openstack

they are two are separate concepts. The technical community has embraced an inclusive governance structure (big tent), while the foundation does work to define "core" for trademark program purposes

We need @eglute @zehicle @sdague @e_monty on this crowdchat to help translate.

The big tent is about being more inclusive of "who is OpenStack" in terms of development process, but less useful than previous designations in terms of navigating. Hence the introduction of tags and the project navigator.

@russellbryant Where are we with the tags? They were supposed to help sort out the projects in the big tent.

a relevant post from last week, "DefCore: What OpenStack Devs Need to Know" -- http://markvoelker.g...

@kenhuiny tags are exposed via the project navigator, a couple of new ones were added to governance in this cycle...

@xsgordon @kenhuiny tags have been progressing, perhaps slowly. Full list of currently defined tags: http://governance.op...

useful data in the project navigator (will be updated soon for Mitaka) https://www.openstac...

@kenhuiny some examples - "assert:supports-upgrade, assert:supports-rolling-upgrade, stable:follows-policy" - check out http://governance.op...

@xsgordon the upgrade tags were a nice addition. they still lack some nuance (upgrades with which configurations or backends?)

Will the tags help users to determine what projects are mature enough for production?

DefCore latest guideline: https://github.com/o... Companies must certify using 2 latest guidelines.

@russellbryant another nice resource to mention is the work @thingee has done to improve the new contributor experience https://twitter.com/...

DefCore 101 Tokyo presentation and slides: https://www.youtube.... http://www.slideshar...

The navigator rolls a number of tags and survey metrics into a maturity "score" - still not really granular enough for operators to navigate IMO

DefCore sets base requirements by defining 1) capabilities, 2) code and 3) must-pass tests for all OpenStack products

my roundup from Vancouver, where we discussed this topic a lot: http://wikibon.com/o...

User Survey showed deployments use 11 projects on average; well beyond the 6 core.

big tent is no longer a topic of discussion and has been fostering the inclusion of additional OpenStack services over the past year. The discussions in DefCore are about ensuring interop between providers of highly-adopted OpenStack services

top attended sessions at recent summits, @coreoslinux attends, @docker is discussed. Kubernetes is the new hot topic.

What do you mean stateless compute? Containers, cloud-native apps, serverless computing?

@stu The user survey shows big increase in use of kubernetes as opposed to docker swarm.

@kenhuiny Swarm seems to be confined to the All-Docker environment, whereas k8s seems to have broader appeal in more frameworks

Meaning the management layer running in containers ie Heat in a rkt container. It allows for versioning of the mgmt layer. But can OpenStack management nodes easily operate in such an infrastructure.

@kenhuiny user survey showed top 3 technologies of interest were 1) Containers 2) NFV & SDN, 3) Bare Metal.

It will certainly be discussed. I see OpenStack as very complementary as a platform to run your containers / orchestration system of choice on

@kenhuiny expect some LIVE DEMOS at the summit involving OpenStack & Kubernetes #shamelessplug

@sparkycollier OpenStack on k8s as an application, or OpenStack with k8s as a container scheduler for applications?

more noise has been made about containers *on* OpenStack but definitely lots of work being done on the backend on containerizing OpenStack itself. Main improvement in Mitaka from this perspective is probably Kolla upgrade support.

Also working on making Kolla a little bit more abstract, easier to plug in different orchestration options (Ansible, Mesos, Kubernetes)

@xsgordon what's the difference between Magnum and Kolla?

@bgracely @xsgordon My understanding is Magnum focuses on containers 4 workloads on OpenStack & Kolla focused on deploying OpenStack using containers.

@kenhuiny yes, there is a bit more flexibility regarding workloads though - folks also using Heat directly or even just rolling their own using e.g. Ansible on older clouds that don't have Murano/Heat yet.

Existing deployments will of course also need an upgrade path to get to containerized OpenStack in the first place (not an issue for greenfield).

Also, User Survey showed Magnum #1 project of interest (44%) followed closely by Manila.

Do we get to a day where Oracle runs on OpenStack and not just only cloud-native apps?

will we see Microsoft or more Google (other than Kubernetes) deeper into OpenStack?

I recently learned of a large bank running Oracle on OpenStack in production, and SAP is also doing a lot to support OpenStack. SAP is doing a keynote in Austin, in fact.

@stu Google seems to be getting more involved. They are Foundation members and recently rolled out Cinder backup to GCP https://goo.gl/2IGQz...

I keep hearing that there is a proposal to add vMotion-like functionality to OpenStack. Is that the last piece needed to get traditional apps on OpenStack?

yes it's great to see google involved, and in fact @cmcluck is joining my keynote to talk about how kubernetes and OpenStack along with @polvi of coreos - Tuesday morning April 26

@bgracely this is delivered by a combination of API hooks for e.g. Nova hooked into existing HA tools (e.g. Pacemaker) number of vendors have this working now, still more failure types to be handled in future though.

@bgracely That and VM HA. @russellbryant has written a bunch about it.

@bgracely features like the "mark host down" API added by OPNFV folks allow external HA tools to do their job more efficiently in terms of recovery times in case of e.g. node failure

@xsgordon OPNFV has been working in the community to help OpenStack evolve to meet the needs of the Telecom industry. Collaboration is key!

Getting there: https://apps.opensta... (and more focused on consuming Oracle versus deploying)

is there a place that people can email you if they'd like to be a guest on theCUBE as an end-user?

I saw @sparkycollier talk about Mitaka being more focused on operations and running the platform. Would love to understand more.

Be particularly interested in hearing what end users are excited about.

OpenStack Client is really maturing as a standard way to interact with OpenStack powered clouds

@sparkycollier That's big. The recent user survey cited APIs as the top reason 4 deploying OpenStack.

From an operational perspective it's great to see rolling upgrade support making its way to more projects, incl. Cinder in Mitaka.

A lot of the user driven improvements were directly from users like Comcast, Time Warner Cable, China Mobile, Yahoo!, Walmart, Deutsche Telekom who've been running in production over time.

also for operators, the ability to combine the distributed virtual router feature with high availability of the layer 3 agent in OpenStack Networking (Neutron)

Deployments are still being cited as pain points. What has improved there with Mitaka?

@xsgordon it's still strange that functionality like "rolling upgrades" doesn't make it into all projects. this is why people struggle with how to understand OpenStack.

@kenhuiny Great point, 97% of users said having a standard API & platform were top business drivers for selecting openstack

@bgracely did someone say something about a big tent?

@bgracely I would hope this would make it into at least the core projects. But then we need to define the core projects. :)

the user survey is a great resource - April 2016 data just released https://www.openstac...

In general, how stable is Neutron in Mitaka? That's an area that is frequently called out as needed more work. Is it production ready?

@bgracely The user survey shows that over 90% (i think?) of deployments are using Neutron, so I'd say so. There's still a lot of variation on the backend, though.

@kenhuiny another aspect is where you can/can't tolerate downtime - arguable different level of need for infra building blocks (neutron, cinder, nova) versus IaaS+ (trove, sahara, etc.) IMO

@xsgordon e.g. Nova rolling upgrades got people a long way, because the VMs stay up while we move everything else around them.

most users are running Neutron in production for the past 2 or 3 releases and with Liberty we actually heard really positive feedback. Mitaka took the stability of Liberty and added some ease of setup improvements for operators.

@russellbryant <- what he said

@sparkycollier it also seems like we are gradually seeing less people stuck many versions behind current releases too - just my eyeball test though