#openstack-meeting-3 log

13:02:15 <claudiub> #startmeeting hyper-v
13:02:16 <openstack> Meeting started Wed May 25 13:02:15 2016 UTC and is due to finish in 60 minutes.  The chair is claudiub. Information about MeetBot at http://wiki.debian.org/MeetBot.
13:02:17 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
13:02:20 <openstack> The meeting name has been set to 'hyper_v'
13:02:26 <domi007> hi all
13:02:28 <claudiub> hello. :)
13:02:33 <sagar_nikam> Hi
13:02:37 <atuvenie> o/
13:02:39 <abalutoiu> hello
13:02:42 <sonu> Hi everyone
13:02:43 <claudiub> sorry I started a bit late, was pinged about nova-rescue. :)
13:02:48 <itoader> o/
13:02:56 <kvinod> hi
13:03:30 <claudiub> ok, so some status updates
13:03:39 <claudiub> #topic monasca Windows support
13:04:06 <claudiub> still in progress. I've fixed the disk checker, so now disk metrics are good.
13:04:19 <claudiub> there were a couple of fixes to do there. yeay.
13:04:47 <sagar_nikam> claudiub: the monasca BP is approved ?
13:04:55 <claudiub> currently looking at process checker, seems to be working. might be dependent on psutil version.
13:04:56 <lpetrut> Hi guys
13:05:12 <claudiub> not yet, for some reason.
13:05:26 <alexpilotti> hi folks
13:05:36 <sagar_nikam> ok, could you discuss this in the last monasca meeting ?
13:05:40 <claudiub> sagar_nikam: I spoke with tpl, he seemed fine with most changes.
13:05:58 <sagar_nikam> alexpilotti: Hi... welcome back ... to IRC meetings
13:06:11 <claudiub> yep, I've brought this up to the last monasca meeting
13:06:16 <sagar_nikam> claudiub: was it roland ?
13:06:19 <claudiub> yep
13:06:30 <sagar_nikam> let me know if i need to follow up
13:06:38 <sagar_nikam> i can try it
13:06:43 <claudiub> sure. I'll ping him today again.
13:06:45 <sonu> are you collecting vnic metrics too on hyper0v?
13:07:40 <claudiub> vNIC metrics are being collected by the hyperv checker, which is still in progress. still have to cleanup the refactor.
13:07:58 <claudiub> as for host NIC metrics, they are collected by the network checker.
13:08:06 <sonu> Yeah it makes sense
13:08:33 <sagar_nikam> claudiub: i think we need the followin metrics to start off .. disk, network, cpu and memory
13:08:42 <sonu> for vnic, do you put the port-id as one of the dim?
13:08:47 <claudiub> anyways, some reviews would be nice.
13:08:54 <claudiub> #link https://review.openstack.org/#/q/status:open+project:openstack/monasca-agent+branch:master+topic:bp/add-windows-support
13:10:08 <sonu> and you would have to revisit this when we have OVS in place on Hyper-V
13:10:10 <claudiub> sagar_nikam: you mean the VMs metrics?
13:10:38 <sagar_nikam> first host metrics should be fine, we can then move to VMs metrics
13:10:47 <claudiub> sagar_nikam: yeah also on the list will be disk.iops and disk.latency
13:10:51 <sagar_nikam> currently vmware driver for monasca does that
13:11:11 <claudiub> sonu: you mean the vNIC metrics, right?
13:11:28 <sonu> then OVS plugin will be useful - https://review.openstack.org/#/c/306621/
13:11:34 <sonu> yes I meant vnic metrics
13:12:25 <sagar_nikam> claudiub: how about monasca metrics if the host is in MSCluster and we are using nova cluster driver
13:12:38 <sagar_nikam> we give CSV metrics ?
13:12:42 <sagar_nikam> instead of disk
13:13:07 <claudiub> sonu: yeah, I should look at that
13:13:46 <claudiub> sagar_nikam: hm, the disk checker takes all the available partitions, and collects metrics
13:14:23 <claudiub> so, if the CSV is mounted on the host and has a drive letter, its metrics should be collected.
13:14:31 <sagar_nikam> claudiub: ok.. should be fine..
13:15:23 <claudiub> but there's a chance the metrics can't collected. for example, for Floppy Disk Drive and CD / DVD Disk Drive, disk metrics can't be collected.
13:15:28 <claudiub> even with drive letter.
