Whats New In OpenStack Juno Q&A
Last week we held our semi-annual "What's New in OpenStack" webinar to talk about the major new features in OpenStack's Tenth Release, Juno. As always we had a great response, and we didn't have a chance to get to all of the questions, so we're providing all of the questions and answers here.
The webinar was hosted by me and my colleague, Christopher Aedo. If you missed it, you can still catch the replay.
Why didn't you cover TripleO?
Nick Chase: According to the Foundation's definitions, TripleO isn't so much a "project" is it is a deployment tool used to deploy OpenStack. Since we already had a plethora of projects to cover, we elected to skip it. That's not to say that the team isn't working hard; I understand they've accomplished a lot this cycle.
What is QuintupleO?
Nick Chase: Ah, QuintupleO. TripleO, or OpenStack on OpenStack, is designed to let people use OpenStack to deploy OpenStack, so that they don't have to go through the fuss of wrangling bare metal servers into submission. However, not everybody has bare metal servers to work with. QuintupleO, or OpenStack on OpenStack on OpenStack, is designed to create a virtual environment where you can create a virtual OpenStack on top of OpenStack. (Did we say "OpenStack" enough in that answer?)
Do the old L2 agents work with the new ML2 plugin?
Christopher Aedo: Yes!
How hard is it to upgrade from one version of OpenStack to another, really?
Nick Chase: Depending on how you've got things architected, not as hard as people would like you to believe. Teams have been working hard to make this easier, and every project comes with a script for upgrading the database for that project. We've got a blog on one method coming up shortly, and we've just published another about an open source project, Pumphouse, that will make it easier as well.
What is Spark?
Nick Chase: Spark is Apache Spark, a data processing tool similar to Hadoop.
When will Fuel support Juno?
Christopher Aedo: Mirantis OpenStack 6.0, which includes Fuel, is due for a preview release supporting Fuel at the end of October, with general availability in November.
I have questions about HA, plz help me (maybe at the end of your presentation): (1). How could I do HA for Instances (VM). (2). Which practical solution for Bock storage. (3). How can I use multiple Public IP range . or the same idea with multiple L3 agent. (4). HA for multiple L3 agent
Christopher Aedo:
1) For help designing HA for instances, see Designing for the cloud.
2) Practical solutions for block storage: Ceph is a good choice, though there are many commercial options including SolidFire, EMC, Equalogic...
3) Public IP pools allow you to add multiple distinct ranges of IP addresses; just add them to additional pools
4) Not a question really, so I'll go with ... YES!
But what IS NFV?
Nick Chase: NFV, or Network Function Virtualization, is one level down from Software Defined Networking. Rather than defining the networks themselves, it provides programmatic control over what actually goes on in those networks, such as load balancing or firewalls. It's designed to enable chaining of these services so that large operators can remove dedicated hardware and become more agile.
Is neutron ready for production? Which opensource backend plugin is recommended, ovs?
Christopher Aedo: Yes, in fact there are several really large clouds running Neutron in production. As for which plugin, right now it's OVS but we may see this change as vendors get more involved.
Will there ever be an automatic evacuation instance feature in case of a nova-compute failure?
Christopher Aedo: It's unlikely to be a native OpenStack thing, as it would require compute nodes to use shared storage for VMs, but the tooling is there to enable this today (for instance using Ceph for vm root storage). A monitor outside the OpenStack environment can watch the health of the compute node and evacuate when appropriate.
Are the networking events hooks only for nova networking or for both nova and neutron networking?
Nick Chase: As far as I know this is only available for nova-network, but if I'm wrong, please say so in the comments.
What does L3 HA really mean?
Nick Chase: L3 agents are responsible for deploying virtual routers. Unfortunately, if a host goes down, the routers they deployed are all lost. L3 HA is the process of making sure that they stay up, and if they go down, making sure that the router functionality doesn't disappear.
What is the difference between IP assignment in IPv4 and IPv6?
Nick Chase: In IPv4, IP addresses are either static, as in assigned specifically on a device or interface, or dynamic, as in assigned by a DHCP server. In that case, the DHCP server has to keep track of what IP addresses are available. In IPv6, it's a whole other ballgame. First, the address space is MUCH larger. Second, the subnet portion is typically advertised by the router, and the device assigns itself the host portion based on its MAC address. (There are actually several ways to do IP assignment in IPv6, but that's the gist of it.)
Whats a good place to learn more about the dvr and how it works.. ?
Christopher Aedo: You ca try the OpenStack wiki, and the mailing list (https://wiki.openstack.org/wiki/Distributed_Router_for_OVS)
With dvr, who/what decides which incoming flows go to which router agent?
Nick Chase: Like normal routing, it's based on the IP addresses in the frame, with the Distributed Virtual Router IP used as the source address. You can find more information here: https://wiki.openstack.org/wiki/Neutron/DVR_L2_Agent#How_Distributed_Routing_Works
Not sure I understood what artifacts are in glance. Is there something more than images/snapshots? Any example?
Nick Chase: Really, anything an be an artifact, but Glance is meant to be involved in deploying new VMs, so perhaps the best example would be a Heat stack.
What is the status of the SSL certificate store project ?
Nick Chase: I'm assuming you're referring to Barbican, which provides keystore services. Barbican is now an incubated project.
What is your next step in integrating Mirantis OpenStack with vSphere for VMware Engineers?
Nick Chase: Watch this space.
When will we get instance tagging in Horizon?
Nick Chase: We weren't able to find a concrete answer to this question, sorry!
I have a question here, IMIO, openstack implements an elastic cloud for cattles, but for NFV, for telco appliances, it seems that openstack is used for pets, any idea?
Christopher Aedo: For general application architecture, most OpenStack architects recommend the "cattle" model where multiple VMs are considered disposable. The main focus of NFV is allowing certain VMs to have the highest performance possible in order to make virtualization of these network functions perform well; the VMs hosting the NFV components could still be treated as cattle as long as they are sharing the necessary state information (i.e. how it's done with L3 HA)
So Juno will support more than one external network interface in the tenant router?
Nick Chase: That's how I'm interpreting the spec, yes. (http://specs.openstack.org/openstack/nova-specs/specs/juno/implemented/nfv-multiple-if-1-net.html)
Will Gnocchi be supported by Gluten? Seriously, when will someone step up and start creating associative product names....! :)
Christopher Aedo: Soon, I hope.
"Gnocchi" is an Italian dish and the right pronunciation is "ñokki" :)
Nick Chase: In the words of The Doctor, "I'm sorry. I'm so sorry."
Is the EMC extremeIO cinder driver replacing EMC SMI-S?
Christopher Aedo: It does not look like it to me as far as I can tell. From what I've seen online, they're driving high IO storage access towards the ViPR driver in the future.
What's about the live migration without shared storage?
Christopher Aedo: Live migration without shared storage is possible, for instance by using block-migration under KVM. Without shared storage, there's potential for the VM to be unavailable for a longer duration, but it's still possible to migrate the VM from one compute node to another.
A question: I've surprised that ryu neutron plugin will be deprecated. Is it mean that ryu cannot be used with OpenStack? Or ryu plugin is migrating to ryu ML2 mechanism driver?
Christopher Aedo: ML2 OFAgent ML2 mechanism driver should allow you to use Ryu with OpenStack/Neutron. (https://ask.openstack.org/en/question/27704/ryu-ml2-support/)
Thanks for joining us, and feel free to ask other questions in the comments. Thanks for joining us!
