[Openstack-operators] Managing pools of pre-allocated instances

Mathieu Gagné mgagne at iweb.com
Wed Feb 25 18:34:42 UTC 2015


On 2015-02-25 1:21 PM, James Penick wrote:
> Hey folks,
>   I have a team that uses our OpenStack VM and BM clusters in their CI
> and CD environments (And yes, they really do need baremetal for some of
> their CI tests, unless I want to support gigantifriggenormous VM flavors).
>
>   CI on baremetal works ok, however the time it takes to image the host
> prior to running the tests slows down their whole CI pipeline. What they
> want to do is boot
> instances in advance then as needed select an unused instance and
> execute their CI job on it. Upon completion of their CI job they'd like
> to reimage the host and return it to an "unused" state.
>
>   I came up with something using instance names as an atomic-ish means
> of managing state. But it's pretty hacky. Before I go off and have to
> build a new webservice myself, has anyone heard of a tool that does
> something like this already?
>

Isn't it what nodepool supposed to do?

The OpenStack CI infra wrote and uses it:
http://ci.openstack.org/nodepool.html
http://ci.openstack.org/nodepool/

-- 
Mathieu



More information about the OpenStack-operators mailing list