13:16:18 <claudiub> anyways. in my list of to do, is to also look at the iis checker
13:16:35 <claudiub> there's some wmi stuff there. :)
13:17:00 <sagar_nikam> claudiub: IIS checker ? is it for web server or some thing else
13:17:04 <claudiub> yep
13:17:44 <sagar_nikam> ok
13:17:51 <claudiub> #topic nova patches
13:18:11 <claudiub> #link https://etherpad.openstack.org/p/newton-nova-priorities-tracking
13:18:17 <sagar_nikam> i saw one nova patch merged ... nice
13:18:22 <claudiub> so, last week something happened, apparently. :)
13:18:26 <claudiub> not just one. :)
13:18:49 <claudiub> since thursday untill now, we've had 6-7 patches merged.
13:19:07 <claudiub> and we have 2 more patches with a +2.
13:19:37 <claudiub> which is great, it came as a nice surprise. :)
13:20:05 <claudiub> also, apparently, we'll have to add support for PCI passthrough on Hyper-V.
13:20:25 <sagar_nikam> cluster driver not in that etherpad ?
13:20:31 <claudiub> it is
13:20:57 <sagar_nikam> ok found it ...
13:21:05 <claudiub> and we'll have to investigate SRIOV as well.
13:21:09 <claudiub> fun times. :)
13:21:15 <sonu> is PCI-SRIOV a requirement?
13:21:29 <claudiub> sonu: what do you mean?
13:21:46 <sonu> I mean which release is this targeted for?
13:22:03 <claudiub> we're targetting Newton.
13:22:21 <claudiub> at the very least, it will land on compute-hyperv, but hopefully it will also land on nova.
13:22:28 <sonu> Have you started work on this already? or yet to start.
13:22:50 <sonu> we were interested in contributing this feature, if you are fine with it.
13:23:27 <claudiub> sonu: I've started investigating for disk passthrough. I already have an env for that. For the rest, I'll need another env. :)
13:24:20 <claudiub> sure, any help is welcome. :)
13:24:36 <sonu> I was aiming for VM NICs as PCI and SRIOV passthrough.
13:24:38 <claudiub> but it will have to done until June 25, if it is to land on nova.
13:25:04 <domi007> so exactly one month from now
13:25:10 <claudiub> sonu: afaik, NIC passthrough is available only on Windows Hyper-V Server 2016.
13:25:15 <claudiub> domi007: yep.
13:25:27 <claudiub> sonu: for the rest, there's only SRIOV
13:26:06 <sonu> so you mean SRIOV has to be done till June 25
13:26:11 <claudiub> yep
13:26:37 <claudiub> anyways, moving on.
13:26:41 <sonu> I shall connect with you offline on this topic thanks.
13:26:41 <claudiub> # os-brick status
13:26:52 <claudiub> sonu: cool. :)
13:27:18 <claudiub> so, there are 2 patches that needs to land in os-brick:
13:27:21 <claudiub> # link https://review.openstack.org/#/c/312999/
13:27:26 <claudiub> this already has a +2, which is nice.
13:28:00 <claudiub> #link https://review.openstack.org/#/c/272522
13:28:15 <claudiub> this one doesn't have any +2s :(
13:28:59 <claudiub> really hope everything gets in asap.
13:29:10 <claudiub> so we can move on with the nova fibre channel patches.
13:29:30 <sagar_nikam> agree
13:30:04 <sagar_nikam> anybody hpe that i need to request for review on os-brick patches ?
13:30:26 <claudiub> hemna. :)
13:30:44 <sagar_nikam> ok
13:31:04 <claudiub> # topic OVS
13:31:16 <claudiub> #topic OVS
13:31:40 <claudiub> so, there's a new blog post that you might like:
13:31:42 <claudiub> #link https://cloudbase.it/open-vswitch-2-5-hyper-v-part-1/
13:32:04 <claudiub> should be interesting
13:32:26 <domi007> :) I've seen this while in progress, good stuff, it works great in our env
13:32:59 <claudiub> as for OVS vif plug driver on nova, we're waiting for os-vif to become a dependency in nova.
13:33:15 <claudiub> domi007: nice, good to know. :)
13:33:32 <domi007> the only thing I wasn't able to figure out is how to run VLAN and VXLAN networks simultenously
13:33:45 <sonu> do you use OVS firewall driver now with this?
13:34:00 <sagar_nikam> claudiub: some months in IRC meeting, we discussed about microsoft certification for OVS, anything happened on that ?
