13:01:25 <claudiub> #startmeeting hyper-v
13:01:31 <openstack> Meeting started Wed Sep 14 13:01:25 2016 UTC and is due to finish in 60 minutes.  The chair is claudiub. Information about MeetBot at http://wiki.debian.org/MeetBot.
13:01:31 <atuvenie> hi all
13:01:32 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
13:01:34 <openstack> The meeting name has been set to 'hyper_v'
13:01:38 <claudiub> helloo
13:01:48 <abalutoiu> hello
13:02:19 <sagar_nikam> Hi
13:02:42 <claudiub> anyone else joining?
13:02:59 <lpetrut> Hi
13:03:18 <sagar_nikam> i think no
13:03:45 <claudiub> #topic monasca patches status
13:04:36 <claudiub> soo, we had a meeting last week
13:04:46 <claudiub> and we discussed some of the patches
13:05:05 <claudiub> they also asked some documentation, on how to setup the monasca-agent on Windows.
13:05:32 <claudiub> #link monasca-agent on Windows documentation: https://review.openstack.org/#/c/370066/
13:06:07 <claudiub> forgot to do git review until a few hours ago.
13:06:25 <claudiub> monasca meeting is in 2 hours, will bug the cores about this again.
13:06:56 <claudiub> other than that, nothing new.
13:07:01 <claudiub> questions?
13:07:24 <sagar_nikam> ok
13:07:26 <sagar_nikam> nice
13:07:35 <sagar_nikam> lets try to get the patches merged
13:07:40 <sagar_nikam> at the earliest
13:07:43 <claudiub> yep
13:07:47 <sagar_nikam> no other questions
13:07:57 <sagar_nikam> from my side on this topic
13:08:04 <claudiub> ok
13:08:10 <sagar_nikam> if you need any help from my side, let me know
13:08:25 <claudiub> sure, let's see how the monasca meeting goes
13:08:59 <claudiub> #topic Release status
13:09:14 <claudiub> sooo, this week there will be a newton release candidate
13:09:21 <claudiub> for all projects
13:09:47 <claudiub> meaning that this cycle is almost over, and the master branch will soon be opened again
13:10:15 <sagar_nikam> ok
13:10:29 <claudiub> final release will be at the beginning of next month
13:10:48 <claudiub> we'll have to rebase quite a few nova patches until then. :)
13:11:24 <claudiub> also, we're going to have a new PyMI release sometime soon, with some performance improvements
13:11:35 <sagar_nikam> FC patch... can that be taken at higher priority for O ... i think only nova patch is required for the feature to be completed
13:11:58 <claudiub> sagar_nikam: yep, that's the only patch remaining.
13:12:39 <lpetrut> that patch will just have to be rebased/reviewed. Hopefully, we can get it in at the beginning of O
13:12:40 <claudiub> sagar_nikam: and yeah, that will be prioritized, alonside the ovs vif plug patch
13:12:52 <sagar_nikam> if we start that at the beginning of the release, we can get it merged
13:13:27 <lpetrut> so, just a small reminder, the patch is done already and properly tested by the os-brick CI, not much else to do than push people to review it :)
13:14:06 <claudiub> i'll throw a glance on that patch as well. I forgot how it looks like. :)
13:14:27 <sagar_nikam> ok
13:15:59 <claudiub> anyways. just a small observation. During testing, we've noticed that Hyper-V Server 2016 has better performance than 2012 r2
13:16:19 <claudiub> so, good news. :)
13:16:29 <sagar_nikam> wow !!!
13:16:31 <sagar_nikam> cool
13:16:41 <sagar_nikam> we dont have 2016 yet
13:16:51 <sagar_nikam> so our scale tests will still be in 2012 r2
13:17:23 <claudiub> well, it wasn't released yet, there's only a technical preview
13:17:32 <claudiub> but we're ready for its release
13:17:41 <sagar_nikam> yes .. i meant we dont have TP
13:17:56 <lpetrut> we've done some improvements on the PyMI side, ensuring that WMI operations are non-blocking (basically allowing more of them to run at the same time when requested within greenthreads) and we had some significant performance boost
13:18:19 <sagar_nikam> lpetrut: nice
13:18:49 <lpetrut> during simple rally tests that were spawning/destroying instances, we got a 27% time improvement on WS 2012 r2 and more than 70% on WS 2016
13:18:56 <sagar_nikam> we are currently on the latest PyMI
13:19:10 <sagar_nikam> whatever version is available on pypi
13:19:22 <lpetrut> yeah, we'll have to do a new PyMI release soon
13:19:30 <claudiub> yeah, already mentioned that. :)
13:19:45 <sagar_nikam> lpetrut: 27% improvement... due to the pyMI change ?
13:19:49 <claudiub> thanks for the exact numbers though, I didn't know them. :)
13:19:56 <lpetrut> yep
13:20:06 <sagar_nikam> ok
13:20:22 <sagar_nikam> does it in any way improve the scale ?
13:20:31 <sagar_nikam> 27% perf improved
13:20:44 <sagar_nikam> how much scale improvement ?
