[openstack-dev] [all][ironic] Kubernetes-based long running processes

Jay Pipes jaypipes at gmail.com
Sat Mar 18 15:36:10 UTC 2017


On 03/18/2017 08:33 AM, Tomasz Pa wrote:
> On 16 Mar 2017 6:21 pm, "Dean Troyer" <dtroyer at gmail.com
> <mailto:dtroyer at gmail.com>> wrote:
>
>     Before implementing something new it would be a good exercise to have
>     a look at the other existing ways to run VMs and containers already in
>     the OpenStack ecosystem.  Service VMs are a thing, and projects like
>     Octavia are built around running inside the existing infrastructure.
>     There are a bunch of deployment projects that are also designed
>     specifically to run services with minimal base requirements.
>
> VMs are having much bigger overhead than containers. Imagine you have
> Ironic cluster with 3000 bare-metal nodes with each having console
> enabled. Overhead on running 3000vms vs 3000 containers is huge.

Would it be possible to run >1 console connection within a single VM? 
I'm thinking here of a single VM running, say, k8s that manages the 
liveness of 3000 lightweight containers inside the VM.

Would that accomplish what Ironic is looking for in this case?

Not saying k8s on baremetal is a bad solution, just throwing some ideas 
around that might be a good compromise between what solutions we 
currently have and what might be a good fit for you.

 > No to
> mention that Kubernetes container highavalability is far ahead when
> compared to openstack vms.
>
> Also Kubernetes is not something new and framework to launch containers
> on top of it would be really light.

Dean wasn't referring to k8s being "new". He was referring simply to 
first examining the existing solutions we have in this space and seeing 
if any of those solutions fit the problem to any extent.

Best,
-jay



More information about the OpenStack-dev mailing list