13:02:51 #startmeeting hyper-v 13:02:52 Meeting started Wed Dec 16 13:02:51 2015 UTC and is due to finish in 60 minutes. The chair is claudiub. Information about MeetBot at http://wiki.debian.org/MeetBot. 13:02:53 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 13:02:55 The meeting name has been set to 'hyper_v' 13:03:00 hello folks! 13:03:07 hey 13:03:12 Hello. Good morning 13:03:20 0/ 13:03:22 o/ 13:03:24 Hi all 13:03:34 o/ 13:03:44 o/ 13:04:20 ok, so let's see... 13:04:31 lpetrut? 13:05:07 Hi 13:05:13 ok, so he joined. we can start. :) 13:05:19 #topic fibre channel 13:05:33 so, lpetrut, any news for us on this regard? :) 13:05:57 yep, the implementation is ready for the passthrough disks scenario 13:06:13 I was actually sending the implementation upstream right now 13:06:36 cool, can you share the link with us then? 13:06:50 so we can start reviewing it :) 13:06:53 I'll send it as soon as I send them 13:07:15 great. Anything else that needs to be done for fibre channel? 13:07:36 yep, we'll have to support the other scenario, when FC ports are exposed to the instance 13:07:57 but that should be easy to implement on top of this work 13:08:13 I see. Any thoughts about when that's going to be up for review? :) 13:08:43 hard to say, best case scenario: next week 13:09:19 Sagar, Sonu: any questions on this topic? 13:09:43 If you need review time from HP, let me know please 13:10:18 that would be great, I'll send you guys the links. Maybe you guys could even test this on your environment 13:10:43 agreed. Sounds like a plan. :) 13:10:56 ok, anything else on this topic? any questions? 13:11:27 i suppose not. moving on. 13:11:44 #topic OVS Microsoft support statement 13:12:17 primeministerp: I hope you can help us on this topic. :) 13:13:40 claudiub: do we have a plan to get WHQL sign 13:13:53 for the OVS port on Windows. 13:15:18 Sonu: it's an interesting question. Would be nice, but I'm not the right person to answer this question. Still waiting for primeministerp's response. :) 13:16:52 hm, he was here when the meeting started. hopefully he will respond on this matter. We will return to this topic later then. Let's not get stuck. :) 13:17:18 #topic Nova OVS vif status 13:17:54 so, I this refers to the ovs plug vif driver for the HyperVDriver. 13:17:55 well, the patches are updated to support live-migration 13:18:26 any blockers on this? 13:18:52 there is a -1 from somebody. I posted a response on the patch. waiting to hear back 13:19:16 if it doesn't move soon, I will attempt to contact the reviewer on irc or mail 13:19:24 well, I suppose you can ping him on irc. 13:20:10 are there any other patches that will have to be up, other than the already existing 2? 13:20:16 to add that live-migration support? 13:20:22 atuvenie: do we have any OVS neutron agent changes for hyper-v OVS 13:20:38 no, the live-migration support was ammended to an already existing patch 13:21:03 sonu: neutron ovs was ported to windows since kilo 13:21:20 great. 13:22:08 Using WMI firewall driver with Hyper-V OVS is a supported scenario. 13:22:24 nice. so, we'll have to review those patches. I'll make sure that they are on the nova mitaka priority list. 13:22:37 The intention is to get VXLAN with Hyper-V 13:23:13 Sonu: yep. 13:23:26 any other questions on this matter? 13:23:34 none from my side. 13:23:50 ok. moving on. 13:24:12 #topic performance enhancements 13:24:19 claudiub: we wanted some time to discuss on neutron blueprint 13:24:45 sure. 13:25:13 k, whenever you are done with your agenda 13:25:24 we can discuss on it 13:25:47 so, pymi is going to replace wmi pretty soon. We've run tests in every project using pymi and it was all green. 13:26:25 and I was wondering if you guys, Sonu and kvinod have tested it in your environments 13:26:45 we are doing the test 13:27:21 have you noticed any issues? how much did the overall performance improve in your envs? 13:28:17 we are yet to receive the result from our teams 13:28:56 however, the prime purpose of the concurrency patch is not only to improve the performance, but also to support concurrent operations 13:29:22 also, I should mention this, we have an approved blueprint on nova. It is the one that will implement "wait for neutron event" mechanism in the HyperVDriver. 13:29:48 basically, on instance spawn, before the VM is started, it will wait for the neutron ports to be bound first. 13:30:29 imagine if the queue builds up, 'cz neutron agent have to handle updates as well, all in the same process. 