13:00:17 <claudiub> #startmeeting hyper-v
13:00:18 <openstack> Meeting started Wed May 11 13:00:17 2016 UTC and is due to finish in 60 minutes.  The chair is claudiub. Information about MeetBot at http://wiki.debian.org/MeetBot.
13:00:20 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
13:00:22 <openstack> The meeting name has been set to 'hyper_v'
13:00:33 <claudiub> hellou OpenStack people. :)
13:00:40 <abalutoiu> hello
13:00:44 <sagar_nikam> Hello
13:00:54 <c64cosmin> Hello all
13:01:15 <atuvenie_> Hi all
13:01:24 <atuvenie_> o/
13:01:49 <claudiub> Soo.. let's try to keep this meeting short. :)
13:02:02 <claudiub> #topic Newton workitems
13:02:51 <claudiub> So, as you well know, we will bring the Windows / Hyper-V world to different projects in the OpenStack ecosystem.
13:03:02 <claudiub> e.g. freezer, monasca, magnum
13:03:21 <claudiub> and designate
13:03:24 <sagar_nikam> claudiub:nice
13:03:56 <claudiub> for monasca I've already started, the 1st patch should be fine, but I still have to create a proper environment and test it.
13:04:14 <sagar_nikam> the windows agent for monasca does not work
13:04:21 <sagar_nikam> could you get it working ?
13:04:28 <sagar_nikam> when you submitted the first patch ?
13:04:47 <sagar_nikam> i mean start agent does not work
13:04:50 <claudiub> sagar_nikam: yep, that's another thing I'll have to do.
13:05:17 <claudiub> as I've mentioned, I don't have a monasca environment yet, so I'll be able to make the agent work when I get my env ready. :)
13:05:25 <sagar_nikam> ok
13:05:37 <claudiub> sagar_nikam: the first patch I've sent right after the summit.
13:05:48 <claudiub> there is also another wmi_checker module in the monasca-agent project that I'll have to take a look into
13:06:05 <sagar_nikam> ok
13:06:27 <sagar_nikam> claudiub: can you please give me the link of that patch
13:06:54 <claudiub> also, I've looked around in the code. they seem to have a module which acts very similarly with futures.thread_pool, that we use in networking_hyperv as well.
13:07:18 <claudiub> I'll probably try to replace that module with futures.thread_pool.
13:07:25 <claudiub> #link https://review.openstack.org/#/c/311925/
13:08:02 <claudiub> basically, this patch is almost identical to the patch I've made in ceilometer, when we wanted to replace the in-tree utils with os-win.
13:08:23 <sagar_nikam> ok
13:08:43 <sagar_nikam> shall i add the required team members of monasca for review ?
13:08:50 <claudiub> anyways. another thing that seemed interesting to me, was the monasca-logging agent. I'll take a look at is as well, see that it will work on WIndows as well.
13:08:51 <sagar_nikam> is it ready to be reviewed ?
13:09:39 <claudiub> sagar_nikam: I've marked it as WIP for the time being. I'll remove it after I see everything working properly. :)
13:09:50 <sagar_nikam> sure
13:10:13 <Sonu> I would like to review. Link?
13:10:29 <sagar_nikam> https://review.openstack.org/#/c/311925/
13:10:35 <Sonu> Thanks.I would like to review the metric lists that we can support.
13:10:38 <claudiub> also, for monasca-logging agent, it might be interesting for it to also fetch Hyper-V events. Will investigate further.
13:11:02 <sagar_nikam> that would be nice to send HyperV events
13:11:18 <sagar_nikam> i think we should not send all events
13:11:26 <sagar_nikam> may be just the relevant ones
13:11:29 <claudiub> Sonu: sagar_nikam: just a heads-up, the monasca inspectors and the ceilometer inspectors are the same, or at least the HyperVInspectors are.
13:11:58 <sagar_nikam> ceilometer was for metering and monasca for monitoring
13:12:15 <sagar_nikam> just wondering how they are same, can you please provide more details
13:12:29 <claudiub> sagar_nikam: yeah, from what I understood, monasca-polling is also able to separate logging / events per tenant, which could be trickier to pull off with hyper-v events.