13:21:41 <claudiub> it means that the boot / destroy actions are 27% faster
13:21:54 <claudiub> meaning that instances will spawn and be destroyed faster
13:22:40 <sagar_nikam> claudiub: i got that.. my question... does it in anyway improve the number of VMs and its networks per hyperv host ?
13:23:06 <sagar_nikam> for same flavors
13:23:32 <claudiub> do you mean if more VMs will be able to fit on a hyper-v host due to these pymi improvements?
13:23:41 <lpetrut> not really, it's just that we're doing more operations in parallel, thus saving time
13:23:42 <sagar_nikam> yes
13:23:58 <sagar_nikam> if disk and cpu and memory are still available
13:24:20 <sagar_nikam> ok ... i will explain... what we had observed with wmi
13:24:27 <claudiub> sagar_nikam: oh, you mean how much disk / cpu / memory is consumed by the host?
13:25:02 <sagar_nikam> the hyperv host has X CPU/memory/disk
13:25:10 <sagar_nikam> with that X
13:25:18 <sagar_nikam> we could spawn only Y VMs
13:25:23 <sagar_nikam> after we moved to pyMI
13:25:39 <sagar_nikam> we could spawn Y+ VMs
13:25:55 <sagar_nikam> since pymi was responding much faster
13:26:05 <sagar_nikam> in the same time frame
13:26:42 <lpetrut> I get now what you mean, yep
13:27:03 <claudiub> oh i see
13:27:20 <sagar_nikam> for example ... in 1 hour, if we could spawn 10 VMs in WMI, in the same 1 hour we could spawn 10+ VMs
13:27:28 <sagar_nikam> dont know if it was 20VMs
13:27:33 <sagar_nikam> or lesser
13:27:42 <sagar_nikam> but yes, we saw a big diff
13:27:55 <sagar_nikam> in the actual number of VMs and netwroks
13:27:59 <sagar_nikam> on the same host
13:28:09 <sagar_nikam> so my question now
13:28:15 <sagar_nikam> with this new change
13:28:20 <sagar_nikam> of pyMI
13:28:27 <sagar_nikam> can we now spawn more VMs ?
13:28:51 <claudiub> yeah, I'd say you'll be able to spawn more VMs within the same timeframe.
13:29:00 <sagar_nikam> ok cool...
13:29:22 <sagar_nikam> let me see if the scale tests using Mitaka, can we use te new pyMI
13:29:28 <sagar_nikam> will let you know
13:29:43 <sagar_nikam> if we use it and how much it improved
13:30:42 <claudiub> ok, cool :) as we've said, this new version of pymi wasn't released yet, still testing, so we make sure there won't be any regressions
13:30:56 <claudiub> but we'll have it ready before newton is released.
13:31:00 <sagar_nikam> yes, we try to use it when released
13:31:12 <sagar_nikam> anyhow the scale test schedule is still not planned
13:31:24 <sagar_nikam> lab needs to be booked
13:31:25 <sagar_nikam> etc
13:31:28 <sagar_nikam> so will take time
13:31:53 <claudiub> well, we might release pymi until then. :)
13:32:22 <claudiub> #topic open discussion
13:32:49 <claudiub> so, nothing new from me
13:33:05 <sagar_nikam> ok
13:33:17 <sagar_nikam> nothing new from me as well
13:33:19 <claudiub> if you have some topics to discuss, now is the time. :)
13:33:30 <sagar_nikam> not much from my side
13:34:06 <sagar_nikam> one question
13:34:09 <claudiub> sure
13:34:14 <sagar_nikam> since lpetrut: is here
13:34:23 <sagar_nikam> lpetrut: how is freezer
13:34:36 <sagar_nikam> i know you will be working on it in O
13:34:54 <sagar_nikam> but i have heard that freezer already has some windows support
13:34:58 <sagar_nikam> have you tried that
13:36:39 <claudiub> afaik, freezer supports windows hosts, but not Hyper-V VMs
13:37:05 <claudiub> and that is one of our targets at the moment
13:37:44 <sagar_nikam> so the plan for O is to add backup of HyperV VMs only ?
13:38:06 <sagar_nikam> if i use freezer as it is today... hyperv host can be backed up ?
13:40:55 <lpetrut> so, freezer will allow you to back up the volume storing the VM files, VSS will be used but it will not interact with the Hyper-V VSS writers directly
13:41:32 <lpetrut> my main goal for freezer during the next release will be to improve the interaction with VSS in order to achieve app consistent guest aware backups
13:41:34 <sagar_nikam> ok.. thanks
13:41:49 <sagar_nikam> guest aware ?
13:41:57 <sagar_nikam> can you explain further please
13:43:10 <lpetrut> sure, when creating VM backups, the guests can take part of the backup process to ensure that the backup will be app consistent
13:43:35 <sagar_nikam> ok
13:43:45 <sagar_nikam> got it
13:44:53 <claudiub> well, if there's nothing else, we can end it here. :)
13:45:16 <sagar_nikam> nothing much from my side
13:45:46 <claudiub> ok then, thanks folks for joining, see you next week!
13:46:00 <claudiub> #endmeeting