13:34:03 <claudiub> atuvenie: ^
13:34:20 <claudiub> hmm, alexpilotti? ^
13:34:26 <sonu> or we should continue using WMI (or MI rather) driver
13:34:47 <domi007> atuvenie told me she is working on trying to make the secgroups MI driver work on Liberty
13:34:52 <domi007> on Mitaka I heard it works thanks to sonu
13:35:15 <claudiub> yeah, I think its because of the enhanced rpc blueprint. :)
13:35:36 <sonu> yeah
13:35:37 <claudiub> sonu: thanks for the blueprint. :)
13:35:39 <domi007> it doesn't work for us on Liberty, getting OVS running needed some cherry-picking as well - mainly because the windows OVS driver wasn't complete and working in Liberty
13:35:42 <alexpilotti> sorry folks, I'm jumping through meetings
13:35:59 <alexpilotti> sagar_nikam: we started that process a few days ago
13:36:22 <sagar_nikam> alexpilotti: thanks.. let us know how it goes
13:37:26 <alexpilotti> sagar_nikam: sure, it will take some time to process, I'll keep you guys updated of course!
13:38:07 <claudiub> domi007: as for the vlan and vxlan networks, I haven't tried it yet.
13:38:28 <claudiub> domi007: simultanously.
13:38:30 <sagar_nikam> claudiub: any plans for DVR ?
13:38:32 <domi007> claudiub: they supposed to work, I'll try to adapt settings from KVM hosts to see if it works
13:38:52 <sonu> DVR is heavy lifting for Hyper-V
13:39:25 <claudiub> domi007: sure, let us know how it goes. :)
13:39:33 <sonu> DVR works on namespaces on Linux compute. We must evolve something equivalent on Windows.
13:40:24 <claudiub> sonu: getting the neutron l3-agent to work properly on Windows is gonig to be tricky. :)
13:41:06 <sonu> yes, But we should have this in roadmap, since sooner we will need it
13:42:20 <claudiub> indeed. we'll look at it once we finish all our work in progress blueprints.
13:42:52 <sonu> claudiub: sure
13:42:55 <claudiub> any other topic you guys want to discuss?
13:43:00 <sonu> yes
13:43:09 <sonu> one question
13:43:19 <claudiub> sure
13:43:33 <sonu> From the source, I see that we handle multiple physnet scenario in Hyper-V.
13:43:47 <claudiub> #topic open discussion
13:43:47 <sonu> meaning physnet:br1, physnet2:b2
13:44:03 <sonu> so all VMs on physnet1 goes over uplink on br1
13:44:19 <sonu> and on physnet2 goes over uplink on br2
13:44:31 <sonu> this is a classic use case in Linux KVM..
13:44:58 <claudiub> ok
13:45:06 <sonu> and using our bridge mapping on hyperv neutron.conf, we can set the physnet and bridge mapping
13:45:17 <claudiub> yep
13:45:28 <sonu> and Ia m hoping that this is a supported scenario to test. correct?
13:45:41 <claudiub> yep
13:45:56 <sonu> we were finding some issues. So wanted blessing to continue our debugging :)
13:46:11 <claudiub> what issues exactly?
13:46:23 <claudiub> I assume it's networking-hyperv, right?
13:46:36 <sonu> Vms on one physnet cannot reach out.
13:46:42 <sonu> yes it is networking-hyperv.
13:46:53 <sonu> We will investigate and root cause the problem tomorrow.
13:47:13 <domi007> I have two little questions as well: when adding a second SMB backend to cinder the following error message can be seen in the log: http://paste.openstack.org/show/4FZfv4AbYIWL9nrmo2zH/ probably originating from this call https://github.com/openstack/cinder/blob/stable/liberty/cinder/cmd/volume.py#L86
13:47:28 <domi007> the second is: have you heard anything from c64cosmin about freerdp?
13:47:31 <claudiub> sonu: do those VMs get an IP?
13:47:37 <sonu> no they don't
13:47:43 <claudiub> c64cosmin: hi. :)
13:47:50 <c64cosmin> hi guys
13:48:06 <domi007> hello :)
13:48:22 <c64cosmin> I'm just waiting to get my PR merged :)
13:48:44 <claudiub> domi007: that's a very weird error
13:48:50 <sagar_nikam> c64cosmin: what is the fix for ?
