Thursday, 2016-01-14

*** aginwala has joined #openstack-neutron-ovn00:03
*** armax has quit IRC00:05
*** salv-orl_ has quit IRC00:31
*** chandrav has quit IRC00:33
*** salv-orlando has joined #openstack-neutron-ovn00:34
*** armax has joined #openstack-neutron-ovn00:39
*** chenli has joined #openstack-neutron-ovn00:43
*** salv-orlando has quit IRC00:45
*** salv-orlando has joined #openstack-neutron-ovn00:46
*** aginwala has quit IRC01:12
*** gongysh_ has quit IRC01:20
*** aginwala has joined #openstack-neutron-ovn01:30
*** chandrav has joined #openstack-neutron-ovn01:35
*** aginwala has quit IRC01:44
*** aginwala has joined #openstack-neutron-ovn01:47
*** fzdarsky has quit IRC01:57
*** chandrav has quit IRC02:03
*** chandrav has joined #openstack-neutron-ovn02:07
*** thumpba has joined #openstack-neutron-ovn02:10
*** chandrav has quit IRC02:16
*** yamamoto has joined #openstack-neutron-ovn02:17
*** aginwala has quit IRC02:22
*** yamamoto has quit IRC02:24
*** yamamoto has joined #openstack-neutron-ovn02:26
*** aginwala has joined #openstack-neutron-ovn02:27
*** yamamoto has quit IRC02:28
*** armax has quit IRC02:29
*** yamamoto has joined #openstack-neutron-ovn02:30
*** thumpba has quit IRC02:32
*** yamamoto has quit IRC02:38
*** thumpba has joined #openstack-neutron-ovn02:42
*** gongysh_ has joined #openstack-neutron-ovn02:50
*** chandrav has joined #openstack-neutron-ovn02:58
*** chandrav has quit IRC03:05
*** chandrav has joined #openstack-neutron-ovn03:06
*** aginwala has quit IRC03:09
*** thumpba has quit IRC03:13
*** aginwala has joined #openstack-neutron-ovn03:14
*** thumpba has joined #openstack-neutron-ovn03:16
*** aginwala has quit IRC03:16
*** fzdarsky has joined #openstack-neutron-ovn03:16
*** thumpba has quit IRC03:17
*** thumpba has joined #openstack-neutron-ovn03:26
*** yamamoto has joined #openstack-neutron-ovn03:28
*** thumpba has quit IRC03:28
*** thumpba has joined #openstack-neutron-ovn03:28
*** chandrav has quit IRC03:45
*** chandrav has joined #openstack-neutron-ovn03:47
*** chandrav has quit IRC03:52
*** fzdarsky has quit IRC03:52
*** fzdarsky has joined #openstack-neutron-ovn04:01
*** chenli has quit IRC04:08
*** salv-orl_ has joined #openstack-neutron-ovn04:10
*** salv-orlando has quit IRC04:12
*** thumpba has quit IRC04:15
*** thumpba has joined #openstack-neutron-ovn04:28
*** chenli has joined #openstack-neutron-ovn04:29
*** fzdarsky has quit IRC04:36
*** chenli has quit IRC04:37
*** chenli has joined #openstack-neutron-ovn04:38
*** thumpba has quit IRC04:44
*** thumpba has joined #openstack-neutron-ovn04:54
*** armax has joined #openstack-neutron-ovn05:06
*** thumpba has quit IRC05:30
*** VikramDham has joined #openstack-neutron-ovn05:33
*** VikramDham has quit IRC05:41
*** thumpba has joined #openstack-neutron-ovn05:41
*** stac has quit IRC05:59
*** stac has joined #openstack-neutron-ovn06:04
*** thumpba has quit IRC06:14
*** aginwala has joined #openstack-neutron-ovn06:14
*** aginwala_ has joined #openstack-neutron-ovn06:23
*** thumpba has joined #openstack-neutron-ovn06:25
*** aginwala has quit IRC06:26
*** aginwala has joined #openstack-neutron-ovn06:27
*** aginwala_ has quit IRC06:30
*** gangil has quit IRC06:31
*** aginwala has quit IRC06:43
*** aginwala has joined #openstack-neutron-ovn06:43
*** armax has quit IRC06:48
*** armax has joined #openstack-neutron-ovn06:49
*** gongysh_ has quit IRC06:53
*** thumpba has quit IRC07:16
*** thumpba has joined #openstack-neutron-ovn07:26
*** aginwala has quit IRC07:34
*** gongysh_ has joined #openstack-neutron-ovn07:37
*** salv-orl_ has quit IRC07:53
*** salv-orlando has joined #openstack-neutron-ovn07:53
*** chandrav has joined #openstack-neutron-ovn08:04
*** thumpba has quit IRC08:30
*** salv-orlando has quit