Re: [openstack-dev] [neutron] IPv4 transition/interoperation with IPv6
On Wed, 2015-05-06 at 02:46 -0400, Mike Spreitzer wrote: While I am a Neutron operator, I am also a customer of a lower layer network provider. That network provider will happily give me a few /64. How do I serve IPv6 subnets to lots of my tenants? In the bad old v4 days this would be easy: a tenant puts all his stuff on his private networks and NATs (e.g., floating IP) his edge servers onto a public network --- no need to align tenant private subnets with public subnets. But with no NAT for v6, there is no public/private distinction --- I can only give out the public v6 subnets that I am given. Yes, NAT is bad. But not being able to get your job done is worse. I would suggest that you talk to your network provider, or apply to your local RIR to obtain Provider Independent address space. It should be relatively trivial to obtain a significant amount of IPv6 address space. Trying to make this work with only a few /64s is going to lead to a world of pain. Cheers, Andrew -- Andrew Ruthven, Wellington, New Zealand and...@etc.gen.nz | linux.conf.au 2015 New Zealand's only Cloud: | BeAwesome in Auckland, NZ https://catalyst.net.nz/cloud | http://lca2015.linux.org.au __ OpenStack Development Mailing List (not for usage questions) Unsubscribe: openstack-dev-requ...@lists.openstack.org?subject:unsubscribe http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
Re: [openstack-dev] [neutron] IPv4 transition/interoperation with IPv6
On the subject of Prefix Delegation - yes, the external system is responsible for the routing. Here¹s a couple of video guides on using PD in Neutron and setting up the Prefix Delegation Server (in this case a dibbler server): Using Neutron PD: https://www.youtube.com/watch?v=wI830s881HQ Configuring the PD server: https://www.youtube.com/watch?v=zfsFyS01Fn0 The patch is up for review at: https://review.openstack.org/#/c/158697 And the networking guide docs: https://review.openstack.org/#/c/178739 John On 06/05/2015 17:57, Carl Baldwin c...@ecbaldwin.net wrote: On Wed, May 6, 2015 at 12:46 AM, Mike Spreitzer mspre...@us.ibm.com wrote: While I am a Neutron operator, I am also a customer of a lower layer network provider. That network provider will happily give me a few /64. How do I serve IPv6 subnets to lots of my tenants? In the bad old v4 days this would be easy: a tenant puts all his stuff on his private networks and NATs (e.g., floating IP) his edge servers onto a public network --- no need to align tenant private subnets with public subnets. But with no NAT for v6, there is no public/private distinction --- I can only give out the public v6 subnets that I am given. Yes, NAT is bad. But not being able to get your job done is worse. Mike, in this paragraph, you're hitting on something that has been on my mind for a while. We plan to cover this problem in detail in this talk [1] and we're defining some work for Liberty to better address it [2][3]. You hit the nail on the head, there is no distinguishing private and public IP addresses in Neutron currently with IPv6. Kilo's new subnet pool feature is a start. It will allow you to create a shared subnet pool including the /64s from your service provider. Tenants can then create a subnet getting an allocation from it automatically. However, given the current state of things, there will be some manual work on the gateway router to route them to the tenant's router. Prefix delegation -- which looks on track for Liberty -- is another option which could fill this void. It will allow a router to get a prefix delegation from an external PD system which will be useable on a tenant subnet. Presumably the external system will take care of routing the subnet to the appropriate tenant router. Carl [1] http://sched.co/2qdm [2] https://review.openstack.org/#/c/180267/ [3] https://review.openstack.org/#/c/125401/ __ OpenStack Development Mailing List (not for usage questions) Unsubscribe: openstack-dev-requ...@lists.openstack.org?subject:unsubscribe http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev __ OpenStack Development Mailing List (not for usage questions) Unsubscribe: openstack-dev-requ...@lists.openstack.org?subject:unsubscribe http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
Re: [openstack-dev] [neutron] IPv4 transition/interoperation with IPv6
Robert Li (baoli) ba...@cisco.com wrote on 05/05/2015 09:02:08 AM: Currently dual stack is supported. Can you be specific on what interoperation/transition techniques you are interested in? We’ve been thinking about NAT64 (stateless or stateful). thanks, Robert On 5/4/15, 9:56 PM, Mike Spreitzer mspre...@us.ibm.com wrote: Does Neutron support any of the 4/6 interoperation/transition techniques? I wear an operator's hat nowadays, and want to make IPv6 as useful and easy to use as possible for my tenants. I think the interoperation/transition techniques will play a big role in this. Is dual stacking working in routers now? At the moment I am still using Juno, but plan to move to Kilo. I want to encourage my tenants to use as much IPv6 as possible. But I expect some will have to keep some of their workload on v4 (I know there is on-going work to get many application frameworks up to v6 speed, and it is not complete yet). I expect some tenants could be mixed: some workload on v4 and some on v6. Such a tenant would appreciate a NAT between his v6 space and his v4 space. This is the easiest cases --- sections 2.5 and 2.6 --- of RFC 6144. I would prefer to do it in a stateless way if possible. That would be pretty easy if Neutron and Nova were willing to accept an IPv6 subnet that is much smaller than 2^64 addresses. I see that my macs differ only in their last 24 bits. Some tenants could put their entire workload on v6, but that workload would be unreachable from customers of all those ISPs (such as mine, CableVision) that