16:06:10 #startmeeting hyper-v 16:06:11 Meeting started Tue Jul 16 16:06:10 2013 UTC. The chair is primeministerp. Information about MeetBot at http://wiki.debian.org/MeetBot. 16:06:12 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 16:06:14 The meeting name has been set to 'hyper_v' 16:06:19 hi everyone 16:06:25 hi there! 16:06:50 hi all 16:07:07 so I think this might be quick 16:07:24 figure we give some quick updates 16:07:37 alexpilotti: I know you are in the process of pushing code 16:07:47 yep 16:08:00 alexpilotti: is there anything specific which we should keep our eyes on 16:08:02 WMIV2 and Dynamic memory, as discussed last time 16:08:14 alexpilotti: so still no changes 16:08:34 I'm going to send the email with the gerrit review urls as soon as those are up 16:08:47 alexpilotti: great 16:08:58 I don't see luis around 16:09:13 so 16:09:37 that means puppet discussion really isn't worth having w/o out him 16:10:12 alexpilotti: any more crowbar related bits that are interesting to note? 16:10:48 primeministerp: we have started sending the first changes 16:10:52 yes, we started pushing for review the Hyper-V support in Crowbar pebbles 16:10:57 ociuhandu: execellent 16:11:15 ociuhandu: would you like to give some details? 16:11:29 no matt ray on the channel right now either 16:11:42 primeministerp: we had to add support for samba, filter some of the linux-specific bits, so that they will not apply to windows 16:12:00 ociuhandu: the usual suspects 16:12:04 ociuhandu: keep up the good work 16:12:21 we try to keep the nodes as a concept transparent, so that we don't treat differently the windows and the linux nodes 16:12:35 but rather apply only what's supported / required 16:13:18 ociuhandu: does that require any refactoring of the base code? 16:13:32 the issues we had were with the ruby version update, as the windows clients were using ruby 1.9 while linux ones are on the 1.8 16:13:58 hanrahat: basically an issue w/ the upstream chef bits for windows being newer than the linux side 16:14:00 and there are syntax changes between them, good thing that 1.8 supports also the 1.9 syntax 16:14:57 hanrahat: if we're talking about the base crowbar code, yes, we need to add in all windows specific parts and make sure that the existing linux-specific ones do not apply on windows nodes 16:16:11 one other reason for not removing part of the roles from the windows clients completely is that in the future those services should be available for windows nodes too 16:16:11 ociuhandu: thanks 16:16:39 ociuhandu: yes, base crowbar code... thanks 16:17:03 sorry, I was late 16:17:10 hi rob 16:17:15 Hey! 16:17:17 zehicle: hi Rob 16:17:40 so moving on 16:17:51 alexpilotti: there were the new bits you had to share 16:18:22 alexpilotti: if you want to mention it, re the runner 16:19:18 alexpilotti: the job runner? 16:19:46 he must be sleeping 16:20:14 ociuhandu: do you want to comment on the glance image cleanup processes as well for garbage collection? 16:20:33 ociuhandu: I know alex wanted to mention it 16:21:41 so 16:21:47 if he's not going to mention it 16:22:03 primeministerp: I believe ociuhandu got disconencted 16:22:12 alexpilotti: you too? 16:22:42 primeministerp: nope I was around 16:22:59 primeministerp: unless you pinged me before and I was disconencted as well :-) 16:23:05 zehicle: hi! 16:23:18 zehicle: ociuhandu was just talking about Crowbar 16:23:41 primeministerp: should we switch to the next topic? 16:23:56 as I don't see any reply from zehicle or ociuhandu 16:24:00 I'm here 16:24:24 alexpilotti: I was waiting for you 16:24:35 alexpilotti: to discuss the glance cleanup 16:24:40 zehicle_: I was pinging you on the wrong nick :-) 16:24:43 missed the earlier thread on CB -> there are pulls to bring HyperV into CB "pebbles" Grizzly 16:25:01 zehicle_: yep, the first batch 16:25:03 me too, looks like colloquy came back to life 16:25:11 I had a dead IRC client that was holding the nick :( 16:25:39 zehicle_: looks like we all have issues today here with IRC except primeministerp :-) 16:25:40 We're coordinating w/ SUSE to review and accept. 16:25:56 hehe 16:26:18 zehicle_: cool, let us know if you'd like to meet on IRC / Skype / etc to discuss them 16:26:43 zehicle_: there's quite an amount of work out there in those patches :-) 16:26:45 zehicle_: one quick thing: the reason for not removing part of the roles from the windows clients completely is that in the future those services should be available for windows nodes too 16:27:07 zehicle_: e.g. nagios 16:27:08 zehicle_: like ipmi, ganglia, nagios 16:27:19 We've got a regular design/plan cadence setup for Crowbar. Plan is this thursday - would be helpful to include you there 16:27:47 +1 16:27:52 we chose to "skip" the recipe until the windows bits get implemented, rather than removing the role and adding it back later on 16:28:19 for the broader, OpenStack community - we're trying setup Crowbar as a quick way to do a Grizzly + HyperV deploy 16:28:25 ociuhandu: did you tell zehicle_ about the IPMI chef issue on the crowbar UI? 16:28:25 that would be great 16:28:58 cool, at what hour is the meeting on Thu? 16:29:06 8 am central 16:29:18 bit.ly/crowbar-calendar 16:29:31 alexpilotti: no, but i suggest we talk on that on the crowbar meeting. As a very short note, we do not have DMI info on the windows nodes so web interface is failing 16:30:37 ok 16:31:26 alexpilotti: do you mention the glance cleanup bits? 16:31:42 er ^want to 16:31:47 oki, added 16:31:47 zehicle_: great, we'll be there 16:33:49 sure 16:33:49 primeministerp: should we change topic? :-) 16:33:55 alexpilotti: i'm waiting on you 16:34:04 #glance cleanup scripts 16:34:06 er 16:34:12 :-) 16:34:13 #topic glance cleanup 16:34:14 txc 16:34:30 alexpilotti: all you 16:34:34 so we ran into an issue with the image cache 16:34:39 on nova compute nodes 16:35:07 we built a vaildation system that accepts arbitrary images sent in with an HTTP REST API call 16:35:44 that image gets downloaded, included in glance, a new image is spawned, a floating ip attached and the user gets notified of the availability 16:36:13 we are using it on a variety of images that need to be validated on Hyper-V 16:36:49 The issue is that the glance image cache used by nova-compute 16:36:49 is not getting "garbage collected" 16:36:59 this is not an issue in a regular environment, where images are relatively static 16:38:38 while it's an issue in a case like this one where potentially hundreds of images are getting added on a node every day 16:38:43 teh result as you can image, is that the host runs out of space sooner or later 16:38:44 as a workaround, I wrote a powreshell script that checks whicj images are not in glance anymore and deletes the corresponding VHD/VHDX files 16:39:04 this is scheduled as a Windows task every 15' 16:39:33 The best solution, would be to add this garbage collection in the nova-compute driver 16:39:48 For Grizzly we can just use the script 16:40:26 while for Havana it'd be nice to include it in the official codebase :-) 16:40:31 comments? 16:40:35 alexpilotti: o 16:40:40 alexpilotti: i'm for it 16:40:43 this is an issues for local cache images - not boot from block? 16:41:10 zehicle_: correct 16:41:15 zehicle_: yes 16:42:07 zehicle_: talking about boot from block, what's the status of the EQL driver? :-) 16:42:22 it's in the pebbles code base 16:42:52 zehicle_: it's not a standalone project? 16:43:03 no, it's part of the cinder barclamp 16:43:25 zehicle_: don't you plan to add it in Cinder? 16:43:33 https://github.com/crowbar/barclamp-cinder/tree/release/mesa-1.6/master/chef/cookbooks/cinder/files/default 16:44:09 zehicle_: so the only way to use it officially is through Crowbar? 16:44:14 that's the plan... would have to be Havana at this point. 16:44:24 zehicle_: so the equallogic doesn't connect directly? 16:44:27 to cinder 16:44:30 *officially* - the code's there 16:44:45 zehicle_: got it :-) 16:44:57 zehicle_: ok 16:45:15 EQL acts just like any cinder plug-in. it sets up the iSCSI targets for the VMs 16:45:35 zehicle_: but it doesn't run on the eql 16:45:39 correct 16:45:50 it's running on the controller/ 16:45:52 ? 16:46:08 right, it connects to the SSH interface for the EQL 16:46:14 got it 16:46:22 so it's like a technology bridge 16:46:53 hmmm, I'd say that it's using SSH to access the API 16:46:56 zehicle_: we have an EQL 6xxx here, I guess we'll give it a try :-) 16:47:09 zehicle_: i just acquired one too, i'm going to give it a try as well 16:47:28 alright anyone have anything addtional to add? 16:48:10 alexpilotti: ociuhandu ? 16:48:18 I have a question about Tempest runs against the HyperV work 16:48:26 haha 16:48:27 ok 16:48:33 zehicle_: shoot 16:48:36 is there parity on it? if not, do you have an idea of the gaps? 16:48:58 no idea 16:49:07 it's on the list of todos 16:49:20 i'm assuming we're going to have to add bits when we get there 16:49:28 no prob - it's something we can check when we spin up the CB deploy (since that's part of the CB install) 16:50:28 zehicle_: anything else to add? 16:50:29 from my work on the Board side, there's going to be more emphasis on status of Tempest tests 16:50:39 primeministerp, hanrahat: we should IMO consider this as part of the work we have to do ASAP 16:50:40 zehicle_: yes we've been following the thread 16:50:41 which, IMHO, is a very good thing for the project 16:51:00 alexpilotti: yes, indeed 16:51:17 We'll need to think if that's a grizzly or havana challenge 16:51:24 there are some changes going on here which will hopefully address this from my perspective 16:51:30 zehicle_: we want to get the driver into B category as soon as the CI is ready 16:51:40 primeministerp: agreed... let's discuss among the three of us later this week. can you set up a meeting? 16:51:52 hanrahat: i need to wait until hashir is back 16:52:05 hanrahat: a lot is dependant on changes on our team 16:52:10 primeministerp: that's fine 16:52:11 hanrahat: from my limited knolege 16:52:22 primeministerp: IMO this is before the CI stuff gets done 16:52:33 primeministerp: we need to be sure that the tests run fine 16:52:40 alexpilotti: definately 16:52:42 primeministerp: as in no Linux dependencies, etc 16:52:48 alexpilotti: agreed 16:53:03 alexpilotti: I'll try to schedule soemthing for the end of the week 16:53:20 zehicle_: when you refer to tempest, you refer to a gate or just compliance with the tests? 16:55:20 alexpilotti: he's referring to tempest as being the scorecard for complaince 16:55:27 yy 16:55:40 alexpilotti: and that will feed into refstack 16:55:50 i'm assuming at some level 16:55:56 primeministerp: yep, but those are two different stages 16:56:00 yes 16:56:18 getting the tests running and "green" is a thing that we can do w/o the CI in place 16:56:30 the gate is the next step that involves the CI 16:56:31 ideally both - we'd like to be able to vote on gating based on multi-node & multi-os deploys 16:56:48 agreed 16:57:00 but even getting a refstack report would be a good step 16:57:04 +1000 16:57:05 alexpilotti: if you want to put resources on tempest then feel free 16:57:22 alexpilotti: it's on the list of todo's and sooner is always better 16:57:22 yep, I will 16:57:28 alexpilotti: excellent 16:57:55 looks like we're almost out of time 16:57:58 Starting already from this week 16:58:04 alexpilotti: good 16:58:16 so additional comments? 16:58:42 alrighty then, i'll end it 16:58:46 thanks everyone for the time 16:58:51 #endmeeting