IRC08:57
*** salv-orlando has joined #openstack-neutron-ovn08:58
*** chandrav has quit IRC09:01
*** chandrav has joined #openstack-neutron-ovn09:04
*** salv-orlando has quit IRC09:05
*** salv-orlando has joined #openstack-neutron-ovn09:05
*** ig0r_ has joined #openstack-neutron-ovn09:17
*** chenli has quit IRC09:22
*** thumpba has joined #openstack-neutron-ovn09:38
*** ig0r_ has quit IRC09:41
*** thumpba has quit IRC09:48
*** chandrav has quit IRC09:52
*** salv-orl_ has joined #openstack-neutron-ovn10:10
*** salv-orlando has quit IRC10:12
*** gongysh_ has quit IRC10:41
*** rtheis has joined #openstack-neutron-ovn12:04
*** gongysh_ has joined #openstack-neutron-ovn12:18
*** thumpba has joined #openstack-neutron-ovn12:47
*** thumpba has quit IRC12:53
*** yamamoto has quit IRC12:54
*** dslevin has quit IRC13:39
*** yamamoto has joined #openstack-neutron-ovn13:39
*** dslev has joined #openstack-neutron-ovn13:48
*** dslev_ has joined #openstack-neutron-ovn13:54
*** dslev has quit IRC13:57
*** dslev_ has quit IRC14:00
*** fzdarsky has joined #openstack-neutron-ovn14:05
*** dslev has joined #openstack-neutron-ovn14:16
mesteryrussellb: Good morning! So, to get networking-ovn workign with stable/liberty required this patch:
mesteryrussellb: I am thinking about a stable/liberty branch of networkign-ovn, though I don't want it ...14:19
*** regXboi has joined #openstack-neutron-ovn14:20
russellbmestery: if that's all it is, what about a variant that can be applied to master?14:20
mesteryrussellb: I'd like that much better I think14:20
russellbeither automatically discover what's needed, or just a config option ...14:21
mesteryregXboi: This is about your patch here
russellbi mean, it's cruft14:21
russellbbut a separate branch to maintain sounds like more work14:21
*** thumpba has joined #openstack-neutron-ovn14:21
mesteryWell, also, the master patch is out for backport to stable/liberty too14:21
mesteryI mean, the neutron patch which caused the change14:21
russellbre: the branch, if that ends up best, i'm OK if you're the maintainer :-p14:21
mestery+1000, I want to avoid a separate branch14:21
* regXboi blinks owlishly14:21
* Sam-I-Am wakes up14:21
* russellb naps14:21
regXboiwhat if we do something hideous like wrap that statement in a try and if it fails, faill back to the old topics?14:22
mesteryThat is hideous, but would be along hte lines of what russellb was suggesting14:22
regXboiI *think* the key change is just from topics.PLUGIN to topics.REPORT14:23
regXboiand all the rest is cruft :)14:23
regXboisigh - brb folks14:23
* regXboi listens to the sounds of construction14:24
regXboiso... have we reached agreement on going hideous or not :) ?14:24
regXboiSam-I-Am: oom14:24
Sam-I-Amis this taking stable/liberty neutron code and putting ovn on top of it?14:25
regXboiSam-I-Am: that it is14:25
Sam-I-Ambuilding ovs from source?14:25
regXboiSam-I-Am: I did it that way14:25
regXboiothers might have done something different14:25
Sam-I-Amsort of have to, since none of the distros package master ovs14:25
Sam-I-Amare you packaging it, or just tarball?14:26
regXboithat, and the networking-ovn devstack plugin does it auto magic14:26
regXboineither: ^^^^14:26
Sam-I-Amso is this devstack or a real install?14:26
regXboiright now this is devstack14:26
Sam-I-Amoh, ok14:26
Sam-I-Amyou got me all excited for a moment14:26
regXboicalm down, the beer isn't flowing yet14:26
Sam-I-Amat this point it needs to be coffee :)14:28
russellbSam-I-Am: i can build you a package if you want one :-p14:29
russellbyou can build debs or rpms from the git tree :)14:30
Sam-I-Amrussellb: yeah... i know :)14:30
Sam-I-Amrussellb: tl;dr, packages play a significant role in the adoption of projects because they lower the barrier of entry14:30
