Re: [openstack-dev] [Neutron] Campus Network Blueprint
Thank you for reading the BP and provide comments. My comments are inline. 2013/7/12 Zang MingJie > Hi Filipe: > > I disagree your ml2-external-port BP > > It is unsuitable to connect multiple l2 networks directly, there may > be ip conflict, dhcp conflict and other problems. although neutron > dhcp agent won't respond dhcp request from unknown source, an external > dhcp may respond vm dhcp request. About conflicts, why is that a problem? Users are suposed to control both the virtual and physical environment. If I put a DHCP server on a VM there may be DHCP conflicts as well. L3 services may then be extended to support external ports. But I'm not considering that for now. > If we move an external port form a > network to another network, how can we ensure that the arp cache is > cleared. > About the ARP tables there are two cases: VMs and physical hosts, and network equipment. About the first ones there's nothing we can or should do. The same happens if I move my laptop from a network to another. About the physical equipment if it is necesary to cleanup ARP tables to speed the process, code implementing the equipment managment can take care of that. > And it will aslo make l2-population bp ( > https://blueprints.launchpad.net/quantum/+spec/l2-population ) more > difficault. Our l2 forwarding works better if the device knows the > whole topology of the network, but the external part is totally > unknown. About your l2-population BP there will be no changes unless you want, because it isn't expected that all the network types implement external ports. If GRE or VXLAN (I think those are the ones supported by your BP) don't implement external ports you don't have to do anything as far as I can see. > So, I suggest a layer-3 solution, where the out world connects to vms > via l3 agent. > > Finally I agree L3 will fit almost allways, but there are cases where L2 is required. Currently one solution is to tunnel L2 over L3 from an external machine to a VM connected to the virtual network. But my purpose is that Neutron can provide that out of the box. > Thank you for sharing the idea > Regards > > Regards ___ > OpenStack-dev mailing list > OpenStack-dev@lists.openstack.org > http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev > ___ OpenStack-dev mailing list OpenStack-dev@lists.openstack.org http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
Re: [openstack-dev] [Neutron] Campus Network Blueprint
Hi Filipe: I disagree your ml2-external-port BP It is unsuitable to connect multiple l2 networks directly, there may be ip conflict, dhcp conflict and other problems. although neutron dhcp agent won't respond dhcp request from unknown source, an external dhcp may respond vm dhcp request. If we move an external port form a network to another network, how can we ensure that the arp cache is cleared. And it will aslo make l2-population bp ( https://blueprints.launchpad.net/quantum/+spec/l2-population ) more difficault. Our l2 forwarding works better if the device knows the whole topology of the network, but the external part is totally unknown. So, I suggest a layer-3 solution, where the out world connects to vms via l3 agent. Thank you for sharing the idea Regards ___ OpenStack-dev mailing list OpenStack-dev@lists.openstack.org http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
Re: [openstack-dev] [Neutron] Campus Network Blueprint
Hi all, As requested in the ML2 meeting I worked out a new blueprint [1] specifying a subset of the Campus Network one eventually able to hit H3. I tried to make this blueprint the most generic I could, so I changed the concepts just a little bit. If you agree with this (even if not for H3) I will change the Campus Network BP accordingly. [1] https://blueprints.launchpad.net/neutron/+spec/ml2-external-port Thank you all Regards Filipe Manco http://about.me/fmanco 2013/7/10 Filipe Manco > I will join you. Thanks! > > Filipe Manco > http://about.me/fmanco > > > 2013/7/10 Kyle Mestery (kmestery) > > On Jul 9, 2013, at 8:40 PM, Filipe Manco wrote: >> > >> > Hi Kyle >> > >> > I already looked at the blueprints and I want to integrate my work with >> ML2, mainly because I want users to able to keep the traditional networking >> models working in parallel (and I think ML2 is the best way to do this >> comparing with the meta plugin). >> > >> > To be honest, the integration with Neutron and the ML2 was what took >> most of the time when writing the blueprint, but although I talk about it >> on the blueprint I still not sure about the best way to do it. >> > >> > About the concept of segments I don't think they fit. From what I >> understand a segment, in the context of ML2, represents part of the virtual >> network that uses some technology to connect a set of VMs, so if I delete a >> segment those VMs will loose connectivity. In the context of my blueprint a >> segment (or cSegment as I called it) represents a domain where I can create >> virtual links across a set of nodes. If I delete a cSegment that doesn't >> mean any VM will loose connectivity, what will happen is that the network >> is remapped and the traffic will cross another cSegment or another set of >> cSegments. >> > >> I think what you've mentioned here is valid, yes, and is a slight >> deviation between the way segments in ML2 work on your cSegments. It looks >> like we would need to add your new constructs into the ML2 API to achieve >> what you're looking for. >> >> > Something that I would like you (or other guys from ML2) to comment is >> how to integrate new operations (beyond the ones related with the base data >> model) in the ML2 interfaces. The MechanismDriver interface supports >> operations on ports and networks, but as you can see I have new entities. >> My idea is that there should be no changes to the original interfaces, >> because this are specific to a type of network. What do you think? >> > >> I'll have a look at this tomorrow and give you more detailed feedback. If >> you want, you can join us in the ML2 meeting at 1400UTV meeting on >> #openstack-meeting and I'll save some time at the end of the meeting to >> discuss your blueprint. >> >> Thanks! >> Kyle >> >> > It would help me if you have the time to take a look at the blueprint >> and comment on the ML2 parts. >> > >> > Thanks for your help! >> > >> > Filipe Manco >> > http://about.me/fmanco >> > >> > >> > 2013/7/9 Kyle Mestery (kmestery) >> > On Jul 9, 2013, at 11:18 AM, Filipe Manco >> wrote: >> > > >> > > Hi all, >> > > >> > > Just published a blueprint for some work I'm starting right now. I >> would appreciate if someone can take a look and give some comments. >> > > >> > > https://blueprints.launchpad.net/neutron/+spec/campus-network >> > > >> > > Thank you >> > > >> > Hi Filipe: >> > >> > At first glance, there appears to be a lot of overlap with ML2 here. >> Have you looked at the ML2 blueprints, and the specific use case of >> multi-segment networks? The ML2 team is working hard to close ML2 >> blueprints in H2/H3, perhaps some of your ideas could be incorporated here? >> > >> > Here is the ML2 Meeting page which has links to the blueprints we're >> tracking: >> > >> > https://wiki.openstack.org/wiki/Meetings/ML2 >> > >> > Thanks, >> > Kyle >> > >> > > >> > > Filipe Manco >> > > http://about.me/fmanco >> > > ___ >> > > OpenStack-dev mailing list >> > > OpenStack-dev@lists.openstack.org >> > > http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev >> > >> > >> > >> > ___ >> > OpenStack-dev mailing list >> > OpenStack-dev@lists.openstack.org >> > http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev >> > >> > ___ >> > OpenStack-dev mailing list >> > OpenStack-dev@lists.openstack.org >> > http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev >> >> >> >> ___ >> OpenStack-dev mailing list >> OpenStack-dev@lists.openstack.org >> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev >> > > ___ OpenStack-dev mailing list OpenStack-dev@lists.openstack.org http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
Re: [openstack-dev] [Neutron] Campus Network Blueprint
I will join you. Thanks! Filipe Manco http://about.me/fmanco 2013/7/10 Kyle Mestery (kmestery) > On Jul 9, 2013, at 8:40 PM, Filipe Manco wrote: > > > > Hi Kyle > > > > I already looked at the blueprints and I want to integrate my work with > ML2, mainly because I want users to able to keep the traditional networking > models working in parallel (and I think ML2 is the best way to do this > comparing with the meta plugin). > > > > To be honest, the integration with Neutron and the ML2 was what took > most of the time when writing the blueprint, but although I talk about it > on the blueprint I still not sure about the best way to do it. > > > > About the concept of segments I don't think they fit. From what I > understand a segment, in the context of ML2, represents part of the virtual > network that uses some technology to connect a set of VMs, so if I delete a > segment those VMs will loose connectivity. In the context of my blueprint a > segment (or cSegment as I called it) represents a domain where I can create > virtual links across a set of nodes. If I delete a cSegment that doesn't > mean any VM will loose connectivity, what will happen is that the network > is remapped and the traffic will cross another cSegment or another set of > cSegments. > > > I think what you've mentioned here is valid, yes, and is a slight > deviation between the way segments in ML2 work on your cSegments. It looks > like we would need to add your new constructs into the ML2 API to achieve > what you're looking for. > > > Something that I would like you (or other guys from ML2) to comment is > how to integrate new operations (beyond the ones related with the base data > model) in the ML2 interfaces. The MechanismDriver interface supports > operations on ports and networks, but as you can see I have new entities. > My idea is that there should be no changes to the original interfaces, > because this are specific to a type of network. What do you think? > > > I'll have a look at this tomorrow and give