13:00:37 #startmeeting hyper-v 13:00:38 Meeting started Wed Jun 1 13:00:37 2016 UTC and is due to finish in 60 minutes. The chair is claudiub. Information about MeetBot at http://wiki.debian.org/MeetBot. 13:00:39 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 13:00:40 woops :D 13:00:43 The meeting name has been set to 'hyper_v' 13:00:44 darn my calendar got daylight-savings-jacked again 13:00:45 no problem. ?) 13:00:51 sorry hyper-v folks 13:00:55 Hi All 13:00:59 no worries. :) 13:01:02 hello folks 13:01:14 o/ hello all 13:01:28 Hi 13:01:31 hi 13:01:42 ok. so let's get to it 13:01:50 #topic monasca patches 13:01:59 #link https://blueprints.launchpad.net/monasca/+spec/add-windows-support 13:02:11 blueprint is approved, there are 12 patches up already. 13:02:21 Roland reviewed a couple of them. 13:02:38 got some review on the process checker 13:03:01 although, the patch on process checker is more of a bugfix, tbh 13:03:07 nice 13:03:15 basically, there are no lower or upper constraints on psutil 13:03:32 and psutil 2.0.0+ and older than 2.0.0 behave differently 13:04:13 in OpenStack requirements, psutil is limited to >=1.2.0<2.0.0, if I'm not mistaken 13:04:48 so that means that if you have other OpenStack services on your nodes, chances are you have psutil < 2.0.0 13:04:58 so, had to address this issue 13:05:53 anyways. the hypervisor checker is still work in progress. had to do some refactor on it, and I have to include hyper-v specific metrics, like disk iops and latency 13:06:03 hi all, sorry I'm late 13:06:20 iis checker and wmi checker require minor updates, I haven't sent the patches for them up yet. 13:06:22 with the 12 patches, are we done or more patches still required ? 13:06:24 any questions? 13:07:03 sagar_nikam: well, one for iis checker, one for wmi checker, for hypervisor checker... don't know yet, 2-3 I think. 13:07:03 claudiub: any more patches required apart from the 12 ? 13:07:10 ok 13:07:25 but the big stuff is pretty much done. 13:07:28 any questions? :) 13:07:32 good 13:07:38 cool, moving on. :) 13:07:42 and it all works well with monasca ? 13:07:47 yep 13:07:56 #topic designate support 13:08:05 abalutoiu: started working on this. 13:08:06 nice .. very good progress in short time 13:08:35 the APIs are not very well documented, so, it might take a little bit of time to get everything right. 13:09:06 from what i know, designate already has some windows support 13:09:10 All the Windows DNS specific code will be in os-win, of course, and it will be used in designate 13:09:57 sagar_nikam: We've met with the designate TPL, he said there isn't. :) 13:10:10 oh ... ok 13:10:16 fine 13:10:18 sagar_nikam also I haven't found anything related to windows in designate 13:10:34 or if it was, it wasn't tested and time flew by, it stopped working properly. 13:11:28 anyways, this is still WIP. let's see how it goes 13:11:44 #topic freezer support 13:12:16 lpetrut: is currently working on this, he's working on a POC, to see how everything works and how it goes 13:12:43 ok 13:13:43 yep, so, I'll have the code ready really soon, at the moment I'm preparing the os-win side, including support for leveraging the Hyper-V VSS writer. As soon as I have this, I'll send blueprints on Freezer and start submitting code there as well 13:13:56 yeay. :) 13:14:05 wanted to see how the code gets shaped first so that the blueprint would include meaningful information 13:14:24 cul. :) 13:14:32 once freezer and monasca are in good shape, we plan to start testing it 13:14:36 both of them 13:14:38 that's great 13:14:43 sounds good. :) 13:14:45 let us know when you think we can start 13:14:56 however the only issue is we are on Mitaka 13:15:04 hopefully we can backport and test 13:15:20 as far as I know, Freezer is undergoing heavy refactoring 13:15:29 i guess that should not be difficult 13:15:33 so, the code submitted won't be compatible with mitaka 13:15:40 oh ... 13:16:06 moving on. 13:16:07 anyhow we can test with devstack and cloudbase MSI 13:16:56 true. :) 13:17:04 #topic Windows Containers 13:17:21 atuvenie: is barely containing her excitement on this subject. :) 13:17:23 haa 13:17:38 ha ha :) 13:17:53 yeah, so I'm looking into kubernetees on Windows right now 13:18:01 am i also keen to know how containers work in windows 13:18:34 There is an active effort within the kubernetees community to make it work with Windows 13:19:17 there are plenty of guides around the internet on windows containers. At the moment, we've submitted some patches to nova-docker for Windows Containers 13:19:23 I think this was in mitaka 13:19:39 the plan is to use/support nova-docker ? 