russellbi do not disagree14:31
Sam-I-Amneutron is packaged, ovs can be packaged... which leaves networking-ovn14:31
russellbwoah now14:31
*** fzdarsky has quit IRC14:32
Sam-I-Amquestion is how long will it take distros to package ovn14:33
russellbit'll fall into place14:33
russellbwell, i mean, the packaging work is done, it's a matter of ovs releases getting out14:34
Sam-I-Ampackaging for ovn is done?14:34
russellbrdo will have latest ovs as soon as it's released14:34
russellbnot networking-ovn, but ovn itself14:34
Sam-I-Amoh, yeah... the part of ovs14:34
russellbi just put networking-ovn packaging on my todo14:34
russellbit's trivial14:34
Sam-I-Amyeah, it seems easy enough14:35
Sam-I-Ami can write up the scenario docs without package names14:35
russellbpackaging a small python thing like this is pretty cookie cutter14:35
russellbmaybe i can get ovs/ovn and networking-ovn into rdo's delorean (that builds master packages regularly)14:36
Sam-I-Amdelorean should be quick, but it does not build for stable releases14:36
Sam-I-Amso for mitaka packages would need to be in a more official place14:36
russellbstable not really relevant this minute14:36
russellbbut yeah, i get it14:37
Sam-I-Amthats sort of what i figured... lots of things not done yet. just thinking ahead.14:37
mesteryrussellb: I'm reading the routed network specs again now, and the new version seems to tie it very tightly to multi-segment networks ... which are in the ML2 plugin.14:37
* russellb nods14:37
mesteryTrying to understand what that means for us now14:37
Sam-I-Ami need to get the ubuntu people on board soon-ish14:37
Sam-I-Amthey dont move quickly14:37
russellbmestery: hrm, i read the latest, but maybe i glossed over some stuff14:37
russellbmestery: i think anything tied to ML2 should be -114:37
mesteryThe multi-segment support in ML2 is critical to this new version14:37
mesteryI'm saying that in my review now14:38
russellbcool thanks14:38
russellbmestery: speaking of rally, have you looked at our rally job results?14:38
mesteryI wrote that code 2 years ago :)14:38
* regXboi continues to watch the gate14:38
mesteryrussellb: I have not, what do they show?14:38
russellbi have no idea how meaningful the results are, but the numbers are lower than ml2+ovs, heh14:38
russellbregXboi the gatekeeper?14:39
russellbno, that's zuul.14:39
regXboirussellb: I was talking to mestery about doing a low level look at the networking-ovn code execution times vis a vis L314:39
regXboiextending that to include ml2 would probably be feasible14:40
regXboiif folks think its worthwhile14:40
russellbregXboi: i'd love any help starting to get performance or scale info14:41
russellbwhatever you think is interesting14:41
russellbwe just need to find bottlenecks so we know where best to invest dev time14:41
regXboirussellb: ack14:41
russellbalso, for anyone interested, reminder: OVN meeting is today in a few hours14:41
regXboicalendar ics link?14:42
*** russellb changes topic to " -=- OVN meeting Thursdays 10:15am Pacific / 1:15pm Eastern #openvswitch -=- Tempest health:"14:42
russellbsorry, no calendar link14:42
regXboino worries, I can add that to my calendar and carry it forward14:42
russellbi haven't created an openstack meeting yet, as it has been valuable to co-locate with the ovs devs14:43
regXboirussellb: no worries - I've got that on my calendar so I should be able to wander in14:44
* regXboi dances - the unit test jobs are fixed in the gate14:44
regXboinow to open the recheck sluices14:44
regXboimestery: just to confirm - your multinode ovn is running straight compute devstack everywhere *but* the controller, which is set up for networking-ovn?14:53
mesteryregXboi: Look here
