16:01:27 <primeministerp> #startmeeting hyper-v
16:01:28 <openstack> Meeting started Tue Sep  3 16:01:27 2013 UTC and is due to finish in 60 minutes.  The chair is primeministerp. Information about MeetBot at http://wiki.debian.org/MeetBot.
16:01:29 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
16:01:31 <openstack> The meeting name has been set to 'hyper_v'
16:01:34 <primeministerp> hi everyone
16:01:53 <primeministerp> we'll wait a couple more minutes for the others to join
16:02:27 <zehicle_at_dell> Hey!
16:02:30 <primeministerp> hey rob
16:02:44 <zehicle_at_dell> Hi Peter
16:03:04 <primeministerp> back from vacation with a bang today
16:03:07 <primeministerp> ;)
16:03:56 <alexpilotti> hola
16:04:02 <primeministerp> hi alex
16:04:12 <primeministerp> i know pedro's time is limited today
16:05:01 <primeministerp> so if we can start
16:05:20 <primeministerp> #topic review backlog
16:06:00 <primeministerp> so obviously we still have the problem of our backlog of reviews
16:06:24 <primeministerp> and it looks like unless we get a priority bump on our blueprints
16:06:40 <primeministerp> we might not get them reviewed
16:07:19 <alexpilotti> I'm frankly quite concerned about the situation
16:07:22 <primeministerp> well
16:07:24 <primeministerp> so am i
16:07:41 <alexpilotti> this will lead straight towards a fork for most drivers IMO
16:07:51 <alexpilotti> we are not the only ones in this situation
16:08:08 <alexpilotti> basically driver specific blueprints are marked as low priority
16:08:10 <primeministerp> has there been any other discussion of this from the other hypervisors, like xen and vmare?
16:08:33 <alexpilotti> in the ML there was already some discussion
16:08:43 <alexpilotti> I forwarded you already some links
16:08:45 <primeministerp> still got 500+ to go
16:08:58 <alexpilotti> in general look for "frustration" in the ML subjects and there are quite a few :-)
16:09:04 <primeministerp> hahaha
16:09:06 <primeministerp> thanks
16:09:11 <primeministerp> unfortunately
16:09:23 <primeministerp> i wish it was something different
16:09:30 <alexpilotti> let me see if we can get Russell
16:09:35 <alexpilotti> russellb: ping
16:11:30 <dansmith> so
16:11:31 <russellb> hey
16:11:36 <primeministerp> hey guys
16:11:39 <dansmith> something has to be prioritized, right?
16:11:39 <primeministerp> thanks for the time
16:11:40 <alexpilotti> hi ru
16:11:43 <alexpilotti> hi russellb
16:11:49 <primeministerp> dansmith: obviously
16:12:06 <alexpilotti> we are not putting into discussion your hard work
16:12:07 <dansmith> we're merging crazy amounts of code, and reviewing even more
16:12:31 <alexpilotti> btw thanks dansmith for the help on the WMIV2 one
16:12:51 <alexpilotti> at the moment, not even a single one of our Nova Bps got merged
16:12:55 <russellb> honestly, the root cause here is how much was delivered right at the end of the cycle
16:13:02 <russellb> (not just you, across the project)
16:13:09 <russellb> review bandwidth doesn't magically increase at the end :-)
16:13:30 <alexpilotti> sure, but blueprints can be retargetted
16:13:41 <russellb> hm?
16:13:53 <alexpilotti> for example by knowing the situation before we could have decided to move stuff to I
16:14:10 <alexpilotti> instead of running aroung on a gazillion reviews and blueprints for nothing
16:14:14 <russellb> the blueprint list for h3 has been insane since mid-cycle
16:14:23 <russellb> and i sent multiple messages to the ML about this
16:14:28 <russellb> and brought it up many times in weekly meetings
16:14:29 <zehicle_at_dell> is there an thought that some of the code may have to move to I if the backlog is not resolved?
16:14:32 <alexpilotti> well, somebody approved them
16:14:46 <russellb> zehicle_at_dell: yes, whatever doesn't make it will move out
16:15:01 <russellb> alexpilotti: an approval isn't a guarantee that it will make it
16:15:04 <alexpilotti> so russellb, what is the realistically expectation for our stuff?
16:15:18 <russellb> hopefully we can get your most important stuff, some of it probably won't make it
16:15:28 <russellb> due to being caught up with so many other things late in the dev cycle
16:15:45 <russellb> i get that your v2 API stuff is important for platform support
16:15:56 <primeministerp> russellb: is there possibilities of the others making it as a feature freeze exception?
16:15:56 <russellb> the others seem less likely
16:16:06 <russellb> you can apply
16:16:16 <russellb> will have to sort through how many applications there are
16:16:18 <alexpilotti> well WMI V2 is required for Nova to work on Hyper-V V2
16:16:22 <russellb> understood
16:16:27 <alexpilotti> sorry, on Hyper-V 2012 R2
16:16:33 <russellb> so that's a good candidate for an exception if it doesn't make it
16:16:47 <primeministerp> russellb: unfortunately most of ours depend on that one
16:16:49 <alexpilotti> the small ceilometer patch is required for… Ceilometer
16:17:04 <alexpilotti> the rest is basically on cascade on the WMI V2 one
16:17:12 <russellb> ok.
16:17:28 <alexpilotti> so we spent one month rebasing patches basically
16:17:50 <alexpilotti> the WMI V2 one has been put as "high" in Neutron, for example
16:18:03 <alexpilotti> while in Nova is just a "low" one among the rest
16:18:40 <alexpilotti> and the WMI V2 is is just to have Nova running on the new Hyper-V
16:18:55 <dansmith> I definitely understand the importance of that,
16:18:56 <alexpilotti> so basically, we risk to end up having 0 new features in Havana
16:19:09 <russellb> i told you V2 is a good candidate for an exception
16:19:12 <alexpilotti> as I wouldn't count compatibility as a feature
16:19:13 <russellb> i think that one can make it at least
16:19:17 <russellb> well, ok.
16:19:24 <russellb> then the lesson is don't deliver at the end of the cycle :-)
16:19:34 <russellb> or you're at risk of this happening
16:19:36 <alexpilotti> July?
16:19:50 <russellb> not V2, the others
16:20:12 <russellb> but anyway
16:20:13 <alexpilotti> only a couple of them got in late (as in Aug 21th)
16:20:16 <russellb> what would you like me to do?
16:20:31 <alexpilotti> exceptions would be great of course
16:20:43 <alexpilotti> all that stuff is simple 100% included in the driver
16:20:46 <russellb> we'll cross that bridge after feature freeze, you're welcome to apply
16:20:51 <alexpilotti> only the API stuff it's outside
16:21:10 <dansmith> exceptions don't increase review bandwidth, and it has to be weighed against other stuff that also needs an exception, right?
16:21:39 <russellb> dansmith: yeah, we'll have to wait to see what exception applications we get
16:21:40 <alexpilotti> if "welcome to apply" means getting ignored, I don't think it's too useful
16:21:56 <primeministerp> I'll second that statement
16:21:58 <russellb> i'm not able to give an answer without having the other applications in front of me
16:22:12 <primeministerp> russellb: that's fair
16:22:13 <russellb> just raging at me isn't going to help :-)
16:22:22 <primeministerp> russellb: not trying to rage at anyone
16:22:27 <dansmith> I take offense at the suggestion that we'd ask you to apply just to ignore you
16:22:28 <alexpilotti> russellb: I'm not raging at you :-)
16:22:29 <russellb> but i promise to review every application and give a response
16:22:31 <primeministerp> russellb: just trying to find a resolution to our current situation
16:22:41 <russellb> you both suggesting that i would ignore you is pretty frustrating
16:22:42 <alexpilotti> nor I think that you should become the scapegoat of this situation
16:22:50 <russellb> i'm taking time out of one of the busiest days of the year to talk here
16:22:58 <russellb> yet i'm ignoring?  come on.
