Monday, 2015-07-13

*** salv-orlando has joined #openstack-neutron-ovn00:49
*** openstackgerrit has quit IRC00:51
*** openstackgerrit has joined #openstack-neutron-ovn00:52
*** salv-orlando has quit IRC00:54
*** salv-orlando has joined #openstack-neutron-ovn02:03
*** salv-orlando has quit IRC02:08
*** salv-orlando has joined #openstack-neutron-ovn02:08
*** salv-orlando has quit IRC02:10
*** armax has joined #openstack-neutron-ovn02:52
*** armax has quit IRC03:21
*** salv-orlando has joined #openstack-neutron-ovn04:02
*** salv-orlando has quit IRC04:10
*** salv-orlando has joined #openstack-neutron-ovn04:13
*** salv-orlando has quit IRC04:18
*** salv-orlando has joined #openstack-neutron-ovn05:16
*** salv-orlando has quit IRC05:24
*** salv-orlando has joined #openstack-neutron-ovn05:25
*** salv-orlando has quit IRC05:29
*** salv-orlando has joined #openstack-neutron-ovn06:43
*** salv-orlando has quit IRC06:46
*** salv-orlando has joined #openstack-neutron-ovn06:51
*** salv-orlando has quit IRC06:56
*** salv-orlando has joined #openstack-neutron-ovn07:48
*** fzdarsky has joined #openstack-neutron-ovn08:58
*** yalie1 has left #openstack-neutron-ovn09:52
*** armax has joined #openstack-neutron-ovn14:00
*** shettyg has joined #openstack-neutron-ovn14:36
*** fzdarsky has quit IRC15:15
*** fzdarsky has joined #openstack-neutron-ovn15:21
*** russellb has quit IRC15:36
*** aluria has quit IRC15:36
*** russellb has joined #openstack-neutron-ovn15:37
*** aluria has joined #openstack-neutron-ovn15:37
*** armax has quit IRC15:48
*** cascardo has joined #openstack-neutron-ovn15:54
*** armax has joined #openstack-neutron-ovn16:01
*** armax has quit IRC16:08
*** armax has joined #openstack-neutron-ovn16:09
*** fzdarsky has quit IRC16:10
*** armax has quit IRC16:13
*** fzdarsky has joined #openstack-neutron-ovn16:14
*** armax has joined #openstack-neutron-ovn16:21
*** fzdarsky has quit IRC16:22
*** fzdarsky has joined #openstack-neutron-ovn16:24
*** fzdarsky has quit IRC16:37
*** fzdarsky has joined #openstack-neutron-ovn16:39
*** armax has quit IRC18:07
*** armax has joined #openstack-neutron-ovn18:13
*** s3wong has joined #openstack-neutron-ovn18:19
openstackgerritAaron Rosen proposed openstack/networking-ovn: DON"T REVIEW TEST..  https://review.openstack.org/20128618:32
russellbpfft, /me reviewed it anyway18:35
*** armax has quit IRC18:46
*** gsagie_ has joined #openstack-neutron-ovn18:50
*** armax has joined #openstack-neutron-ovn18:58
*** nlahouti has joined #openstack-neutron-ovn19:05
*** nlahouti has left #openstack-neutron-ovn19:06
*** arosen has joined #openstack-neutron-ovn19:11
arosenhiya russellb19:11
russellbarosen: o/19:11
gsagie_Hi Aaron19:11
arosenhi19:11
russellbarosen: just wanted to check in on your ml2 conversion ... seems to be passing tests just as well as the ml2 version19:12
russellbwant to just merge what you have?19:12
russellbor anything known not working yet?19:12
arosenrussellb: Sorry I was working on some other internal stuff. Let me hack on what i have up for a few hours and we can merge it tomorrow.19:12
russellbarosen: no problem at all!  just didn't want you to feel like you had to make it pass all tempest or anything19:13
arosenis anyone hitting this btw?19:13
russellbbecause it doesn't now19:13
arosenObtaining file:///opt/stack/networking-ovn19:13
arosen    Complete output from command python setup.py egg_info:19:13
arosen    error in setup command: Invalid environment marker: (python_version=='2.7' # Apache-2.0)19:13
arosen19:13
russellbarosen: that came up on openstack-dev ... try updating pip19:13
russellband pbr19:13
russellband whatever else, heh19:13
russellbit was something like that19:13
gsagie_טקשי פנר19:13
gsagie_yeah pbr19:14
gsagie_its also failing on neutron windows dependencies19:14
arosenrussellb: Sounds good. I wanted to make sure that everything that passed with ML2 passes with the new plugin.19:14
russellbarosen: yeah, that makes sense :)19:14
arosenThat said a lot of the stuff that is passing with ML2 doesn't really work. Just the API passes.19:14
russellb:-|19:14
arosenso i've been trying to sort through that a little.19:14
arosenyea :(19:15
russellbwell that's no good19:15
russellbsounds like the tests aren't any good then IMO19:15
arosenthe tests just test that the API is working.19:15
russellbgotcha19:16
russellboh well19:16
russellband here i was thinking we had so much working ...19:16
arosenThere are also several tests in tempest on the nova side that assume neutron implement security groups.19:16
russellbmagically19:16
arosenAnd the nova-compute <-> neutron integration only works if we have security groups :(19:16
arosenSo for now we just have to lie it looks like.19:17
arosenanyways, I'm getting closer.19:17
arosenhopefully by the end of today I should match parity with ml2.19:17
russellbnice19:17
arosenThe other thing I'm planning on working next is the docker integration stuff in  the OVN plugin.19:18
russellbarosen: the binding:profile thing?19:18
aroseni've been talking to Guru and a few other internal guys and we realized we can have the vlan allocation stuff allocated in the plugin directly rather than making the caller specifiy the vlan.19:19
russellbi never actually tested that with containers ... just verified that it configured OVN like i expected it to19:19
arosenyup19:19
russellbwell ...19:19
