09:29:57 #startmeeting XenAPI 09:29:58 Meeting started Wed Sep 16 09:29:57 2015 UTC and is due to finish in 60 minutes. The chair is BobBall. Information about MeetBot at http://wiki.debian.org/MeetBot. 09:29:59 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 09:30:01 The meeting name has been set to 'xenapi' 09:30:05 johnthetubaguy: ping! 09:30:20 Afternoon jianghuaw, huazhihao huanxie 09:30:45 @bob: good morning 09:30:50 hello Bob 09:31:00 :bobball Good morning. 09:31:02 Agenda at https://wiki.openstack.org/wiki/Meetings/XenAPI 09:31:33 So - I don't think there were any actions from the last meeting 09:31:41 #topic Blueprints and reviews 09:32:01 First thing to talk about is the vGPU support - jianghuaw Could you give an update on that? 09:32:20 yes. 09:32:36 hey, I am around, sorry, just lurking 09:32:54 the prototype work is in progress. 09:33:31 Can we have the gerrit link jianghuaw? 09:33:54 johnthetubaguy: If you have any feedback on the WIP code, feel free to contribute, but we do intend to submit a spec for M 09:34:06 we are able to populate the vgpu info to the compute-node info and it's able to sue the extra_spec to filter out host which support the specific vGPUs. 09:34:40 yes. the link is: https://review.openstack.org/#/c/223426/ 09:35:08 #link https://review.openstack.org/#/c/223426/ 09:35:12 BobBall: can't look at that right now, focusing on reviewing all the bug fixes 09:35:33 thats sucks, I know :( 09:36:00 no worries - look forward to discussing the spec with you 09:36:25 So the next step is to prototype the code that connects the vGPU to the instance when spawning, is that right jianghuaw? 09:36:45 exactly 09:36:59 Great, thanks. 09:37:19 Next topic - huanxie - could you give an update on the Neutron integration / fixes you've been working on? 09:37:28 Sure 09:38:09 With devstack we can install xenserver + neutron successfully, with ML2 as plugin, OVS driver, VLAN type 09:38:34 That needs an new devstack fix, right? Do you have the link for that? 09:38:54 yes 09:39:09 https://review.openstack.org/#/c/217651/ 09:39:18 #link https://review.openstack.org/#/c/217651/ 09:39:18 This is devstack script changes 09:39:39 (If you prefix things with #link then the bot which logs the meetings will include it as a link in the minutes) 09:40:10 Great news - so the next steps we're going to look into for this support are the tempest tests, on the route to setting up a Neutron CI 09:40:16 And we also found a problem of netwrap which resides in dom0, we use this to talk to dom0 for OVSDB monitor 09:40:40 Link to the bug report? 09:41:08 Yes, currently I can use #link https://github.com/citrix-openstack/qa/blob/master/install-devstack-xen.sh to install Xenserver + Neutron successfully 09:41:17 Sure, hang on 09:41:44 huanxie has also written a document detailing her debugging experience on this problem. 09:42:02 I am curious hows that going 09:42:09 huanxie: Could you add this to https://github.com/citrix-openstack/blogentries so we can review it there and eventually publish to blogs.citrix.com ? 09:42:10 are we basically restoring the old method to make this work? 09:42:29 BobBall: Sure 09:42:55 Yes johnthetubaguy: Daemon mode can't work without major changes to the daemon since it doesn't have the concept of ssh'ing to a remote host to run the command 09:43:51 The second issue is that that the optimisation to minimize polling by using ovsdb monitoring again cannot work in the current architecture because the command is being run over XenAPI which doesn't support streaming stdout; it will collect stdout and return it from the plugin once the process is complete 09:44:34 true, you would have to have your own deamon that ssh-ed into dom0, which sounds scary 09:44:49 There are some neutron configuration options that huanxie has found that will mean we can get support working again so we can set up a CI and then work on improving the integration to support daemon mode or the ovsdb monitor commands 09:44:49 Yes, thanks for Bob's explaination 09:45:22 Daemon mode rootwrap is in oslo as well - so it's not even neutron specific, so getting some custom code in there to support the SSH concept will be fun. 09:45:28 But something we can look at soon I hope :) 09:46:27 Hi, #link https://bugs.launchpad.net/neutron/+bug/1495423 09:46:28 Launchpad bug 1495423 in neutron "Cannot get dom0's ovsdb monitor result" [Undecided,New] 09:46:46 OK, great. Thanks for the update huanxie 09:47:10 BobBall: yup, Nova has just added deamon mode support 09:47:30 sounds like a good plan forward, from where I am standing 09:47:30 I don't think we need to rootwrap anything from Nova though do we ? 09:47:38 I mean for XenAPI into dom0 09:47:54 nope, not that I know of 09:48:07 For Neutron, there is also daemon mode, and the q-domu agent can use daemon mode 09:48:14 OK, good. Then it's just Neutron we need to worry about too much for now 09:48:28 But q-agt which talks to dom0 cannot use it in our case now 09:49:02 Anyway - So huanxie is also going to be looking at verifying Neutron support in a non-devstack environment so we don't just rely on devstack's documentation 09:49:25 Yes, Bob 09:49:38 #topic Bugs & QA 09:49:54 huanxie has already mentioned a couple of the recently discovered bugs in Neutron support 09:50:06 but jianghuaw has also discovered some bugs relating to the next release of XenServer 09:50:39 We've shifted to a CentOS-7 based release for the alphas that are available from xenserver.org 09:50:39 so big ask from me, any release blockers that you can see? 09:51:11 No blockers johnthetubaguy. 09:51:23 It would be good to get https://review.openstack.org/#/c/213112/ in if we can 09:51:59 It's the closest thing to a blocker for us in my mind 09:52:09 but probably wouldn't be considered a blocker by most in nova core :D 09:52:29 yeah, I should take another peak at that 09:52:58 The impact is that more people will be using HVM guests to run their compute going forward (for various reasons) and if they are using older versions of XS they will definitely be hitting the bug that huanxie fixed here 09:53:01 I think it was just some issues with the tests 09:53:32 huanxie: Are all comments addressed on that issue? 09:55:05 Bob: yes, I have modified with these comments 09:55:16 huanxie: the only reason I knew to look for that sleep bug is I made the same mistake myself testing similar code loops, a few times! 09:55:21 OK, great, so I will have another look today and then johnthetubaguy if you can look at it then it'd be great. 09:55:33 #topic Open Discussion 09:55:47 Time for an update huazhihao! :) 09:56:07 Could you let us know the status of the Mirantis Fuel plugin development? 09:57:13 We are gonna submit the plugin to Mirantis for review this weekend. 09:57:53 We are doing some health checks. 09:58:30 So we have a fully developed plugin now, and a separate XenCenter plugin to add the host internal management network to VMs in a nice way 09:58:54 As well as the health checks, we're just updating some more of the docs, right? 09:59:10 Yes 09:59:30 Could we have a link to the automated runbook in case anyone wanted to try the early docs that you wrote? 09:59:50 Of course, there have been some improvements and modifications to the plugins since those docs were written, so use them at your own risk :) 10:00:04 https://docs.google.com/document/d/1dwy0q04KndLUi-HXEr2enfl1EovBvzgc7NmOHCWWuO8/ 10:00:15 Thank you for reminding me of that ;) 10:00:35 Great - thanks for the update! 10:00:42 cool stuff 10:00:45 But we find a critical problem. guest instance cannot obtain ip from br100. 10:00:58 you got a link to the repo with the plugin in? 10:01:05 huazhihao: Probably just a configuration issue; we will have a look at fixing that 10:01:14 #link http://ca.downloads.xensource.com/OpenStack/Plugins/ <-- binaries 10:01:19 github.com/citrix-openstack/xenserver-fuel-plugin 10:01:33 huazhihao's link is the source for the Fuel plugin yes 10:01:52 #link https://github.com/citrix-openstack/xencenter-himn-plugin <-- source code for management network XenCenter plugin 10:02:03 github.com/citrix-openstack/xencenter-himn-plugin that is for the xc plugin that adds host internal management network 10:02:13 Sorry huazhihao - I just beat you to it! :) 10:02:38 Haha 10:02:40 OK - we're running over time, so we'll have to close it there unless there are other urgent items to raise? 10:03:11 That is all from me. 10:03:11 As a final note, we've all got approval to travel to Tokyo, so we can introduce you to the whole team there johnthetubaguy :) 10:03:41 OK great, thanks all :) 10:03:45 #endmeeting