16:23:07 <russellb> anyway, i hope we can review some of your stuff
16:23:12 <russellb> definitely want to get V2 in for you
16:23:21 <russellb> the rest, please apply for exceptions and we'll sort it out soon
16:23:23 <russellb> thanks guys
16:23:27 * russellb goes back to reviewing
16:23:30 <primeministerp> russellb: thank you for your time
16:23:31 <alexpilotti> add please the ceilometer one
16:23:46 <primeministerp> russellb: the intent wasn't to offend
16:23:58 <primeministerp> russellb: dansmith: sorry if we did
16:24:19 <primeministerp> ok
16:24:29 <alexpilotti> well, back to the meeting
16:24:35 <primeministerp> yep
16:24:36 <primeministerp> so
16:24:47 <alexpilotti> what I fear is that we're gonna have a fork
16:25:04 <alexpilotti> all the stuff that won't make it, will go into our github repo
16:25:14 <alexpilotti> the installer will be based on that
16:25:22 <primeministerp> not what i wanted
16:25:27 <alexpilotti> and the same will probably happen for the other hypervisors
16:25:45 <alexpilotti> I don't want this to happen
16:25:52 <alexpilotti> but I don't see any way out
16:26:25 <primeministerp> let's wait a day or two, see where we stand then decide what needs a feature freeze request
16:26:40 <primeministerp> and start that process
16:27:07 <alexpilotti> one day actually, as the freeze in on the 4th :-)
16:27:12 <hanrahat> i have a long term concern, too.  i don't think this was a case of late submissions.  the process is overwhelmed.  how does this get fixed in future releases... 'I' and beyond... i've followed the discussion about options, but there wasn't any resolution.
16:27:47 <primeministerp> hanrahat: that's a discussion above this meeting
16:28:06 <primeministerp> this is most likely going to be an issue next release too
16:28:21 <primeministerp> and not just for us
16:29:28 <alexpilotti> the solution IMO is increasing the review bandwidth
16:29:43 <alexpilotti> we're preparing a guy for this work
16:29:49 <alexpilotti> to do our part
16:30:13 <alexpilotti> the other important part of the solution is not to approve so many blueprints
16:30:21 <alexpilotti> if you already know you cannot make it
16:30:53 <alexpilotti> and the last one is better communication with the blueprint owners
16:31:01 <primeministerp> alexpilotti: i think there's always optisism that they will make it
16:31:39 <alexpilotti> sure, hope beer dies :-)
16:31:43 <alexpilotti> loool
16:31:49 <alexpilotti> damn autocorrector
16:31:51 <primeministerp> haha
16:31:56 <primeministerp> was wondering what that meant
16:31:59 <alexpilotti> meant "never"
16:32:16 <alexpilotti> I see this as a positive sign ;-)
16:32:22 <primeministerp> me too
16:32:32 <primeministerp> let's move on then
16:32:40 <alexpilotti> agree
16:32:42 <primeministerp> only other thing i had was to talk puppet bits
16:32:46 <primeministerp> but luis didn't make it today
16:33:05 <primeministerp> so I'll hold off on that, as I still have much to get cought up on.
16:33:23 <primeministerp> so one other thing
16:33:29 <primeministerp> to discuss is live-migration
16:33:36 <primeministerp> #topic live-migration
16:34:00 <alexpilotti> thanks primeministerp
16:34:08 <primeministerp> alexpilotti: you had something to add to this
16:34:13 <alexpilotti> so there's a new cool feature added by vishy
16:34:24 <alexpilotti> which most probably won't make it
16:34:47 <alexpilotti> but as there was a chance to get a shot at it, we implemented it in Hyper-V anyway
16:34:57 <primeministerp> great
16:35:05 <alexpilotti> live-migration is a "host" snapshot of a VM
16:35:16 <alexpilotti> basically, what we call snapshots in Hyper-V
16:35:35 <alexpilotti> the VM gets snapshotted, the image + memory dump uploaded in Glance
16:35:47 <alexpilotti> and from there you can spawn new images out of it
16:36:00 <alexpilotti> the advantage is that VM boot time is almost 0
16:36:00 <primeministerp> so storing running state as an image
16:36:37 <alexpilotti> think about VDI, possibly in conjunction with RemoteFX
16:36:44 <primeministerp> yep
16:36:54 <primeministerp> very interesting use case for it
16:36:56 <alexpilotti> you can have running VDI images in a matter of 1-2 seconds
16:37:37 <alexpilotti> and since a differential disk is used also for the snapshot disks
16:37:47 <primeministerp> disk image is tiny too
16:37:49 <alexpilotti> there's basically no disk usage
16:38:01 <alexpilotti> it's a great feature IMO
16:38:17 <alexpilotti> it will not make it for a single reason:
16:38:25 <primeministerp> being?
16:38:39 <alexpilotti> it's not ready in libvirt / qemu
16:38:48 <primeministerp> ahh
16:38:49 <alexpilotti> as in the project, not the Nova driver
16:38:53 <primeministerp> got it
16:39:01 <alexpilotti> while the APIs are perfectly working
16:39:30 <alexpilotti> but they don't want to merge them as there's no driver implementing it
16:39:49 <alexpilotti> that's why I gave a shot at it
16:40:01 <russellb> fwiw guys, i'm going to have a design summit session on the general future of compute drivers in nova
16:40:02 <primeministerp> good thinking
16:40:18 <alexpilotti> russellb: cool
16:40:38 <primeministerp> russellb: great, i'm hoping to have team there, I'll make sure it's on the required viewing list
16:40:40 <alexpilotti> russellb: since you are around, no chance I guess for the live-snapshot apis?
16:41:02 <russellb> no
16:41:42 <primeministerp> alexpilotti: do we have anything else to discuss
16:41:43 <alexpilotti> russellb: lol, I had to ask but I didn't have big hopes :-)
16:42:03 <alexpilotti> primeministerp: yes, the positive news, aka what merged:
16:42:10 <primeministerp> alexpilotti: go for it
16:42:23 <alexpilotti> pnavarro's Cinder patch merged (yeiii)
16:42:31 <alexpilotti> tx pnavarro! :-)
16:42:37 <primeministerp> alexpilotti: yes he brough me up to speed on that this morning
16:42:43 <primeministerp> thanks again pnavarro !
16:42:47 <alexpilotti> ceilometer merged
16:42:52 <primeministerp> also great news
16:43:02 <alexpilotti> Neutron Hyper-V V2 merged
16:43:16 <alexpilotti> a bunch of bug fixes
16:43:30 <alexpilotti> and a bunch of Nova pieces
16:43:43 <alexpilotti> none of them make an entire blueprint unfortunately yet
16:43:53 <primeministerp> gotcha
16:44:05 <alexpilotti> johnthetubaguy was incredibly helpful yesterday and today in helping out with WMI V2
16:44:25 <primeministerp> johnthetubaguy: thanks for the help!
16:44:26 <alexpilotti> so we have one +2 to go for having this feature merged
16:44:47 <alexpilotti> and basically this is it
16:44:50 <johnthetubaguy> no worries, happy to help out
16:44:52 <zehicle_at_dell> let me know when we're ready to talk on the Crowbar HyperV progress
16:45:02 <primeministerp> zehicle_at_dell: we can add that as a topic as well
16:45:05 <primeministerp> let's do it now
16:45:07 <zehicle_at_dell> thanks
16:45:12 <primeministerp> #topic crowbar/hyper-v
16:45:15 <alexpilotti> zehicle_at_dell: sure
16:45:28 <primeministerp> zehicle_at_dell: the floor is yours
16:45:52 <zehicle_at_dell> There are a lot of pull requests in the queue for HyperV support
16:46:46 <alexpilotti> zehicle_at_dell: you mean our requests or pulls on our barclamp?
16:46:57 <zehicle_at_dell> both
16:47:15 <zehicle_at_dell> it looks like they are going to settle pretty soon
16:47:24 <zehicle_at_dell> and we're ready to branch and start the havana work in ernest
16:47:36 <alexpilotti> ok
16:48:22 <alexpilotti> we merged lately some pulls on the Hyper-V barclamp
16:48:26 <zehicle_at_dell> great!
16:48:32 <ociuhandu> hy all, sorry for being late
16:48:40 <alexpilotti> basically in the day we got them, afaik
16:48:44 <zehicle_at_dell> for Havana, we'll use PFS
16:49:11 <zehicle_at_dell> so, we should be able to bring in the code even if it has not been through review & into trunk
16:49:11 <alexpilotti> ociuhandu: do you have some more details on the pulls on the Hyper-V barclamp?