arosenSo one neutron change we need to make is have neutron allocate the vlans on the port that are unused.19:19
russellbyeah i guess that's fine, not a huge win because the caller still has to know about all the VLANs in use19:19
russellbso that it can configure ovs in the VM properly19:19
arosenSo the neutron api only needs to reciece the parent_port_id and we'll allocate you a free vlan19:19
arosenrussellb:  We can make it so they can also specify the vlan and if it isn't used they can use their own.19:20
russellbok, i guess if it gets auto allocated, it should (could) be in the API call response19:21
russellbso you'd know what VLAN to use without having to do another API call19:21
arosenrussellb: right but this also makes it easier for them because now they don't need to keep track of all the vlans in use on their end.19:21
russellbyeah19:21
arosenyup, i think we should put the vlan in the response.19:21
russellbagree :)19:21
russellbwfm19:21
russellbwould also love to get this integrated with kubernetes19:22
russellbthat really shows the value19:22
russellbi know some others have looked at neutron integration with k8s in general19:22
russellbmaybe docker networking support will get us there?19:23
* russellb hasn't really looked at this stuff enough19:23
russellbprovider networks are still making me shed a few tears19:23
arosenYea I think getting this integrated in other places should be pretty cool19:24
arosenshettyg: has been working on the docker network stuff19:24
russellbyeah, i've seen his stuff, but haven't tried it yet :(19:25
* russellb should start building a real team here for ovn19:25
arosenrussellb:  hrm do you have that email handy about updating pip? I just search for the error in my email and google but i'm not seeing it :(19:27
gsagie_aarosen : how is the vlan thing going to, because its the containers management system that needs to bind the containers to the correct VLAN port on OVS (in the VM hosting the containers)19:28
gsagie_and OVN/neutron has no access there, right?19:28
russellbarosen: the thing i had in mind was "[openstack-dev] [all][requirements] Whats with the "python_version=='2.7'" stuff thats showing up from today?"19:28
russellbarosen: but that doesn't seem to have the answer19:29
arosengsagie the vlan id can be passed in there along with what/ever is getting the response to start the container.19:29
russellbarosen: oh btw, check out the vlan-aware-vms spec19:29
russellbcan't remember if it got approved yet or not19:29
russellbbut i was thinking that covers our use case19:29
russellband it definitely didn't include auto vlan id allocation19:29
arosenrussellb:  Yea i think it's somewhat similar19:31
gsagie_arosen : so you are just managing VLAN allocation across local switches (or networks) in that VM right?19:32
gsagie_but you need to do it per VM19:32
arosengsagie per port.19:32
russellbarosen: i think it's our best hope of a generic API for what we want anyway19:32
arosenthat is what my thought was.19:32
gsagie_yes per port19:32
gsagie_per VM port19:32
arosenyup19:32
arosenbrb19:32
arosenrussellb:  it looks like this change is what is breaking me19:37
arosen-ovs>=2.4.0.dev0 # Apache-2.019:37
arosen+ovs>=2.4.0.dev0;python_version=='2.7' # Apache-2.019:37
russellbyep19:38
arosencommit abb9c14ffc4da5906cbfe02cf984c184da202a3f19:38
arosenAuthor: OpenStack Proposal Bot <openstack-infra@lists.openstack.org>19:38
russellbmaybe ask in -infra19:38
russellbi don't remember ..19:38
gsagie_nice idea regarding the VLAN allocations19:40
gsagie_\guess lib network support returning data for the container creation19:40
*** thumpba has joined #openstack-neutron-ovn19:42
*** armax has quit IRC19:48
*** gsagie_ has quit IRC19:51
openstackgerritAaron Rosen proposed openstack/networking-ovn: Convert plugin away from ML2  https://review.openstack.org/19532620:01
shettygI have been looking at Kubernetes integration of OVN for a month now. Without OVN providing a LB, there are are no clean integration points with Kubernetes v1.020:09
russellbshettyg: ah ok ... if it's neutron integration with OVN on the backend, we might be able to make neutron LBaaS work ... but I'm not sure how mature that is20:11
russellbnever tried it20:11
shettygI had a meeting with Tim Hockin from Google-Kubernetes and he was looking at changing the way they do LB in Kubernetes (without changing the source IP of the packet) post v1.0. If that happens, we can provide the first round of integration20:12
shettygwherein Kubernetes still does LB, OVN only provides L2 + L320:12
shettygrussellb: I also do not know how Neutron LB works or how good that is. Kubernetes needs a LB for both north-south as well as east-west traffic.20:26
russellbreference implementation of the lbaas API is called octavia ... I think it's a bit young20:28
russellbit spawns up VMs that run haproxy, basically20:28
shettygrussellb; looking at providing LB as part of OVN. Have the basic idea. But need to brainstorn on how it fits the pipeline20:31
russellbwell we don't have L3 at all in OVN yet20:31
shettygrussellb: yes, along with L3.20:31
shettygThe idea is to use conntrack to track stateful connections and then to IP address translation to LB VIP20:31
russellbgreat20:32
*** fzdarsky has quit IRC21:15
*** armax has joined #openstack-neutron-ovn21:27
*** thumpba has quit IRC21:38

Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!