Wrapped the previous comments in the bifrost-test-vm role
and added an additional note about the future possibility
of using the inventory it's self to define the nodes.
Removed a comment noting that the format was conceptual
for the execution of os_ironic_node as that module had
not yet been written when the playbook was initially
written.
Cleaned up variable list to be a bit more clear to understand
what the variable actually is, as well as added the ironic_url
variable, and noted that the testing variable is defaulted to
false.
Updated the url for the ipmi retry patch to represent a newer
version of the patch which does contain one fix from the previous
version, as well as unit tests.
Updated the Bifrost validation role to have default variables
set and updated the documentation to indicate variables that are
likely to differ on customized environments, and included a sample
playbook.
Changed the name of the ironic-deploy role to
bifrost-validate-host-for-deploy to be in-line
with what the role actually does which is perform
basic node validation before the process proceeds.
Upon realizing that using the testing variable on the commandline was
unreliable, I set it into the test-bifrost.yaml playbook to assume
it by default and set the variable for the playbooks it is invoking.
Changed out the patch that Bifrost retreives for installation as the
patch that was present landed in Ironic today. The new patch is to
address ipmi retries on busy BMCs where multiple nodes share the same
BMC address.
Added ironic-delete step into the test-bifrost.yaml playbook sequence
and updated documentation to indicate that the node was being removed
from active use and then deleted from ironic.
Additionally noted in the README that cleaning is disabled for the
bifrost-test playbook.
Turning off cleaning for test-bifrost.yaml as the cleaning step
is a lengthy IO intensive operation, and the testing sequence
does not need cleaning at this point in time.
Added a basic role which, in the same style as other roles,
unprovisions the nodes created.
Added to the test-bifrost.yaml playbook as it will be useful
in the testing sequence.
Changed the Ironic client and Shade installations to utilize pip
by default however with a flag that could be set in order to change
the logic over to utilizing source based installation.
Added basic details on use of the test-bifrost.yaml playbook.
With the previous addition of a self contained basic test sequence
and basic documentation, this commit should close out issue #4.
Added two roles and appropriate changes to test-bifrost.yaml in
order to add the new hosts, typically which should be a single
virtual machine, to the running inventory, and then connect
to them and verify that the hostname is NOT ubuntu.
This is in support of issue #4.
Changed deploy.yaml to not invoke sudo for the configdrive and
deployment roles in order to allow a user's key to be pulled into
the configuration drive instead of the root user key by default.
Changed the configdrive generation steps to utilize sudo so the base
user context remains the same as the running user the playbook
and thus the user's key can be incorporated without needing root
privliges
This is part of issue #5
Adding temporary addition of https://review.openstack.org/#/c/166903/
to the ironic installation in order to be able to utilize bifrost.
This effectively closes issue #6.
Due to a race condition in CoreOS, any pre-existing configuration
drive that is present on a system is being attached to the node.
The only documented way to prevent CoreOS from attaching a
pre-existing configuration drive is to pass the kernel parameter.
This combined with an update that has been made to IPA, closes #2.
Additional details can be found at
https://bugs.launchpad.net/ironic/+bug/1433812