deny IPv6 service to their customers. There are techniques for coping, and Teredo looks like a pretty good one. It has been shipped in Windows for years. Yet I can not find a Windows machine where the Teredo actually works. What's up with that? If Windows somehow got its Teredo, or other, act together, that would be only half the job; Teredo requires something from the server side as well, right? Supposing a focus on mobile, where IPv6 is much more available, and/or progress by Microsoft and/or other ISPs, my tenant might face a situation where his clients could come in over v6 but some of his servers still have to run on v4. That's section 2.3 of RFC 6144. While I am a Neutron operator, I am also a customer of a lower layer network provider. That network provider will happily give me a few /64. How do I serve IPv6 subnets to lots of my tenants? In the bad old v4 days this would be easy: a tenant puts all his stuff on his private networks and NATs (e.g., floating IP) his edge servers onto a public network --- no need to align tenant private subnets with public subnets. But with no NAT for v6, there is no public/private distinction --- I can only give out the public v6 subnets that I am given. Yes, NAT is bad. But not being able to get your job done is worse. Sean M. Collins s...@coreitpro.com wrote on 05/05/2015 06:26:28 AM: I think that Neutron exposes enough primitives through the API that advanced services for handling your transition technique of choice could be built. I think that is right, if I am willing to assume Neutron is using OVS --- or build a bunch of alternatives that correspond to all the Neutron plugins and mechanisms that I might encounter. And it would feel a lot like Neutron implementation work. Really, it is one instance of doing some NFV. Thanks, Mike __ OpenStack Development Mailing List (not for usage questions) Unsubscribe: openstack-dev-requ...@lists.openstack.org?subject:unsubscribe http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
Re: [openstack-dev] [neutron] IPv4 transition/interoperation with IPv6
On Wed, May 6, 2015 at 12:46 AM, Mike Spreitzer mspre...@us.ibm.com wrote: While I am a Neutron operator, I am also a customer of a lower layer network provider. That network provider will happily give me a few /64. How do I serve IPv6 subnets to lots of my tenants? In the bad old v4 days this would be easy: a tenant puts all his stuff on his private networks and NATs (e.g., floating IP) his edge servers onto a public network --- no need to align tenant private subnets with public subnets. But with no NAT for v6, there is no public/private distinction --- I can only give out the public v6 subnets that I am given. Yes, NAT is bad. But not being able to get your job done is worse. Mike, in this paragraph, you're hitting on something that has been on my mind for a while. We plan to cover this problem in detail in this talk [1] and we're defining some work for Liberty to better address it [2][3]. You hit the nail on the head, there is no distinguishing private and public IP addresses in Neutron currently with IPv6. Kilo's new subnet pool feature is a start. It will allow you to create a shared subnet pool including the /64s from your service provider. Tenants can then create a subnet getting an allocation from it automatically. However, given the current state of things, there will be some manual work on the gateway router to route them to the tenant's router. Prefix delegation -- which looks on track for Liberty -- is another option which could fill this void. It will allow a router to get a prefix delegation from an external PD system which will be useable on a tenant subnet. Presumably the external system will take care of routing the subnet to the appropriate tenant router. Carl [1] http://sched.co/2qdm [2] https://review.openstack.org/#/c/180267/ [3] https://review.openstack.org/#/c/125401/ __ OpenStack Development Mailing List (not for usage questions) Unsubscribe: openstack-dev-requ...@lists.openstack.org?subject:unsubscribe http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
Re: [openstack-dev] [neutron] IPv4 transition/interoperation with IPv6
Hi Mike, Currently dual stack is supported. Can you be specific on what interoperation/transition techniques you are interested in? We’ve been thinking about NAT64 (stateless or stateful). thanks, Robert On 5/4/15, 9:56 PM, Mike Spreitzer mspre...@us.ibm.commailto:mspre...@us.ibm.com wrote: Does Neutron support any of the 4/6 interoperation/transition techniques? I wear an operator's hat nowadays, and want to make IPv6 as useful and easy to use as possible for my tenants. I think the interoperation/transition techniques will play a big role in this. Thanks, Mike __ OpenStack Development Mailing List (not for usage questions) Unsubscribe: openstack-dev-requ...@lists.openstack.org?subject:unsubscribe http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
Re: [openstack-dev] [neutron] IPv4 transition/interoperation with IPv6
I think that Neutron exposes enough primitives through the API that advanced services for handling your transition technique of choice could be built. -- Sean M. Collins __ OpenStack Development Mailing List (not for usage questions) Unsubscribe: openstack-dev-requ...@lists.openstack.org?subject:unsubscribe http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
[openstack-dev] [neutron] IPv4 transition/interoperation with IPv6
Does Neutron support any of the 4/6 interoperation/transition techniques? I wear an operator's hat nowadays, and want to make IPv6 as useful and easy to use as possible for my tenants. I think the interoperation/transition techniques will play a big role in this. Thanks, Mike__ OpenStack Development Mailing List (not for usage questions) Unsubscribe: openstack-dev-requ...@lists.openstack.org?subject:unsubscribe http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev