15:59:39 #startmeeting hyper-v 15:59:40 Meeting started Tue Nov 13 15:59:39 2012 UTC. The chair is primeministerp. Information about MeetBot at http://wiki.debian.org/MeetBot. 15:59:41 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 15:59:42 The meeting name has been set to 'hyper_v' 15:59:48 Hi Everyone 16:00:20 going to wait a couple more minutes for some others to join 16:02:17 ociuhandu: are you going to stand in for apilotti? 16:03:22 primeministerp: yes, he's joining also but i guess he got delayed in traffic 16:03:26 ahh 16:03:27 ok 16:04:06 alexpilotti: alessandro! 16:04:19 great let's begin 16:04:25 hi everybody! 16:04:36 sorry for being late, just arrived! 16:04:37 so I sent out a brief agenda 16:04:42 alexpilotti: no worries 16:04:53 ok so updates 16:05:05 let's start w/ the new bits that have been added 16:05:13 #topic cloudrive 16:05:45 so has the clouddrive code made it through approvaL 16:05:53 did anybody have the occasion to check out the configdrive / cloudinit bits? 16:06:14 alexpilotti:I know you were working on that yesterday 16:06:52 primeministerp: still waiting for reviewer's love :-) 16:06:52 sorry configdrive 16:06:56 not clouddrive 16:07:01 ok 16:07:04 I had to split it further 16:07:08 who do we need to poke to get it to happen 16:07:26 o into smaller commits? 16:07:32 we need to ping the nova core guys 16:07:43 primeministerp: yep. 2 of them got approved 16:07:48 2 are waiting 16:07:58 vishy: ping 16:08:23 and I also added documentation for the testing framework: https://review.openstack.org/#/c/15930/ 16:09:53 ociuhandu and me tested it quite well on ubuntu, finding a couple of bugs in cloudinit 16:10:04 we submitted them and those are going to be fixed 16:10:21 in the meantime we are using a patched version of cloudinit 16:10:35 heh 16:10:37 sory about that 16:10:40 er sorry 16:10:56 cloudinit, being outside of Openstack, is not benefitting of Jenkins / OpenStack 16:10:57 ok 16:11:02 sorry, SmokeStack 16:11:12 alexpilotti: what did i miss 16:11:21 so regressions and generic errors are common 16:11:57 primeministerp_: nothing special, just a rough debugging weekend ;-) 16:11:59 ok 16:12:04 i guess moving on 16:12:16 about cloudinit, I took a look at the Windows versions available 16:12:20 is there anything else to discuss re: configdrive 16:12:25 IMO none of them are up to the task 16:12:26 or cloudinit? 16:12:37 alexpilotti: did you already cover cloudinit? 16:12:52 ahh thx 16:12:53 cloudinit / configdrive? 16:12:53 that's what i missed 16:13:16 so the idea is to write a new windows implementation or to port the Linux version 16:13:17 alexpilotti: so, suggestions? 16:13:31 alexpilotti: new windows implemenation? 16:13:41 I'd like an implementation that can be easily injected in a VHD w/ dependencies 16:13:45 so no .Net 16:13:48 agreed 16:14:04 so you want something regardless of guest platform 16:14:07 I'd either go w a port of the Python code or a C++ version 16:14:31 alexpilotti: ok 16:14:33 cloudinit is written in quite modular and decoupled way 16:14:53 so it could be possible to port it w/o wasting time 16:15:05 alexpilotti: ideally we don't waste any time 16:15:08 on the other side, it's probably faster to come up with a C++ version 16:15:18 the objectives are: 16:15:26 1) hostname config 16:15:45 2) user creation / password / group memmbership 16:15:53 3) static network injection 16:16:15 4) userdata scripts (powershell in this case) 16:16:16 which basically all means, just a harness to run powershell 16:16:19 ;) 16:16:34 not really, I'd avoid powershell as a dependency as well 16:16:58 and use wmi to calls to create users 16:17:14 and such 16:17:20 or simple "net user