mesteryThe Vagrant setup runs compute on all nodes *Except* the ovsdb-server node which is running ovsdb-server and ovn-northd14:54
mesterySo, compute on control node as well14:54
mesteryThat's easy enough to change though14:54
regXboiyeah: I was going to leave my controller as just controller+ovsdb14:55
mesteryregXboi: I suggest running a separate node with the DB and ovn-northd14:55
mesteryThat is what got me to a stable system14:55
mesteryEither that, orm ove your q-dhcp somewhere else14:55
regXboimestery - I'm not going to scale past 100 at this point - is it really that necessary?14:55
mesteryWell, then, yeah, you're fine14:55
Sam-I-Amrussellb: whats the time frame for rolling out 2.5?14:59
russellbunknown, "soonish"14:59
russellbrelies heavily on internal vmware QA ...14:59
russellbso not the clearest of visibility14:59
* mestery wonders who's booked their Austin hotel yet15:03
mesteryThinking JW Marriott, looks like conference things there and the Hilton downtown15:03
Sam-I-Ami have not even thought about it15:05
Sam-I-Amalthough it probably should be a thing15:05
Sam-I-Amrussellb: who wrote ovn-architecture.7 ?15:05
mesteryLooks like either the JW Marriott (which is super nice) or hte Hilton15:05
mesteryBoth within 2 blocks of the convetion center15:05
russellbSam-I-Am: mainly Ben Pfaff15:06
russellbothers have contributed some15:06
Sam-I-Amthats what i thought. he's damn good at writing.15:06
Sam-I-Amme likey15:06
russellbhe's done a lot of academic papers, too15:06
Mic22Guys my vote probably doesn't count but if you want to gain adoption or even more testing you should definatelly target liberty in the main branch. Also because the development semms to be really active and maintaining this activity in multiple branches risk to become a nightmare.15:10
Sam-I-AmMic22: backporting should be considered once ovn is feature-complete and stable15:12
mesteryAt this point, having a branch would be more of a pain than useful15:12
mesterySo if we can get even a hack in that enables networking-ovn in master and stable/liberty, that would be best.15:13
Sam-I-Amtoo many neutron things are released without significant vetting, or documentation.... which leads to FUD in the community :/15:13
Mic22"feature-complete and stable": it would be better ofc but is it really in a schedule that don't make this unuseful? I mean this is less useful if Mitaka come out stable before15:15
Mic22don't you think?15:15
Sam-I-Amcomplex things pushed for release just to make schedules never works15:17
Mic22Your decision ofc but speeking from my point of view (a company poit of view) OVN can already be a fit into a lot of deployments even without being feature complete. This can lead to a lot more testing that usually speed things up.15:22
Mic22I agree that push things just to release is stupid but whoever has a liberty deployment now will probably upgrade to Mitaka so maintaining liberty after mitaka is out seems to me a little bit "work for noone"15:24
Mic22but still is just my idea15:24
Sam-I-Amits not clear-cut15:27
Sam-I-Amthe key is having good documentation for what does and doesn't work15:27
Sam-I-Amthen making sure whatever a user needs to change down the road to enable the last set of features doesn't require re-deploying everything from scratch15:28
Sam-I-Amimo, the audience we want to adopt ovn has high expectations given prior experiences with neutron15:32
Mic22that's the point neutron can be a pain in the ass and OVN is making this a lot more simple and a lot more stable15:36
Mic22giving out this feeling will lead to have a lot more people testing, trying failing and giving new ideas15:37
Sam-I-Amyes, but right now some of the big advantages of ovn don't work ... like the built-in agents that replace L3/DHCP etc15:37
russellbour tempest reliability seems to have taken a hit ....15:37