13:19:56 sagar_nikam: there are 2 directions here 13:20:02 so there is nova-docker 13:20:04 i thought we were going with magnum 13:20:08 but, there is also magnum 13:20:14 ok 13:20:17 fine 13:20:27 but magnum uses kubernetes, docker swarm and mesos 13:20:42 yep. 13:20:44 we are looking into kubernetees on windows 13:20:57 patches in nova-docker already merged ? 13:21:04 and into docker swarm, but the main focus right now is kubernetees 13:21:40 #link https://review.openstack.org/#/c/219657/ 13:21:58 that is the patch on nova-docker regarding Windows Containers. 13:22:43 so the plan is that after we get kubernetees to work on windows, we can go ahead with magnum 13:22:44 claudiub: can we have a blog post on how to use nova-docker for windows ? 13:22:44 so, our work on nova-docker is done, we are focusing on magnum right now 13:23:13 the problem with magnum, is that the heat templates are heavily linux oriented 13:23:14 people can start using/testing it 13:23:55 sagar_nikam: I'm not sure we have made a blog post about this 13:24:03 I can't seem to find it at least. 13:24:53 claudiub: i requested for a blog post 13:24:54 yeah, I think we should do it, if time allows it 13:25:00 it is not there yet 13:25:21 ok agree 13:25:46 sagar_nikam: I don't think we've made one yet. Will take the suggestion into account. :) 13:26:06 #topic buuugs 13:26:08 thanks 13:26:30 is sonu joining us? 13:26:49 he is on vacation this week 13:26:57 ah, I see. 13:27:36 #link Fixes for multiple physnets in HyperV compute https://review.openstack.org/#/c/322068/ 13:27:52 so, I've reviewed this, I left some comments on it. 13:28:21 primarely, there is some usecase that wasn't considered. 13:28:31 sure i will ask the team to check 13:28:53 #link Fix for Neutron agent hang on nova boot https://review.openstack.org/321452 13:29:25 I haven't commented on this yet. I will want to verify and reproduce the bug 13:29:35 ok 13:29:41 the bug report is very nicely explained 13:29:46 and the steps are quite detailed 13:29:47 some of these bugs are from our scale test setup 13:30:18 but from what I can see in the fix, it will monkey patch the threading module 13:30:24 and native threads as well 13:30:36 which will have a negative performance impact. :( 13:30:56 so, I'll try to see exactly who, where, why this hang happens. 13:31:01 ok, let me check with sonu on it 13:31:27 hopefully, we won't have to monkey patch the threads. 13:31:51 #link Provide backwards compatibility for sec_group driver https://review.openstack.org/322117 13:31:56 this is an easy one 13:32:57 basically, the story is that in Liberty, the Neutron controller would have 'icmpv6' security group rules, while in mitaka and newer, they are called 'ipv6-icmp' 13:33:44 so, having a Liberty Neutron controller with Mitaka / Newton neutron-hyperv-agent would have been bad 13:34:08 with this, it will work. 13:34:27 anyways. 13:34:29 moving on. 13:34:34 #topic FreeRDP 13:34:41 c64cosmin: hellou. :) 13:34:44 o/ 13:35:02 any news? :) 13:35:08 sort of 13:35:15 well I'm been stuck in a delay 13:35:30 so I'm sorry to tell sagar and domi that RDP will be delayed a bit 13:35:42 oh .... 13:35:46 I hope I will finish by the end of this week 13:36:03 have we hit some complex issue ? 13:36:15 end of this week should be fine 13:36:35 i have requested my test team to pick up the new MSI whenever it is ready 13:36:42 so we upgraded to the new boost/openssl/cpprestsdk 13:36:49 until now that seems find 13:37:01 I cannot merge the the PR until those are ok 13:37:14 domi007: you around? 13:37:26 also we will build for x86&x64 in MSI and zip format too 13:37:32 zip is important for Nano 13:37:42 deb and rpm will follow after 13:37:53 the good news is that I've managed to isolate the crash bug 13:38:10 it's hard to replicate though 13:38:19 deb will be available ... that is great news.... 13:38:45 as I said, you and will find out that the MSI is out 13:38:50 and I will take care of that personally 13:38:52 c64cosmin: have you seen the issue where keystrokes dont work ? 