xxx yyy /add" invocations 16:17:41 at that point why not just use powershell 16:17:52 if you're invoking those commands 16:18:01 wmi is a bit of a PITA when it's about usr management 16:18:12 no, just a plain shell 16:18:32 think about Windows 2003 guests 16:18:42 ahh 16:18:44 gotcha 16:18:46 I want to be sure that every possible Windows workload is supported 16:18:49 i forget about those 16:18:53 ok 16:18:55 fair enough 16:19:10 w/o forcing the installation of any additional component 16:19:10 native os only 16:19:13 I agree 16:19:17 that's the idea 16:19:24 perfect then 16:19:32 Python is a good compromise, as we can do a simple xcopy deployment 16:19:45 so I'm looking at those two options. 16:19:50 any ideas? 16:19:50 ok 16:19:52 :-) 16:20:03 python sounds good to me 16:20:17 JAVA ! no, I'm joking... 16:20:21 pnavarro: any input? 16:20:23 haha 16:20:29 o man 16:20:34 too funny 16:20:35 pnavarro: lol 16:21:04 nothing to add.. 16:21:14 alexpilotti: pnavarro: so the two of you discussed quantum bits last week. 16:21:19 #topic quantum 16:21:30 any positive outcome? 16:21:38 pnavarro: I know you are extremely busy 16:21:58 on our we got blocked by the cloudinit bugs / review 16:21:58 alexpilotti: moving ahead as planned? 16:22:14 alexpilotti: understood 16:22:20 alexpilotti: more work than expected 16:22:22 yeah, we wasted a few days but things are moving as fast as usual ;-) 16:22:40 are we still on timeline for G1 or offically looking at G2 16:23:06 I've started a basic implementation from the basis of the linux bridge one 16:23:08 G2 at this point, there's no need IMO in spending night coding 16:23:17 pnavarro: cool! 16:23:22 alexpilotti: please don't 16:23:36 but, as I have the volume bits to commit... 16:23:38 alexpilotti: save that for the week before "G" 16:23:51 pnavarro: o please get those in asap 16:23:59 pnavarro: +1 16:24:16 pnavarro: I wasn't aware that they didn't get in yet 16:24:31 they are ready for commit, just live migration testing are failing because I have no the environment ready 16:24:43 pnavarro primeministerp: should we give the plugin a different name and kkep them both? 16:24:58 like appending a "V2" at the end? 16:25:08 alexpilotti: that works for me for backward compat 16:25:18 primeministerp: yep, that was the idea 16:25:21 pnavarro? 16:25:40 what to you mean by V2? 16:25:48 plugin version 2 16:25:50 pnavarro: bascially have version 1 and version 2 16:26:03 so if the class name is 16:27:12 I wanted to paste the real class name, but my VM is stuck :-D 16:27:21 hehe 16:27:22 volumeutils ? 16:28:03 pnavarro: in a nutshell having multiple versions of the driver 16:28:16 pnavarro: going forward ... i.e. win9 16:28:25 pnavarro: there is no wmiv1 namespace 16:28:37 pnavarro: whatever, I'll find it 16:28:57 pnavarro: the idea is to give the user a choice, or even better automatically choose the version based on the OS 16:29:08 I'd go with the second 16:29:21 you just need to add a simple factory 16:29:56 mmm ok, I thought the no-WMI version should be deprecated 16:30:08 pnavarro: in driver.py 16:30:14 when you do: self._volumeops = volumeops.VolumeOps() 16:30:33 you replace that with afactory method that goes like: 16:30:44 if os < 2012: 16:30:56 pnavarro: return volumeops.VolumeOps() 16:30:59 else: 16:31:05 return volumeops.VolumeOpsV2() 16:31:15 ok, I got it 16:31:29 pnavarro: what do u think? 16:32:11 it's ok for me 16:32:37 pnavarro: great tx! we are going to do the same also the vmops and the rest 16:32:54 ok 16:32:55 good, 16:33:00 for example VHDX require the WMI api V2 16:33:04 no pnavarro get that code in 16:33:06 er so 16:33:10 so we'll have to rewrite spawn and all the rest 16:33:10 ;) 16:33:45 ok 16:33:53 esp if we're going to work on anything going fwd 16:33:56 ;) 16:33:57 ok 16:34:03 anything on that? 16:34:14 I've realised that tests stubs are duplicated 16:34:33 pnavarro: which ones? 16:35:26 you know, the stubs changed depending on they are launched from /nova/tests or /nova 16:35:42 ?? 16:35:45 so they fully qualified name of the stub changes 16:36:05 I got it 16:36:17 d'oh 16:36:22 https://github.com/openstack/nova/tree/master/nova/tests/hyperv/stubs 16:37:18 ok 16:37:24 #topic ci update 16:37:39 so, we should add this convention in some dev doc 16:38:01 alexpilotti: perfect timing i guess 16:38:04 pnavarro: we should fix that! 16:38:18 alexpilotti: +1000 ! 16:38:19 alexpilotti: re: devdoc and the mocs 16:38:34 ok ci stuff 16:38:56 alexpilotti: and ociuhandu are coming to cambridge the week of dec 2 16:39:22 we'll be bringing up the smokestack/devstack tests and lighting up the ci infrastructure 16:39:34 pnavarro: also, they are asking to change the format from pickle to json 16:39:49 pnavarro: I'll work on giving you access to the infrasture at that time 16:40:10 that'd be great ! 16:40:10 mordred: did you want to come up? 16:40:10 as a general rule, let's execute them in nova\tests for the moment 16:40:21 alexpilotti: ok ! 16:41:20 alexpilotti: does that require much work? moving to json 16:41:44 primeministerp: just changing the serialization layer 16:41:44 alexpilotti: from the pickle files 16:42:03 primeministerp: 2 lines if the python code works well 16:42:07 alexpilotti: and it would align with the project better 16:42:16 alexpilotti: as they are already using a lot of json 16:42:27 alexpilotti: hence the request i'm sure 16:42:28 primeministerp: a lot of swearing if it happens like with Python 2.6 16:42:31 primeministerp: maybe - I've got to sort out a different work thing today before I can know 16:43:14 alexpilotti: thierry carrez mentioned the other day in Paris that the TComittee started discusions to forget Python 2.6 16:43:15 primeministerp: and alos if would be asy to read them and inject manually code 16:43:15 mordred: we'll if you're interested let me know 16:43:30 alexpilotti: yes 16:43:35 alexpilotti: no gz 16:43:51 I'd like to get rid of the gzs 16:44:11 alexpilotti: i'm sure every one else would too 16:44:14 ;) 16:44:26 it's the source of a lot of questions 16:44:29 if the json files don't bloath as much as the picled ones! :-) 16:45:03 alexpilotti: we need to have a discussion after this if possible 16:45:14 does anyone have anything else to add? 16:45:22 sure 16:45:32 let me say thanks to sagar_nikam! 16:45:40 o yes 16:45:40 he's doing a great testing work! 16:45:53 sagar_nikam: thanks for your help w/ testing 16:46:01 i have started the tests for live migration 16:46:04 sagar_nikam: you're doing a great job 16:46:11 sagar_nikam: feel free to file bugs 16:46:15 thanks all 16:46:19 sagar_nikam: also thanks for keeping me in the loop 16:46:24 and talking about testing, we need to fix glance_client. Volunteers? ;-) 16:46:48 alexpilotti: it's broken? 16:46:49 pnavarro: actually the discussion should start at today's meeting. 16:46:54 primeministerp: we are going to open a page to file bugs for the installer 16:47:09 alexpilotti: good 16:47:16 ttx: thanks for the note ! 16:47:19 primeministerp: badly. It doesn't work on Hyper-V 16:47:33 ttx: which one? 16:47:33 jaypipes: hi! 16:47:44 the py2.6 one 16:47:48 ahh 16:47:49 jaypipes: we have some issues with glance_client :-) 16:48:55 guys, I have to go ! 16:49:03 bye ! 16:49:49 bye pedro 16:50:00 alexpilotti: email him? 16:50:12 ok i"m going to call it 16:50:18 alexpilotti: let's chat now if that's ok 16:50:23 #endmeeting