you more detailed feedback. If > you want, you can join us in the ML2 meeting at 1400UTV meeting on > #openstack-meeting and I'll save some time at the end of the meeting to > discuss your blueprint. > > Thanks! > Kyle > > > It would help me if you have the time to take a look at the blueprint > and comment on the ML2 parts. > > > > Thanks for your help! > > > > Filipe Manco > > http://about.me/fmanco > > > > > > 2013/7/9 Kyle Mestery (kmestery) > > On Jul 9, 2013, at 11:18 AM, Filipe Manco > wrote: > > > > > > Hi all, > > > > > > Just published a blueprint for some work I'm starting right now. I > would appreciate if someone can take a look and give some comments. > > > > > > https://blueprints.launchpad.net/neutron/+spec/campus-network > > > > > > Thank you > > > > > Hi Filipe: > > > > At first glance, there appears to be a lot of overlap with ML2 here. > Have you looked at the ML2 blueprints, and the specific use case of > multi-segment networks? The ML2 team is working hard to close ML2 > blueprints in H2/H3, perhaps some of your ideas could be incorporated here? > > > > Here is the ML2 Meeting page which has links to the blueprints we're > tracking: > > > > https://wiki.openstack.org/wiki/Meetings/ML2 > > > > Thanks, > > Kyle > > > > > > > > Filipe Manco > > > http://about.me/fmanco > > > ___ > > > OpenStack-dev mailing list > > > OpenStack-dev@lists.openstack.org > > > http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev > > > > > > > > ___ > > OpenStack-dev mailing list > > OpenStack-dev@lists.openstack.org > > http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev > > > > ___ > > OpenStack-dev mailing list > > OpenStack-dev@lists.openstack.org > > http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev > > > > ___ > OpenStack-dev mailing list > OpenStack-dev@lists.openstack.org > http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev > ___ OpenStack-dev mailing list OpenStack-dev@lists.openstack.org http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
Re: [openstack-dev] [Neutron] Campus Network Blueprint
On Jul 9, 2013, at 8:40 PM, Filipe Manco wrote: > > Hi Kyle > > I already looked at the blueprints and I want to integrate my work with ML2, > mainly because I want users to able to keep the traditional networking models > working in parallel (and I think ML2 is the best way to do this comparing > with the meta plugin). > > To be honest, the integration with Neutron and the ML2 was what took most of > the time when writing the blueprint, but although I talk about it on the > blueprint I still not sure about the best way to do it. > > About the concept of segments I don't think they fit. From what I understand > a segment, in the context of ML2, represents part of the virtual network that > uses some technology to connect a set of VMs, so if I delete a segment those > VMs will loose connectivity. In the context of my blueprint a segment (or > cSegment as I called it) represents a domain where I can create virtual links > across a set of nodes. If I delete a cSegment that doesn't mean any VM will > loose connectivity, what will happen is that the network is remapped and the > traffic will cross another cSegment or another set of cSegments. > I think what you've mentioned here is valid, yes, and is a slight deviation between the way segments in ML2 work on your cSegments. It looks like we would need to add your new constructs into the ML2 API to achieve what you're looking for. > Something that I would like you (or other guys from ML2) to comment is how to > integrate new operations (beyond the ones related with the base data model) > in the ML2 interfaces. The MechanismDriver interface supports operations on > ports and networks, but as you can see I have new entities. My idea is that > there should be no changes to the original interfaces, because this are > specific to a type of network. What do you think? > I'll have a look at this tomorrow and give you more detailed feedback. If you want, you can join us in the ML2 meeting at 1400UTV meeting on #openstack-meeting and I'll save some time at the end of the meeting to discuss your blueprint. Thanks! Kyle > It would help me if you have the time to take a look at the blueprint and > comment on the ML2 parts. > > Thanks for your help! > > Filipe Manco > http://about.me/fmanco > > > 2013/7/9 Kyle Mestery (kmestery) > On Jul 9, 2013, at 11:18 AM, Filipe Manco wrote: > > > > Hi all, > > > > Just published a blueprint for some work I'm starting right now. I would > > appreciate if someone can take a look and give some comments. > > > > https://blueprints.launchpad.net/neutron/+spec/campus-network > > > > Thank you > > > Hi Filipe: > > At first glance, there appears to be a lot of overlap with ML2 here. Have you > looked at the ML2 blueprints, and the specific use case of multi-segment > networks? The ML2 team is working hard to close ML2 blueprints in H2/H3, > perhaps some of your ideas could be incorporated here? > > Here is the ML2 Meeting page which has links to the blueprints we're tracking: > > https://wiki.openstack.org/wiki/Meetings/ML2 > > Thanks, > Kyle > > > > > Filipe Manco > > http://about.me/fmanco > > ___ > > OpenStack-dev mailing list > > OpenStack-dev@lists.openstack.org > > http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev > > > > ___ > OpenStack-dev mailing list > OpenStack-dev@lists.openstack.org > http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev > > ___ > OpenStack-dev mailing list > OpenStack-dev@lists.openstack.org > http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev ___ OpenStack-dev mailing list OpenStack-dev@lists.openstack.org http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