13:39:01 yep 13:39:05 and only mouse click works 13:39:09 ok 13:39:16 that one seems to be from FreeRDP library itself 13:39:17 so is there a fix for that as well 13:39:30 if you are using get-vmconsole from powershell, you'd get the same problem 13:39:49 I will get in touch with the FreeRDP people and see if there is a resolution 13:40:02 ok ... that is a critical defect 13:40:19 indeed, it seems that HyperV doesn't do a 'keyboard refresh' for this one 13:40:21 without keystrokes working.. RDP console is not so useful 13:40:40 so this issue can be triggered from the console 13:40:47 but can be solved 13:40:49 c64cosmin: do you have a defect id for that already raised ? 13:41:17 id? 13:41:31 i mean bug number 13:41:37 which i can follow 13:42:03 to see how the progress 13:42:16 not yet, I'm not sure on which branch we are building the CI with 13:42:45 updating the CI might solve this issues 13:43:23 ok 13:43:25 this issue with the keyboard also appears on the console 13:43:34 if you leave the console while a key is pressed 13:43:54 you mean on VM console on HyperV host ? 13:44:01 the html5 13:44:16 ok 13:44:28 don't remember if that is upstream or not 13:44:53 anyhow, I never got a problem with the keyboard on both of the console 13:45:28 i hit the issue often 13:46:11 with horixon console 13:46:16 horizon 13:47:28 anything else? 13:47:34 moving on, then. 13:47:45 #topic os-brick patches 13:47:48 sure i am done 13:47:54 #link https://review.openstack.org/#/q/topic:bp/os-brick-windows-support 13:48:03 one of the patches merged, which is great. :D 13:48:19 and the other one was reviewed once by hemna. 13:48:28 and it was split into 3 patches, in order to make it easier to review 13:48:36 ok 13:48:39 it was extremely big before. :) 13:48:42 there was a nova patch as well 13:48:53 i had requested hemna to review it 13:48:57 yeah. that's waiting for these patches to get in first. 13:49:05 ok 13:49:19 hopefully we will have FC support in Newton 13:49:21 after these patches merge, we should request an os-brick release. 13:49:27 we have been trying from liberty 13:49:39 sagar_nikam: yep, I hope so too. :) 13:49:58 #topic Open Discussion 13:50:08 ok, so, a couple of things.. 13:50:14 you might want to read this: 13:50:20 that's embarasing, back 13:50:26 #link http://www.slideshare.net/alessandropilotti/puppet-windows-nano-server?qid=79617a6c-e608-417e-b8ba-a6e3dc6c48cf&v=&b=&from_search=1 13:50:44 it's about Windows Nano and how to use it 13:51:05 sure 13:51:11 will read it 13:51:18 fyi, Nano is an installation option available on the Windows Server 2016 images, including TP5. 13:51:35 last time i mentioned about nano not support MSI 13:51:43 yep 13:51:50 that is what I was going to say next. :) 13:51:51 what's the plan for it 13:52:17 we do some magic. :) 13:52:31 #link Hyper-V Compute Installer https://cloudbase.it/downloads/HyperVNovaCompute_Mitaka_13_0_0.zip 13:52:41 yeah, so the solution is to use the zip installer. 13:52:46 that's it. 13:53:18 ok 13:53:34 i had a question on cluster driver 13:53:35 fair warning, it uses python 3.4. We have fixed the majority of python 3.4 issues in this installer for Mitaka (patches still pending on Nova on Newton). 13:54:04 but there might be some issues that are not apparent. 13:54:18 sagar_nikam: sure, go ahead. 13:54:49 now that many nova patches are merged, can we have cluster driver in higher priority for core reviewers 13:55:02 we are very keen on cluster driver getting merged in newton 13:55:11 also, we are going to attend the OpenStack CEE Day in Budapest next week. 13:55:14 #link http://openstackceeday.com/ 13:55:16 lets try to get it merged in newton 13:55:47 sagar_nikam: sure, it needs to be reviewed. The sooner, the better. :) 13:56:07 sagar_nikam: some help on the reviewing part would be great. 13:56:13 for us that is the most important patch for mitaka 13:56:19 i mean newton 13:56:27 yes sure we will review 13:57:07 anything else? 13:57:21 nothing much 13:57:30 all discussion done... early today 13:57:37 :) 13:57:37 indeed. :) 13:57:42 fastest one. :) 13:57:51 yes by 3 mins 13:57:53 there was an email on the ML about Hyper-V vs KVM 13:57:57 trying to find it 13:58:11 intrested about it 13:58:19 what is being discussed 13:58:36 #link http://www.gossamer-threads.com/lists/openstack/dev/54736 13:58:42 i think this should be it. 13:59:21 bassically, the guy tried out some DB performance tests on hyper-v / kvm 13:59:40 hyper-v was more stable, faster, and lower host cpu usage. :) 13:59:49 :) 14:00:19 and I don't think he used any performance enhancing tricks that I've included in my reply. 14:00:24 anyways 14:00:34 thanks folks for joining today. :0 14:00:35 :) 14:00:38 #endmeeting