Open Stack

Hi friends,

The two devstack patches mentioned below have had the latest patchset up for a week now, and still only have one +2 on them. The other patches (and the ultimate goal of getting CI running on this driver) are blocked these patches.

Could a devstack core please review these soon? Thanks!

// jim

On August 7, 2014 at 3:51:21 PM, Jay Faulkner (jay at jvf.cc) wrote:

Hi all,



At the recent Ironic mid-cycle meetup, we got the first version of the ironic-python-agent (IPA) driver merged. There are a few reviews we need merged (and their dependencies) across a few other projects in order to begin testing it automatically. We would like to eventually gate IPA and Ironic with tempest testing similar to what the pxe driver does today.



For IPA to work in devstack (openstack-dev/devstack repo):

 - https://review.openstack.org/#/c/112095 Adds swift temp URL support to Devstack

 - https://review.openstack.org/#/c/108457 Adds IPA support to Devstack

 

Docs on running IPA in devstack (openstack/ironic repo):

 - https://review.openstack.org/#/c/112136/

 

For IPA to work in the devstack-gate environment (openstack-infra/config & openstack-infra/devstack-gate repos):

 - https://review.openstack.org/#/c/112143 Add IPA support to devstack-gate

 - https://review.openstack.org/#/c/112134 Consolidate and rename Ironic jobs

 - https://review.openstack.org/#/c/112693 Add check job for IPA + tempest 



Once these are all merged, we'll have IPA testing via a nonvoting check job, using the IPA-CoreOS deploy ramdisk, in both the ironic and ironic-python-agent projects. This will be promoted to voting once proven stable.



However, this is only one of many possible IPA deploy ramdisk images. We're currently publishing a CoreOS ramdisk, but we also have an effort to create a ramdisk with diskimage-builder (https://review.openstack.org/#/c/110487/) , as well as plans for an ISO image (for use with things like iLo). As we gain additional images, we'd like to run those images through the same suite of tests prior to publishing them, so that images which would break IPA's gate wouldn't get published. The final state testing matrix should look something like this, with check and gate jobs in each project covering the variations unique to that project, and one representative test in consuming project's test pipelines.



IPA:

 - tempest runs against Ironic+agent_ssh with CoreOS ramdisk

 - tempest runs against Ironic+agent_ssh with DIB ramdisk

 - (other IPA tests)

 

IPA would then, as a post job, generate and publish the images, as we currently do with IPA-CoreOS ( http://tarballs.openstack.org/ironic-python-agent/coreos/ipa-coreos.tar.gz ). Because IPA would gate on tempest tests against each image, we'd avoid ever publishing a bad deploy ramdisk.



Ironic:

 - tempest runs against Ironic+agent_ssh with most suitable ramdisk (due to significantly decreased ram requirements, this will likely be an image created by DIB once it exists)

 - tempest runs against Ironic+pxe_ssh

 - (what ever else Ironic runs)

 

Nova and other integrated projects will continue to run a single job, using Ironic with its default deploy driver (currently pxe_ssh).

 

 

Using this testing matrix, we'll ensure that there is coverage of each cross-project dependency, without bloating each project's test matrix unnecessarily. If, for instance, a change in Nova passes the Ironic pxe_ssh job and lands, but then breaks the agent_ssh job and thus blocks Ironic's gate, this would indicate a layering violation between Ironic and its deploy drivers (from Nova's perspective, nothing should change between those drivers). Similarly, if IPA tests failed against the CoreOS image (due to Ironic OR Nova change), but the DIB image passed in both Ironic and Nova tests, then it's almost certainly an *IPA* bug.



Thanks so much for your time, and for the Openstack Ironic community being welcoming to us as we have worked towards this alternate deploy driver and work towards improving it even further as Kilo opens.



--

Jay Faulkner

_______________________________________________  
OpenStack-dev mailing list  
OpenStack-dev at lists.openstack.org  
http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev  
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack-dev/attachments/20140814/e7e76650/attachment.html>