13:30:58 we have already implemented neutron notifier based VIF notif in nova 13:31:28 Sonu: please let us know what the results are. It would also be great to gauge the performance difference between current hyper-v neutron agent + pymi and concurent binding hyper-v neutron agent + pymi. 13:32:43 claudiub: as I had said, the purpose of concurrency patch is to support concurrent users doing operations. 13:32:45 also, I'm thinking that a greenpool might be an interesting alternative to explore as well. 13:33:41 our intention was to use additional CPUs on the servers to do new port binding, while the updates will be handled in original process 13:34:46 separation of new request handling, and update handling will be needed to support concurrency theme. 13:35:31 I see. so, I'm gonig to test the blueprint as well and see how it goes. At the very least the implementation will have to be rebased. :) 13:35:43 I am glad 13:35:44 claudiub : I would suggest you also to take the concurrency patch sets and try on your setup. 13:36:15 On the same note, I will get you the results of PyMI tests we have run 13:36:58 kvinod: yep, will do. although, it needs to be rebased, also, I think I have a concern on one of the patches. I don't remember if I added a comment on this, but it looked to me like ports were being processed twice. Will have to re-review. 13:37:10 Sonu: excellent, thank you. :) 13:38:11 anything else on this topic? 13:38:19 ok, you can try out with latest path sets, I uploaded one more version after I sent out mail to you 13:38:51 ok, great. :) 13:38:57 primeministerp: here, by any chance? 13:39:53 #topic open discussion 13:40:22 if you have any other topics you wish to discuss, now's the time. :) 13:41:36 atuvenie: I was asking about this neutron patch - https://review.openstack.org/#/c/67316/ 13:42:42 sonu: that patch is abandoned for some time from what I see. I don't know anything about it. I will investigate 13:43:20 in any case, I did not apply them on my setup. And basic use cases had so far worked for me 13:45:11 seems to be related to an OVS related firewall, if I'm understanding the commit message properly. 13:45:59 I don't think it affects HyperVSecurityGroupsDriver. 13:46:21 claudiub: on your statement, on applying on you setup 13:46:27 thanks. 13:46:39 the neutron patch that sonu_ mentioned 13:47:02 kvinod: what about it? 13:47:20 you said you will try on your setup 13:47:37 vinod: you are referring to the neutron blue print 13:47:44 anyways, alexpilotti: one topic remained undiscussed, the OVS Microsoft support statement one. Any ideas on this? 13:48:00 kvinod: I will. :) 13:48:02 yes 13:48:30 it will be good to have some results in the next irc 13:48:33 thanks 13:49:08 kvinod: sure. next meeting will be in 23th december. Will you be able to attend it? 13:49:49 Yes will try to log in 13:50:16 If not me Sonu will be there 13:50:49 great. :) 13:50:50 sure 13:51:08 so open item is Microsoft support statement for OVS. 13:51:42 claudiub, we're going to need to get it signed, and certified 13:52:20 primeministerp: great. Any idea about when that is going to happen? 13:52:21 claudiub, I'll start a thread internally regarding the OVS support statement 13:52:28 claudiub, we would have to do it 13:52:38 claudiub, probably as part of the ovs ci 13:52:44 and build process 13:53:21 we'll also have to have some internal discussion i'm sure 13:53:30 I'll start that process now 13:53:38 as it usually moves pretty slow 13:54:09 primeministerp: great, thank you! 13:54:36 np 13:54:37 anything else folks? 13:54:50 primeministerp, thanks. Would you say with OVS 2.5 we will have a singed driver 13:55:03 correction-signed 13:55:52 for the record, here's the os-win side of the FC implementation: https://review.openstack.org/#/c/258512/ Note that there's a chain of a few patches, this is the last of them. I'm sending the Nova one as well 13:56:11 lpetrut: great, thanks! 13:56:18 #link https://review.openstack.org/#/c/258512/ 13:56:53 thanks 13:57:07 seems that ovs 2.5 should be released pretty soon. 13:57:43 yes I am hoping 13:57:56 for OVS 2.5 will be out soon with conntrack 13:59:02 Thanks everyone. I have to logout. 13:59:08 anyways, the end meeting time draws near. we will have to end it soon. 13:59:29 so, thank you all for joining in! 13:59:33 can you send us the meeting notes 13:59:34 see you next week! 13:59:37 bye 13:59:42 they are automatically logged. 13:59:45 #endmeeting