16:49:39 <zehicle_at_dell> that's part of the goal - I think it would be easier to get reviews if you'd been able to show integratation
16:50:14 <ociuhandu> yes, we have only 2 left, one for SSL and one that's from today, some code cleanuo
16:50:15 <alexpilotti> integration? as in having a CI framework in place?
16:51:04 <primeministerp> alexpilotti: was that for me?
16:51:10 <primeministerp> alexpilotti: that last question?
16:51:19 <alexpilotti> no, for zehicle_at_dell
16:52:01 <alexpilotti> zehicle_at_dell: ping
16:52:12 <zehicle_at_dell> sorry
16:52:23 <alexpilotti> np!
16:52:56 <zehicle_at_dell> our primary goal has been to make sure they are not breaking other stuff
16:53:28 <zehicle_at_dell> we have people who shoudl be able to do that - I'll get them connected to you
16:53:53 <zehicle_at_dell> that's grizzly target'
16:54:06 <zehicle_at_dell> but I'm interested in migrating that work to Havana soon
16:54:18 <zehicle_at_dell> because I think it will be easier to make adjstments and fast feedback
16:54:54 <alexpilotti> ok, that's not too complicated for us
16:54:58 <zehicle_at_dell> so, expect a branch where we can work on Havana.  then we can target the hyperv code to your github
16:55:10 <alexpilotti> ok
16:55:13 <primeministerp> zehicle_at_dell: perfect
16:55:22 <zehicle_at_dell> the key is that I want to align our internal tests, they need to be able to get a workable build
16:55:32 <zehicle_at_dell> because they are not in a position to integtrate directly from the code
16:55:39 <alexpilotti> one question: let's say that some of the Havana stuff under review does not make it for Havana
16:55:48 <alexpilotti> do we still include it in Crowbar?
16:55:53 <zehicle_at_dell> yes
16:55:59 <alexpilotti> good
16:56:10 <primeministerp> I agree to that statement
16:56:15 <zehicle_at_dell> we do that for advanced features, it's one of the reasons that we use PFS
16:56:21 <zehicle_at_dell> because sometimes, you're working ahead of trunk
16:56:25 <zehicle_at_dell> (esp w/ these backlogs)
16:57:01 <zehicle_at_dell> IMHO, if we can show it's working against trunk on a multi-node system then it should be easier to get reviews to flow
16:57:13 <primeministerp> zehicle_at_dell: agreed
16:57:45 <primeministerp> ok guys we're coming to the end of our time
16:57:59 <alexpilotti> zehicle_at_dell (and primeministerp) how do you plan to organise the testing?
16:57:59 <zehicle_at_dell> I know that there's a balance between the Grizzly work that's trying to land and getting Havana integrations
16:58:24 <primeministerp> alexpilotti: that's still under discussion right now
16:58:28 <zehicle_at_dell> I've been waiting for a green light from you that we've got a usable build
16:58:40 <zehicle_at_dell> and then have that team work 1x1 so that they can replicate your environemtn
16:58:54 <zehicle_at_dell> first, they'll baseline that build w/ the defaults
16:59:10 <zehicle_at_dell> and then add in the HyperV parts in collabroation w/ you
16:59:29 <zehicle_at_dell> I'm expecting that we'll create a doc in the process so the community can replicate the process
16:59:47 <alexpilotti> cool
16:59:53 <ociuhandu> zehicle_at_dell: sounds great
16:59:54 <zehicle_at_dell> originally, I'd thought to do that w/ grizzly bits but I think we may want to consider Hanava
16:59:55 <primeministerp> zehicle_at_dell: ideally we'll be able to share our test plans/cases as well
17:00:00 <zehicle_at_dell> whioch would be more helpful?
17:00:24 <primeministerp> zehicle_at_dell: probably havana
17:00:30 <zehicle_at_dell> primeministerp: yes, that makes sense
17:00:47 <zehicle_at_dell> +1 on Havana.  We just have to get the baseline set
17:00:54 <zehicle_at_dell> ok, we're out of time.  this was helpful.  Thanks
17:01:03 <primeministerp> ok guys, let's free up the channel, we can continue the discussion int he hyper-v channel if necessary
17:01:09 <primeministerp> thanks everyone
17:01:15 <primeministerp> #endmeeting