Tuesday, 2016-09-06

openstackgerritZongKai LI proposed openstack/networking-ovn: Update segment_data for TestOvnSbSync  https://review.openstack.org/36581401:49
*** armax has quit IRC02:13
*** salv-orl_ has joined #openstack-neutron-ovn02:29
*** salv-orlando has quit IRC02:32
*** dkehn_ has joined #openstack-neutron-ovn02:44
*** dkehn_ has quit IRC03:49
*** dkehn_ has joined #openstack-neutron-ovn03:51
*** mickeys has joined #openstack-neutron-ovn04:49
*** mickeys has quit IRC04:49
*** irenab has joined #openstack-neutron-ovn05:04
*** irenab_ has joined #openstack-neutron-ovn05:06
*** irenab has quit IRC05:08
*** irenab_ is now known as irenab05:08
*** mmirecki has joined #openstack-neutron-ovn06:41
*** pcaruana has joined #openstack-neutron-ovn06:48
*** dkehn_ has quit IRC06:56
*** dkehn_ has joined #openstack-neutron-ovn07:01
openstackgerritBabu Shanmugam proposed openstack/networking-ovn: OVN trunk driver to support vlan-aware-vms  https://review.openstack.org/35640907:09
*** fzdarsky has joined #openstack-neutron-ovn07:22
openstackgerritZongKai LI proposed openstack/networking-ovn: Add sync support for DHCP_Options  https://review.openstack.org/35220908:00
*** gongysh has quit IRC08:25
*** gongysh has joined #openstack-neutron-ovn08:26
*** salv-orlando has joined #openstack-neutron-ovn08:29
*** salv-orl_ has quit IRC08:32
*** roeyc has joined #openstack-neutron-ovn08:32
*** gongysh has quit IRC09:19
*** gongysh has joined #openstack-neutron-ovn09:21
*** gongysh has quit IRC09:26
*** roeyc has quit IRC09:38
*** roeyc has joined #openstack-neutron-ovn10:18
*** roeyc has quit IRC10:38
*** roeyc has joined #openstack-neutron-ovn10:45
*** HenryG_ is now known as HenryG10:46
*** roeyc has quit IRC11:14
*** rtheis has joined #openstack-neutron-ovn11:22
*** dkehn_ has quit IRC11:23
*** roeyc has joined #openstack-neutron-ovn11:23
*** dkehn_ has joined #openstack-neutron-ovn11:35
*** salv-orl_ has joined #openstack-neutron-ovn11:52
*** salv-orlando has quit IRC11:52
*** zefferno has joined #openstack-neutron-ovn12:38
*** tongli has joined #openstack-neutron-ovn12:40
*** pcaruana has quit IRC12:50
rtheismestery, Sam-I-Am, russellb: n-ovn gate fix for recent update to neutron network segment support: https://review.openstack.org/#/c/365814/12:57
russellbrtheis: approved12:58
rtheisrussellb: thank you12:58
russellbthanks for bringing it up12:58
*** brad_behle has joined #openstack-neutron-ovn13:27
*** zkassab has joined #openstack-neutron-ovn13:31
*** irenab has quit IRC13:31
*** irenab has joined #openstack-neutron-ovn13:32
rtheisFYI: n-ovn meeting weekly meetings start on Thursday: https://review.openstack.org/36390613:34
*** woodburn has joined #openstack-neutron-ovn13:37
*** roeyc has quit IRC13:46
*** roeyc has joined #openstack-neutron-ovn13:48
russellbi'll be in an all day meeting, but hopefully will be able to  join anyway since it's irc13:50
*** mlavalle has joined #openstack-neutron-ovn13:51
*** gongysh has joined #openstack-neutron-ovn13:52
openstackgerritMerged openstack/networking-ovn: Update segment_data for TestOvnSbSync  https://review.openstack.org/36581413:55
*** brad_behle has quit IRC14:00
*** banix has joined #openstack-neutron-ovn14:11
*** gongysh has quit IRC14:20
*** salv-orlando has joined #openstack-neutron-ovn14:29
*** salv-orl_ has quit IRC14:32
openstackgerritZongKai LI proposed openstack/networking-ovn: Add sync support for DHCP_Options  https://review.openstack.org/35220914:37
openstackgerritZongKai LI proposed openstack/networking-ovn: Add DB sync support for lrouter port networks  https://review.openstack.org/35850114:38
openstackgerritZongKai LI proposed openstack/networking-ovn: Enhance _get_port_dhcpv4_options  https://review.openstack.org/36272614:39