Sam-I-Amusing a built-in L3 agent that includes redundancy and performance increases is a *big* deal15:38
Sam-I-Amrussellb: howso?15:38
russellbjust seeing failures more regularly15:38
Sam-I-Amanything in particular or just transient?15:39
Mic22true but that's already a lot compared to the plain neutron with L2 agents, ovs agents, etc... that crash and fuckup not every day but a lot of times. Or just a single fail that can lead to a downtime of a minute in the whole network.15:40
Mic22This stuff is for example the main reason why in my company we are still not using neutron15:41
Sam-I-Amovs is still going to be fun when it breaks15:42
Sam-I-Amhence why a lot of operators adopt neutron+linuxbridge15:42
Sam-I-Amseeing ovn being tested at some level of scale before release is a really great thing15:43
Sam-I-Ami hate to repeat the story of DVR...15:43
Mic22yes but how do you want to gain this tests when the only possible senario of testing is devstack?15:44
regXboirusselb: what's the patch to fix devstack blowing chunks on tox?15:46
regXboier russellb: ^^^^^15:46
* russellb was not aware of any chunks being blown15:46
Sam-I-Amme neither15:47
Sam-I-Amis your devstack drunk?15:47
*** armax has quit IRC15:47
regXboiSam-I-Am: i'm looking at stack failing because of recursion blow up when installing tox15:48
*** armax has joined #openstack-neutron-ovn15:52
*** armax has quit IRC15:52
*** VikramDham has joined #openstack-neutron-ovn15:59
*** fzdarsky has joined #openstack-neutron-ovn16:00
*** yamamoto has quit IRC16:01
*** gongysh_ has quit IRC16:02
russellband kind of related:
*** armax has joined #openstack-neutron-ovn16:06
*** chandrav has joined #openstack-neutron-ovn16:07
*** salv-orlando has joined #openstack-neutron-ovn16:09
*** salv-orl_ has quit IRC16:13
*** chandrav has quit IRC16:15
*** yamamoto has joined #openstack-neutron-ovn16:17
*** VikramDham has quit IRC16:21
mesteryrussellb: Checkout kevinbenton's comments replying to mine on routed networks16:27
mesteryTo me, routed networks adds segments to neutron's logical model16:27
mesteryThus, it's more than just ML2 like he's claiming16:27
mesteryMaybe I'm wrong, circling with carl_baldwin on it now :)16:27
russellbabout to grab lunch, but will read up this afternoon16:28
mesteryenjoy lunch!16:28
Mic22russelb: very good now mistakes are fixed, have a good lunch!16:29
*** yamamoto has quit IRC16:53
*** salv-orlando has quit IRC16:58
*** salv-orlando has joined #openstack-neutron-ovn16:58
*** dslev has quit IRC16:58
*** salv-orlando has quit IRC17:03
*** salv-orlando has joined #openstack-neutron-ovn17:04
*** dslev has joined #openstack-neutron-ovn17:06
regXboimestery: I'm going to try and restack on master rather than stable/liberty17:06
*** dslev has quit IRC17:09
regXboiSam-I-Am: moo - what's failing on my stack is tox --notest -efull17:18
regXboiI thought we had this fixed, but this stable/liberty stack appears to not know about it17:18
Sam-I-Ampatch not backported?17:21
*** dslevin has joined #openstack-neutron-ovn17:23
*** salv-orlando has quit IRC17:29
*** salv-orlando has joined #openstack-neutron-ovn17:29
*** azbiswas_ has joined #openstack-neutron-ovn17:56
russellbovn meeting in 5 minutes18:08
*** fzdarsky has quit IRC18:08
*** fzdarsky has joined #openstack-neutron-ovn18:08
Sam-I-Amrussellb: i really need some input from OVS people regarding the note in the networking guide about version requirements18:09
Sam-I-Amthe backport is getting a -1 now18:10
Sam-I-Amthere's 2 bugs, one for openstack-manuals, and one in devstack... regarding roughly the same thing.18:10
Sam-I-Amthe bugs in question are in the comments on this review18:11
regXboirussellb: here or in #openvswitch?18:14
Sam-I-Amprobably #openvswitch18:15