Re: [openstack-dev] [Neutron] Campus Network Blueprint
Hi Kyle I already looked at the blueprints and I want to integrate my work with ML2, mainly because I want users to able to keep the traditional networking models working in parallel (and I think ML2 is the best way to do this comparing with the meta plugin). To be honest, the integration with Neutron and the ML2 was what took most of the time when writing the blueprint, but although I talk about it on the blueprint I still not sure about the best way to do it. About the concept of segments I don't think they fit. From what I understand a segment, in the context of ML2, represents part of the virtual network that uses some technology to connect a set of VMs, so if I delete a segment those VMs will loose connectivity. In the context of my blueprint a segment (or cSegment as I called it) represents a domain where I can create virtual links across a set of nodes. If I delete a cSegment that doesn't mean any VM will loose connectivity, what will happen is that the network is remapped and the traffic will cross another cSegment or another set of cSegments. Something that I would like you (or other guys from ML2) to comment is how to integrate new operations (beyond the ones related with the base data model) in the ML2 interfaces. The MechanismDriver interface supports operations on ports and networks, but as you can see I have new entities. My idea is that there should be no changes to the original interfaces, because this are specific to a type of network. What do you think? It would help me if you have the time to take a look at the blueprint and comment on the ML2 parts. Thanks for your help! Filipe Manco http://about.me/fmanco 2013/7/9 Kyle Mestery (kmestery) > On Jul 9, 2013, at 11:18 AM, Filipe Manco wrote: > > > > Hi all, > > > > Just published a blueprint for some work I'm starting right now. I would > appreciate if someone can take a look and give some comments. > > > > https://blueprints.launchpad.net/neutron/+spec/campus-network > > > > Thank you > > > Hi Filipe: > > At first glance, there appears to be a lot of overlap with ML2 here. Have > you looked at the ML2 blueprints, and the specific use case of > multi-segment networks? The ML2 team is working hard to close ML2 > blueprints in H2/H3, perhaps some of your ideas could be incorporated here? > > Here is the ML2 Meeting page which has links to the blueprints we're > tracking: > > https://wiki.openstack.org/wiki/Meetings/ML2 > > Thanks, > Kyle > > > > > Filipe Manco > > http://about.me/fmanco > > ___ > > OpenStack-dev mailing list > > OpenStack-dev@lists.openstack.org > > http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev > > > > ___ > OpenStack-dev mailing list > OpenStack-dev@lists.openstack.org > http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev > ___ OpenStack-dev mailing list OpenStack-dev@lists.openstack.org http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
Re: [openstack-dev] [Neutron] Campus Network Blueprint
On Jul 9, 2013, at 11:18 AM, Filipe Manco wrote: > > Hi all, > > Just published a blueprint for some work I'm starting right now. I would > appreciate if someone can take a look and give some comments. > > https://blueprints.launchpad.net/neutron/+spec/campus-network > > Thank you > Hi Filipe: At first glance, there appears to be a lot of overlap with ML2 here. Have you looked at the ML2 blueprints, and the specific use case of multi-segment networks? The ML2 team is working hard to close ML2 blueprints in H2/H3, perhaps some of your ideas could be incorporated here? Here is the ML2 Meeting page which has links to the blueprints we're tracking: https://wiki.openstack.org/wiki/Meetings/ML2 Thanks, Kyle > > Filipe Manco > http://about.me/fmanco > ___ > OpenStack-dev mailing list > OpenStack-dev@lists.openstack.org > http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev ___ OpenStack-dev mailing list OpenStack-dev@lists.openstack.org http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
[openstack-dev] [Neutron] Campus Network Blueprint
Hi all, Just published a blueprint for some work I'm starting right now. I would appreciate if someone can take a look and give some comments. https://blueprints.launchpad.net/neutron/+spec/campus-network Thank you Filipe Manco http://about.me/fmanco ___ OpenStack-dev mailing list OpenStack-dev@lists.openstack.org http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev