15:00:27 <johnthetubaguy> #startmeeting XenAPI
15:00:28 <openstack> Meeting started Wed Jan  8 15:00:27 2014 UTC and is due to finish in 60 minutes.  The chair is johnthetubaguy. Information about MeetBot at http://wiki.debian.org/MeetBot.
15:00:29 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
15:00:31 <openstack> The meeting name has been set to 'xenapi'
15:00:33 <johnthetubaguy> Happy new year everyone!
15:00:39 <johnthetubaguy> who is around for today's meeting?
15:00:45 <BobBall> And to you too John
15:00:48 <BobBall> I might be
15:00:51 <thouveng> Hi, happy new year :)
15:00:56 <BobBall> although I've only just got back to my desk
15:01:00 <BobBall> so I'm still mostly at home.
15:01:13 <BobBall> I've been in the office for a grand total of 61 minutes!
15:01:25 <johnthetubaguy> BobBall: oh, just back? or not been well?
15:01:43 <BobBall> Long holiday because I wasn't allowed to take holiday near the conference  last year ;)
15:01:56 <johnthetubaguy> lol, you do right
15:02:04 <matel> hi, happy new year to everyone.
15:02:13 <johnthetubaguy> BobBall: I am just reviewing all your patches
15:02:25 <johnthetubaguy> OK, so lets get into the agenda
15:02:30 <johnthetubaguy> #topic Blueprints
15:02:32 <BobBall> Happy new year to you, and matel and thouveng !
15:02:46 <johnthetubaguy> thouveng: hows it going?
15:03:02 <thouveng> I posed a draft for the bp https://review.openstack.org/#/c/63816/2
15:03:08 <johnthetubaguy> sweet
15:03:24 <thouveng> now resource tracker is up to date
15:03:48 <matel> thouveng: is it a private patch?
15:03:48 <BobBall> I've started to review the patch thouveng
15:03:49 <thouveng> I can see that db is well updated when I create/remove instance
15:03:50 <johnthetubaguy> I would put get_passthrough_devices in xenhost I think
15:04:04 <BobBall> but I've got lots of stuff I have to do today to catch up
15:04:14 <johnthetubaguy> but anyways, yeah, I will take a look
15:04:23 <BobBall> thouveng: yes - I agree with john now - I figured out how to do it just before Christmas.
15:04:44 <johnthetubaguy> cool, so this ensures resource trackers knows about all PCI device?
15:04:44 <BobBall> thouveng: it's ugly, but you can have both old and new style plugins
15:04:54 <thouveng> ok I need to check this so
15:05:22 <thouveng> I don't look xenhost but I will do
15:05:33 <johnthetubaguy> old and new style plugins?
15:05:49 <BobBall> thouveng: Check out https://review.openstack.org/#/c/60253/6/plugins/xenserver/xenapi/etc/xapi.d/plugins/xenhost
15:06:08 <BobBall> johnthetubaguy: we didn't want to add new calls with the "old" non-serialised style
15:06:14 <BobBall> but xenhost is 100% old style
15:06:20 <BobBall> until my patch is accepted :)
15:06:21 <johnthetubaguy> BobBall: ah, I see
15:06:50 <BobBall> Check out line 408 of that link
15:06:57 <johnthetubaguy> cool, sounds like good progress is happening
15:07:06 <thouveng> ok I see. So you support both serialized and non-serialized that are old and new style plugin right?
15:07:08 <BobBall> Unfortunately the XAPI side of the plugin code is ugly to say the least
15:07:14 <BobBall> yes thouveng
15:07:27 <johnthetubaguy> BobBall: yuck
15:07:37 <johnthetubaguy> anyways
15:07:59 <BobBall> johnthetubaguy: it's cuz XAPI actually forceably returns after the "dispatch" call
15:08:10 <johnthetubaguy> nice
15:08:15 <BobBall> johnthetubaguy: so one file can't have both unless you do the horrid hack I added
15:08:19 <BobBall> not return - _exit_
15:08:24 <BobBall> because it's a process
15:08:30 <BobBall> gotta love those hardcoded assumptions
15:08:34 <johnthetubaguy> yup
15:08:43 <thouveng> So we agree that I should move the code into xenhost?
15:08:49 <johnthetubaguy> yeah, I think so
15:08:58 <BobBall> well - when you say yuck johnthetubaguy, are you happy witht he approach in my patch?
15:09:19 <johnthetubaguy> BobBall: yeah, I will take a look, but its what we have to do, so its fine
15:09:38 <BobBall> ok
15:09:43 <thouveng> BobBall: that means that my patch will depend on yours right?
15:09:47 <johnthetubaguy> re-writing all the plugins, while tempting, is a waste of everyones time
15:10:13 <johnthetubaguy> touveng: I would just copy the hack, and deal with the clash when it happens, but its up to you really
15:10:16 <BobBall> well I hope not thouveng - mine might take a while to get in.  I'd say let's both just have the same code in there and one of us will have to rebase
15:10:25 <johnthetubaguy> +1
15:10:34 <thouveng> ok for me
15:10:37 <johnthetubaguy> cool
15:10:38 <BobBall> it's only a few lines of nastyness
15:10:42 <johnthetubaguy> what is next for this blueprint?
15:11:05 <thouveng> I'm currently looking how to attach the pci devices during the spawn process
15:11:12 <thouveng> I updated the etherpad
15:11:13 <johnthetubaguy> ah, yes, lol, we do need that
15:11:16 <johnthetubaguy> cool
15:11:22 <johnthetubaguy> I am just looking at the blueprint
15:11:28 <johnthetubaguy> I see the two work items now, thats cool
15:11:49 <thouveng> my concern is that I'm doing for GPU so I will use something like VGPU.create I guess
15:12:25 <johnthetubaguy> ah, I see, then SRIOV says no
15:12:40 <thouveng> Exact.
15:12:56 <BobBall> uhhhh no
15:13:04 <thouveng> And currently you will see in my patchset that I'm using type-PCI
15:13:09 <BobBall> we're doing PCI pass through - not vGPU aren't we?
15:13:24 <BobBall> so as far as dom0 and nova are concerned it's just any PCI device
15:13:27 <johnthetubaguy> well, thats a good point, vGPU is sepearte
15:13:33 <johnthetubaguy> erm, seperate
15:13:50 <BobBall> vGPU will indeed need a separate XAPI call - but for now don't we just set other-config:pci-whatever-it-is?
15:14:28 <thouveng> I agree with you BobBall, we are doing PCI passthrough but my though was to first do GPU and then SR-IOV and other device.
15:14:41 <thouveng> Ah ok if you uyse other-config:pci-whatever-it-is
15:15:02 <BobBall> Agreed - but my point is that if it's GPU rather than vGPU then it shouldn't be any different to SR-IOV... I think
15:15:05 <BobBall> :)
15:15:27 <johnthetubaguy> its more complicated than that sadly
15:15:32 <johnthetubaguy> well eventually anyways
15:15:45 <johnthetubaguy> the virtual device might need config from neutron :(
15:16:04 <johnthetubaguy> but anyways other-config:pci-… seems like the best first step
15:16:11 <BobBall> oh this is where the fun starts! :)
15:16:26 <thouveng> ok so I will go for other-config:pci-
15:16:31 <johnthetubaguy> yeah, I am doing daily meetings this week to try and setup some stuff
15:17:44 <johnthetubaguy> anyways, sounds like we are all good
15:17:46 <matel> Are we done with this topic?
15:17:48 <johnthetubaguy> any more blueprinty stuff?
15:17:53 <thouveng> I think so
15:17:53 <johnthetubaguy> +1
15:18:10 <johnthetubaguy> #topic QA
15:18:18 <johnthetubaguy> so, upstream testing...
15:18:22 <matel> Official XenAPI testing: https://review.openstack.org/#/c/60581/
15:18:28 <johnthetubaguy> matel: how goes it
15:18:35 <matel> This is the script, that makes xs work in the "cloud"
15:18:49 <matel> It's a really complicated thing, and I am worried about it.
15:19:04 <johnthetubaguy> yeah, lets try simplify it over time
15:19:11 <johnthetubaguy> but we have a possible solution
15:19:17 <BobBall> Is there a particular part you're worried about matel?
15:19:18 <matel> I also experienced issues, when launching instances from snapshot, and the network devices were messed up.
15:19:28 <johnthetubaguy> ah, thats bad
15:19:39 <johnthetubaguy> subnet changed maybe?
15:19:46 <johnthetubaguy> different gateways, etc?
15:19:48 <matel> BobBall: look at the script - I tried to make it readable, but it's not the best thing to look at.
15:20:00 <matel> john: no, device numberings were changed somehow.
15:20:06 <BobBall> Agreed.  It's a confusing script.
15:20:06 <johnthetubaguy> oh, boy, nice
15:20:31 <matel> john: If I experience the same, I will ask you to look under the hood.
15:20:45 <johnthetubaguy> sure, I might need some tips on what to look for mind
15:20:56 <matel> Anyhow, what I am working on: Come up with an XVA, that will be able to run devstack.
15:21:05 <matel> That will be a saucy.
15:21:16 <johnthetubaguy> OK
15:21:26 <johnthetubaguy> an XVA rather than waiting for the full install?
15:21:34 <matel> And after that, I will come up with a localrc, that works.
15:21:52 <matel> This XVA is the equivalent of a cloud image.
15:21:58 <BobBall> hang on - don't we have an XVA that can run devstack already?  What do you mean you're creating?
15:22:13 <BobBall> ah - an image including XenServer and a devstack XVA
15:22:20 <matel> So, we expect to be able to access the devstack domU.
15:22:41 <johnthetubaguy> XVA is just the devstack VM right?
15:22:43 <matel> No, it is the appliance which will listen on the public IP.
15:22:53 <matel> yes, that's the devstack vm.
15:22:56 <johnthetubaguy> cool
15:23:29 <matel> We already have one, but: it's not a saucy, it needs more ram, etc.
15:23:55 <matel> At the moment the scripts that generate this xva live in a qa repository
15:24:03 <matel> I am extracting those bits to separate repos.
15:24:29 <matel> BobBall: any comments?
15:24:45 <johnthetubaguy> sounds good
15:24:56 <johnthetubaguy> I think
15:25:01 <BobBall> I'm not sure I understand yet
15:25:11 <matel> I would need some political help from you john, to somehow get more attention to these patches.
15:25:26 <BobBall> are you saying - apart from refactoring - the work you're doing is updating to saucy and giving the existing devstack XVA more ram?
15:25:29 <johnthetubaguy> OK, but I fear is not a group I have much influence with, but I can try
15:25:46 <johnthetubaguy> different networking config I guesS?
15:25:49 <matel> BobBall: Yes, that's the main change: more ram, more disk.
15:25:59 <matel> BobBall: and go to saucy.
15:26:16 <BobBall> OK
15:26:19 <matel> John: networking is set by the script.
15:26:28 <johnthetubaguy> OK, cool.
15:26:39 <johnthetubaguy> I would guard against more disk
15:26:44 <matel> So I would like to have a localrc by next week.
15:26:48 <johnthetubaguy> could we not just add an extra data disk?
15:27:13 <johnthetubaguy> anyways, we can just do it the easy way for now
15:27:59 <matel> I updated the nodepool patches, give them some love: https://review.openstack.org/#/c/61463/
15:28:08 <matel> #link https://review.openstack.org/#/c/65261/
15:29:19 <matel> I would like to note, that if we want all this to happen, we have to push it seriously - at least that's what I see.
15:29:26 <johnthetubaguy> Yep
15:29:39 <johnthetubaguy> so what is blocking us right now, review bandwidth?
15:29:49 <matel> Yes, the nodepool changes.
15:29:55 <BobBall> Agreed
15:29:56 <johnthetubaguy> can we test more of this in a "dev"/fake environment first?
15:29:57 <BobBall> ok
15:30:10 <BobBall> I'll commit to reviewing it this week
15:30:11 <matel> Yes, that's missing.
15:30:28 <matel> So trying out this whole stuff with nodepool would be nice.
15:30:33 <BobBall> then we can push in -infra to get some progress
15:30:37 <BobBall> I guess that setup would be hard
15:30:42 <matel> But I first wanted to get some feedback from the infra guys
15:31:05 <matel> I don't know, BobBall, I haven't tried it.
15:31:11 <johnthetubaguy> Ok, we need to go into -infra and ask nicely I think
15:31:31 <johnthetubaguy> I think we should try that suggested dev setup first though
15:31:41 <johnthetubaguy> I think you found the link somewhere
15:31:45 <matel> john: I would like to get some feedback as well.
15:31:54 <johnthetubaguy> yeah, lets try both in parallel
15:31:56 <BobBall> it wouldn't surprise me if -infra had some useful dev setups they could spin this through easily
15:32:05 <johnthetubaguy> true
15:32:09 <johnthetubaguy> we can but ask
15:32:29 <matel> I would expect these software pieces to be easily testable.
15:32:32 <johnthetubaguy> are you guys good driving this, I am traveling next week, but should be able to help now and then, if its useful
15:33:31 <matel> We are doing our best, and will try to do the political side as well.
15:34:12 <matel> next()
15:34:28 <johnthetubaguy> yep, am seeing good progress
15:34:32 <johnthetubaguy> thank you!
15:34:49 <johnthetubaguy> lets keep moving this forward
15:35:00 <johnthetubaguy> else we will be out of tree in under a month or so
15:35:23 <matel> well...
15:36:05 <matel> Okay, are we done here?
15:36:33 <johnthetubaguy> possibly
15:36:41 <johnthetubaguy> #topic Open disscussion
15:36:44 <johnthetubaguy> any more?
15:36:47 <matel> Nope
15:36:51 <johnthetubaguy> about next weeks meeting
15:36:52 <BobBall> not from me
15:36:56 <thouveng> not for me
15:36:57 <johnthetubaguy> I am probably not around
15:37:03 <matel> Bob?
15:37:09 <johnthetubaguy> do you guys want to go ahead without me, and I will do my best to join?
15:37:22 <BobBall> Yes
15:37:26 <BobBall> I'll chair it
15:37:29 <johnthetubaguy> cool, thanks
15:37:38 <johnthetubaguy> I am back the following week, all being well
15:37:38 <matel> Okay, have a nice day!
15:37:39 <BobBall> given that thouveng is joining regularly it would be a shame not to have a chat! :)
15:37:46 <johnthetubaguy> +1
15:37:48 <thouveng> :)
15:37:52 <johnthetubaguy> cool, thanks all
15:37:55 <johnthetubaguy> happy coding
15:37:56 <johnthetubaguy> etc
15:38:00 <johnthetubaguy> #endmeeting