bug #1341420
The bug #1341420 can cause the driver to try to deploy onto a node which
already has an instance associated with it, the way Ironic reserve a
node to a specific instance is associating the UUID of that instance
with the node via the instance_uuid field. This patch just makes the
request to associate that field non-retriable when the api returns HTTP
409 (Conflict) (that's the HTTP error indicating that the node is
already associated with another instance), so it can fail fast in
case the node is already pick and the RetryFilter could then pick
another node for that instance.
Partial-Bug: #1341420 562182bbb3ceb8c b160e84c312
Change-Id: I697c5129893b18