13:49:03 <c64cosmin> several fixes
13:49:10 <domi007> claudiub: I agree, it should work just fine with any number of backends I guess
13:49:39 <sagar_nikam> c64cosmin: once they are merged, can we have a MSI in stable branch, our QA can pick and test
13:50:05 <c64cosmin> sagar_nikam: totally
13:50:13 <claudiub> domi007: yeah, I would assume so. what backends did you configure?
13:50:33 <domi007> claudiub: I have the config file here, got it from a colleague, could it be caused because of having the same mount point base? http://paste.openstack.org/show/0Hle4B0gZIK0nBZiDDbY/
13:50:38 <sagar_nikam> c64cosmin:let me know when the MSI is ready, we pick it up
13:50:47 <domi007> sagar_nikam, c64cosmin +1
13:51:42 <sagar_nikam> c64cosmin: if deb is available, we can pick and test that as well... i know that is in plans... we can try both MSI and deb
13:52:17 <c64cosmin> sagar_nikam: in your case, the deb might be useful
13:52:22 <sagar_nikam> domi007: how is your tests going with freerdp behind haproxy ?
13:52:27 <claudiub> domi007: it might be possible, actually.
13:52:52 <sagar_nikam> c64cosmin: yes eagerly waiting for deb... can test as soon as it is available
13:53:03 <domi007> sagar_nikam: I decided to wait for the new MSI, currently busy with networking and this SMB issue
13:53:09 <claudiub> domi007: do you have a full log by any chance?
13:53:25 <c64cosmin> sagar_nikam: as I said last week, you and domi007 will be anounced directly when available :)
13:53:39 <domi007> claudiub: I'll try to create one and get it to you, thanks
13:53:40 <claudiub> wondering if lpetrut knows more about this.
13:53:47 <sagar_nikam> domi007: i think you should try deb and haproxy instead of MSI and haproxy
13:53:48 <domi007> c64cosmin: really appreciate it
13:53:58 <claudiub> domi007: sure, I'll let him know.
13:54:01 <sagar_nikam> c64cosmin: thanks
13:54:14 <domi007> sagar_nikam: it doesn't make a difference, I can try the deb as well, we have Ubuntu machines running as controller nodes
13:54:28 <c64cosmin> thank you
13:55:20 <claudiub> sonu: still here?
13:55:25 <sagar_nikam> alexpilloti: claudiub: how are the tests going with WIN 2016
13:55:42 <sagar_nikam> i believe we cant run MSI on it
13:55:56 <claudiub> sagar_nikam: really? why?
13:56:21 <sagar_nikam> i remember reading it somewhere on the net that it will not be supported
13:56:24 <claudiub> sagar_nikam: we had a lot of tempest test runs on win 2016 until now.
13:56:33 <claudiub> for each cycle and release.
13:56:44 <sagar_nikam> cluadiub: installed on WIN2016 using MSI ?
13:56:48 <claudiub> yep
13:56:53 <sagar_nikam> ok.
13:57:20 <claudiub> what exactly is the issue?
13:57:25 <claudiub> is there any error, or?
13:57:35 <claudiub> can you provide the installer logs?
13:57:45 <sagar_nikam> claudiub: not tried it.. just read it
13:58:14 <sagar_nikam> claudiub: can you provide me the location from where i can download the latest TP of WIN2016
13:58:41 <claudiub> https://www.microsoft.com/en-us/evalcenter/evaluate-windows-server-technical-preview
13:58:52 <sagar_nikam> thanks
13:58:57 <claudiub> np. :)
13:59:13 <domi007> claudiub: maybe this: "Nano Server does not include MSI as an installation technology due to dependencies" https://blogs.technet.microsoft.com/windowsserver/2015/11/16/moving-to-nano-server-the-new-deployment-option-in-windows-server-2016/
13:59:26 <sagar_nikam> yes
13:59:30 <sagar_nikam> i meant nano server
13:59:35 <sagar_nikam> of WIN 2106
13:59:46 <sagar_nikam> sorry i should have been more clear
14:00:03 <sagar_nikam> have you tried tests on nano ?
14:00:04 <domi007> np my Googleing was fruitful anyway :)
14:00:27 <sagar_nikam> how does it work ?
14:00:31 <claudiub> domi007: oh, i see what you mean. for nano, you cannot use a msi
14:00:53 <sagar_nikam> claudiub: yes, we cant use MSI
14:01:48 <claudiub> anyways, we have to end the meeting. :)
14:01:56 <sagar_nikam> thank you all
14:02:03 <claudiub> thanks for joining. :)
14:02:06 <claudiub> #endmeeting