*** lrichard__ has joined #openstack-neutron-ovn14:54
*** lrichard_ has joined #openstack-neutron-ovn14:54
*** lrichard__ has quit IRC14:54
*** dkehn_ has quit IRC14:57
*** cryptarium has joined #openstack-neutron-ovn15:06
*** dkehn_ has joined #openstack-neutron-ovn15:10
*** regXboi has joined #openstack-neutron-ovn15:26
*** zefferno has quit IRC15:29
*** doonhammer has joined #openstack-neutron-ovn15:29
*** armax has joined #openstack-neutron-ovn15:29
*** doonhammer has quit IRC15:35
*** pcaruana has joined #openstack-neutron-ovn15:42
*** dkehn__ has joined #openstack-neutron-ovn15:54
*** dkehn__ has quit IRC16:09
*** dkehn__ has joined #openstack-neutron-ovn16:29
*** lrichard has quit IRC16:31
*** lrichard has joined #openstack-neutron-ovn16:34
*** mickeys has joined #openstack-neutron-ovn16:38
*** mickeys has quit IRC16:40
*** mickeys has joined #openstack-neutron-ovn16:40
*** chandrav has joined #openstack-neutron-ovn16:41
*** mickeys has quit IRC16:45
*** roeyc has quit IRC16:47
*** chandrav has quit IRC16:55
*** chandrav has joined #openstack-neutron-ovn16:59
*** fzdarsky is now known as fzdarsky|afk17:01
*** chandrav has quit IRC17:03
*** armax has quit IRC17:11
*** armax has joined #openstack-neutron-ovn17:26
*** banix has quit IRC17:49
*** banix has joined #openstack-neutron-ovn17:51
*** chandrav has joined #openstack-neutron-ovn17:52
*** chandrav has quit IRC17:55
*** banix has quit IRC18:00
*** s3wong has joined #openstack-neutron-ovn18:02
*** dkehn__ has quit IRC18:03
*** dkehn_ has quit IRC18:03
*** banix has joined #openstack-neutron-ovn18:03
*** chandrav has joined #openstack-neutron-ovn18:09
*** regXboi has quit IRC18:10
*** regXboi has joined #openstack-neutron-ovn18:13
*** dkehn__ has joined #openstack-neutron-ovn18:15
*** dkehn_ has joined #openstack-neutron-ovn18:15
*** chandrav has quit IRC18:20
*** mmirecki has quit IRC18:51
*** banix has quit IRC19:02
*** fzdarsky|afk is now known as fzdarsky19:08
*** pcaruana has quit IRC19:11
*** banix has joined #openstack-neutron-ovn19:20
regXboi@russellb: if you are around - has anybody talked about how to represent the segments that make up a routed network in one of OVN's logical switch structure?19:36
regXboi(er structures)19:36
russellbi thought there have been people working on that already19:40
russellband we've merged a number of patches related to that19:40
russellbi don't recall the current state exactly though ...19:40
russellbgoal was to have routed networks support developed in parallel with core neutron19:40
regXboiwell, I see a big TODO in the code and I didn't see any patches in the queue that addressed it19:42
regXboihence my question :) - a casual glance makes me think that segments should map to logical switches and have the LS decorated with the owning network UUID in either the other_config or external_ids column19:45
regXboirtheis: you are the author of the TODO comment in the mech_driver - am I missing something?19:45
rtheisregXboi: segments map indirectly to logical switches.  segment is part of a network which has a logical switch19:47
rtheisThe TODO is for multi-provider network option which I'm not sure if we need for routed networks19:47
rtheiscarl_baldwin: does networking-ovn need to support the multi-provider network option for routed networks?19:48
carl_baldwinrtheis: Kind of but not strictly. In order to support routed networks, it will need to support multiple segments. There are now two ways to create multi-segment networks. The multi-provider network extension and create through the segments extension.19:51
carl_baldwinrtheis: So, by the time OVN supports multiple segments, it might as well support the multi-provider extension.19:51
rtheiscarl_baldwin: thanks for the explanation.  Do you know if Hong Hui was planning to work on this?  I added the TODO at his request.19:52
carl_baldwinWhen xiaohhui was at IBM, he was planning to work on this. He has switched companies and I'm not sure if his priorities have changed.19:53
rtheisI'll have to ask him19:54
russellbah, i see19:56
russellbi think that's who i was assuming was doing the ovn work :)19:56
regXboirussellb: and that's why I was asking about it - don't want to see the ball get dropped19:57
carl_baldwinregXboi: Thanks for looking out for this. I was getting a little worried myself.20:02