regXboiyeah, I'm not seeing traffic in either place18:15
*** chandrav has joined #openstack-neutron-ovn18:38
*** azbiswas has joined #openstack-neutron-ovn18:41
*** azbiswas_ has quit IRC18:55
*** fzdarsky has quit IRC19:25
*** shettyg has joined #openstack-neutron-ovn19:26
mesteryQuestion for folks: Is the tenant tunnel type configurable in OVN, or is GENEVE all we've got?19:27
Sam-I-Amgeneve or stt?19:27
mesterythat's it?19:27
Sam-I-Amvxlan can't carry the metadata iirc19:27
mesteryI know that19:27
Sam-I-Amso its only there for vtep19:27
mesterySo, STT does work?19:27
Sam-I-Amthe docs says its there... :)19:27
Sam-I-Ami havent tried it19:28
Sam-I-Amdocs say it might perform better in some cases19:28
mesterystt is much sexier19:28
azbiswasovs-vsctl --no-wait set open_vswitch . external-ids:ovn-encap-type="geneve"19:28
azbiswasThat's how the ovn-controller determines encap19:28
* mestery needs to try this with STT19:29
* Sam-I-Am makes notes for docs19:29
Sam-I-Amit'll be a bit different setting things via ovs-vsctl instead of config files19:29
Sam-I-Amlike provider nets19:30
*** gangil has joined #openstack-neutron-ovn19:31
*** gangil has quit IRC19:31
*** gangil has joined #openstack-neutron-ovn19:31
russellbmestery: Sam-I-Am yes STT is supported, in theory, but i'm not sure anyone ever tests it19:50
mesterythanks russellb19:51
russellbvxlan comes up in requests sometimes, mainly around "what about older hardware that doesn't have geneve offload yet"19:52
russellbvxlan-gpe and nsh support is in progress for OVS.  it seems like we could use that in theory to be able to support vxlan for OVN19:52
russellbbut that's a ways off, if at all19:53
*** gangil has quit IRC19:59
Guest89435mestery: unless you're going through a firewall  ;)20:02
*** Guest89435 is now known as arosen20:02
*** arosen has quit IRC20:03
*** arosen has joined #openstack-neutron-ovn20:03
* arosen device that does state tracking. 20:03
*** gangil has joined #openstack-neutron-ovn20:05
*** gangil has quit IRC20:05
*** gangil has joined #openstack-neutron-ovn20:05
*** armax has quit IRC20:09
*** armax has joined #openstack-neutron-ovn20:09
*** roeyc has joined #openstack-neutron-ovn20:25
flaviofmestery hey!20:32
mesteryflaviof: Hey my friend!20:32
flaviofmestery wrt to stt, you will need "Linux OVS tree", right?20:32
flaviofnot in linux upstream...20:32
mesteryI guess so yes20:32
flaviofmestery ack, just learning from my gurus as i 'listen'.20:33
* flaviof found a nice table of what is supported where in the ovs faq (
mesterySweet, thanks flaviof!20:41
*** chandrav has quit IRC20:46
*** chandrav has joined #openstack-neutron-ovn20:49
*** roeyc has quit IRC21:05
*** gangil has quit IRC21:12
*** gangil has joined #openstack-neutron-ovn21:27
*** gangil has quit IRC21:27
*** gangil has joined #openstack-neutron-ovn21:27
*** zhouhan has joined #openstack-neutron-ovn21:35
*** roeyc has joined #openstack-neutron-ovn21:37
*** thumpba has quit IRC21:44
*** thumpba has joined #openstack-neutron-ovn21:45
*** thumpba has quit IRC21:49
*** armax has quit IRC21:55
*** rtheis has quit IRC21:57
*** thumpba has joined #openstack-neutron-ovn22:01
*** thumpba has quit IRC22:02
*** salv-orl_ has joined #openstack-neutron-ovn22:09
*** salv-orlando has quit IRC22:12
*** azbiswas has quit IRC22:18
*** chandrav has quit IRC22:55
*** chandrav has joined #openstack-neutron-ovn22:57
*** zhouhan has quit IRC23:01
*** regXboi has quit IRC23:06
*** shettyg has quit IRC23:08
*** gongysh_ has joined #openstack-neutron-ovn23:08
*** chandrav has quit IRC23:46
*** chandrav has joined #openstack-neutron-ovn23:48
*** chandrav has quit IRC23:53
*** chandrav has joined #openstack-neutron-ovn23:54

Generated by 2.14.0 by Marius Gedminas - find it at!