regXboicarl_baldwin: well, I'm still a bit scared because I'm thinking through the upgrade use case for provider networks and getting really worried :(20:02
carl_baldwinregXboi: upgrade from what?20:02
carl_baldwinregXboi: I probably just don't understand the upgrade scenario that you're think about.20:03
regXboicarl_baldwin: assume that I start with a cloud with a provider network with one segment today20:03
regXboinow I add a second segment to the provider network - what happens?20:03
russellbcarl_baldwin: how's the core neutron stuff going, all done?20:04
carl_baldwinregXboi: The first problem is that the subnets on the original provider network are not attached to any segment (even the original one).20:05
carl_baldwinregXboi: Currently, a subnet can only be associated with a segment on subnet create. So, we might need to think about relaxing this constrint.20:06
carl_baldwinrussellb: It is done to a point. The big missing piece is that we don't have any Nova scheduling support in Newton. It is looking like we'll get that in Ocata though.20:07
regXboi@carl_baldwin: agreed - specifically, I am getting hung up on the phyiscal_interface_mappings setting in the ml2_conf.ini and how to override it20:07
russellbcarl_baldwin: cool, well congrats on the progress20:07
carl_baldwinrussellb: Thanks, it has been a pretty fun project. It should be considered experimental in Newton but what's there works pretty well as far as I can tell.20:08
russellbi wonder if we can sneak in the rest for OVN, if it's not done ...20:08
russellbi thought it was though, for some reason20:08
*** cryptarium has quit IRC20:08
regXboi@russellb: I think it's going to be tricky20:08
carl_baldwinrussellb: I'm sorry that I didn't stay current on the OVN side of the work. I remember it was going well to a point.20:09
regXboirussellb: I"m trying to figure out how to test it right now with OVN and having questions :)20:09
russellbno worries, not your task20:09
carl_baldwinregXboi: If you want to convert an existing single-segment provider network to a routed network, we'll have to provide something for that.20:09
rtheisrussellb: I just tested creating a network and adding a segment to it.  That works fine, but we are missing the multi-provider extension piece.  I don't think that should be too hard to add20:10
regXboicarl_baldwin: unfortunately, that is exactly the scenario that I'm looking at20:10
*** banix has quit IRC20:11
regXboior I should say trying to figure out how to test20:11
rtheisregXboi: how about adding your upgrade case to https://etherpad.openstack.org/p/routed-provider-networks-notes for consideration in Ocata?20:11
regXboirtheis: sure I'll add it20:12
carl_baldwinregXboi: It should be doable. We just can't blindly allow attaching an existing subnet to a new segment. It will take just a little thinking.20:13
regXboicarl_baldwin: ok - that brings me back to the question of the ovs bridges, etc.20:14
regXboirtheis: upgrade case added at the bottom20:16
regXboiI haven't put any ideas of what needs to happen since I'm not 100% sure how to state them yet :)20:18
regXboicarl_baldwin: I"m going to wander away and think about how to make the original network the first segment of a new network (as I think that comes before the subnet question)20:18
carl_baldwinregXboi: It can't be that hard. I could convert a provider network to a routed network with one command in the DB manually. We just have to make sure that we have the right checks in place to do it automatically.20:20
carl_baldwinregXboi: Thanks for adding that.20:21
carl_baldwin... to the etherpad I mean.20:21
regXboicarl_baldwin: Ok, I've put a couple of steps in the etherpad as well20:21
regXboior at least what I think some of the steps are...20:21
carl_baldwinregXboi: I'm already marking them up with my thoughts.20:25
regXboiand ditto replies :)20:25
carl_baldwinregXboi: I'm trying to wrap my head around that last concern. Let me know if I'm still missing some understanding.20:28
regXboicarl_baldwiN: We are getting there20:29
* regXboi can't type20:29
regXboicarl_baldwin: ^^^20:29
carl_baldwinregXboi: Should be discuss here instead of etherpad?20:30
*** salv-orl_ has joined #openstack-neutron-ovn20:30
regXboiyeah, this is logged as well :)20:30
regXboiin fact, it's logged better20:30
regXboicarl_baldwin: what I'm trying to get to is doing all of this *AND* minimize hits20:31
carl_baldwinregXboi: With how things are today, you would have to restart the compute node. But, you'd also have to live migrate existing instances that are on the original segment away from the compute node in order to be able to move it to another segment.20:31
regXboiso I really don't want to have to change ml2_conf.ini files because that's going to get very messy20:31
regXboiyeah, I'm not looking to move a compute from one segment to another20:31
regXboiI'm trying to wrap my head around what is going to happen when the second segment gets added20:32
carl_baldwinThen the only reason to add a new segment is to add new compute nodes (or am I missing another reason).20:32
regXboiyou are missing another reason20:32
*** salv-orlando has quit IRC20:33
* carl_baldwin is on the edge of his seat. :)20:33
regXboisorry - interruption20:33
*** mmirecki has joined #openstack-neutron-ovn20:33
regXboiright now I'm looking at deploying a 400 compute node cloud with a single provider network - I want to convert that to a routed network and add three segments to get to the point of having a segment go to (on average) 25% of the cloud20:34
regXboibecause I can do some things to help me at that point20:34
regXboi*but* I'm not allowed to redeploy the cloud in anything other than an upgrade path20:34
regXboithe "things that help me at that point" == OVN's conditional monitoring20:35
regXboithe break-even point for that is 80% of the cloud - so I can't use it right now, but I can once I have routed networks20:35
regXboiand that's the reason for adding a new segment that doesn't involve adding new computes20:36
regXboibtw - the "three" above is just a nice round number I pulled out of the sky - if I can add "one" network then I get the win20:36
carl_baldwinregXboi: Let me see if I follow...20:36
carl_baldwinIn the original provider network, you won't use all 400 computes?20:37
regXboiI have to assume that some number of computes will be bound to the original provider network and I'm not looking to change that20:37
carl_baldwinI'm not sure I understand what you said about the "break-even point"20:37
regXboiIf I have a network that is bound to less than 80% of my cloud, then conditional monitoring buys me a performance improvement20:38
regXboiand if it is more than 80%, I pay a penalty.20:38
regXboiso, when I have a single provider network that *might* go to all of the nodes of the cloud, I can't turn it on20:39
*** chandrav has joined #openstack-neutron-ovn20:39
regXboibut if I can convert to routed networks and add a single segment, then I can turn it on20:39
regXboiI'm hoping to do this before too many of the compute nodes are bound to the original provider network20:40
regXboiand with that - I need to wander away for today20:42
regXboiwe'll pick this conversation up later this week20:42
carl_baldwinregXboi: I don't fully understand yet. We can talk later.20:42
* regXboi wanders off into the not-yet sunset20:42
*** regXboi has quit IRC20:42
*** mmirecki has quit IRC20:46
*** fzdarsky has quit IRC20:53
*** s3wong has quit IRC20:56
openstackgerritRichard Theis proposed openstack/networking-ovn: Add unit tests for neutron-ovn-db-sync-util command  https://review.openstack.org/36508820:59
*** s3wong has joined #openstack-neutron-ovn21:08
*** rtheis has quit IRC21:33
*** chandrav has quit IRC21:37
*** banix has joined #openstack-neutron-ovn21:48
*** banix has quit IRC21:49
*** banix has joined #openstack-neutron-ovn21:50
*** banix has quit IRC22:00
*** banix has joined #openstack-neutron-ovn22:01
*** chandrav has joined #openstack-neutron-ovn22:04
*** mickeys has joined #openstack-neutron-ovn22:09
*** chandrav has quit IRC22:12
*** banix has quit IRC22:39
*** learner has joined #openstack-neutron-ovn23:52

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