13:13:44 <claudiub> sagar_nikam: sure. there are a few differences between them. first of all, monasca boasts a much better response time when doing queries
13:14:27 <claudiub> since ceilometer uses mongodb, after a few thousands or more metrics, the response times can get a bit crazy.
13:14:53 <sagar_nikam> ok
13:15:23 <claudiub> monasca uses another database type, I'm forgetting which one at the moment, but it works better with large ammounts of data
13:15:27 <sagar_nikam> my question was on how the same/similar code in hyperv can be used in both monasca and ceilometer
13:15:29 <Sonu> influxdb
13:15:39 <claudiub> Sonu: yep, that's the one.
13:15:52 <Sonu> they have vertica integration as well...but for enterprise
13:16:15 <claudiub> anyways, it can handle a lot more data, which is very important, since monasca polls metrics every 30 seconds by default, while ceilometer does it every 600 seconds.
13:17:14 <claudiub> alternatively, we can use the ceilometer-polling agent with the monasca backend. this the so-called ceilasca. :)
13:18:03 <sagar_nikam> claudiub: so which approach is planned for hyperv ? monasca or ceilasca ?
13:18:09 <claudiub> + monasca-agent seems to be polling / monitoring other metrics as well. (e.g. services status)
13:19:14 <claudiub> sagar_nikam: ceilometer-polling is already working on hyper-v, so ceilasca should already work. but we should also have monasca-agent working as well.
13:19:27 <claudiub> so, users can choose their favorite toys. :)
13:19:55 <sagar_nikam> claudiub: agree... i think the first priority should be monasca agent
13:20:30 <claudiub> anyways. I'll have my env ready by next week hopefully. won't take long to have everything working after that.
13:20:55 <sagar_nikam> claudiub:as of today does ceilsca work ? for hyperv
13:21:11 <sagar_nikam> since we already have ceilometer agent for hyperv
13:21:24 <claudiub> plus, the monasca core are pretty cool with deadlines and deadlines, which is neat. :)
13:21:54 <claudiub> sagar_nikam: I've mentioned before: it should work, but I haven't tested it yet. will try it.
13:22:11 <sagar_nikam> ok
13:22:42 <claudiub> ok.. so... for other projects, abalutoiu has been designated to lead the effort for the project named designate. :)
13:23:12 <Shivakumar> Hi, with native thread patch applied, hyperv agent hangs sometimes.
13:23:23 <claudiub> lpetrut is freezing with enthusiasm over the prospect of working on freezer. :)
13:24:13 <claudiub> Shivakumar: ok, will tackle that as soon as we change the topic.
13:24:35 <sagar_nikam> claudiub: nice to see some progress on monasca, designate and freezer
13:25:14 <claudiub> ok, so I guess that's it for this topic.
13:25:25 <claudiub> #topic networking-hyperv
13:26:32 <claudiub> ok, so, first of all, just a small announcement, a little while ago, networking-hyperv has been removed from neutron's governance, as it was calling the Hyper-V APIs, which is not "open-source".
13:27:25 <claudiub> this happened because the neutron core decided that they can't vouch for some networking agents / plugins in their governance, if not everything is opensource.
13:27:49 <claudiub> so, networking-hyperv is not the only one in this situation, there were quite a few that were removed from governance because of this.
13:28:04 <claudiub> buut, since today, networking-hyperv is under winstackers governance, so yeay. :)
13:28:29 <claudiub> that means we have a little bit more liberty and control with the project.
13:29:09 <claudiub> Shivakumar: ok, so you're saying you have issues with native threads? what branch? are you using pymi?
13:29:15 <claudiub> what os-win version?
13:29:55 <Shivakumar> actually, we are using pymi with stable liberty hyperv code
13:29:59 <claudiub> because this hasn't happened in the Hyper-V CI since it was originally merged in Mitaka, like 2-3 months ago I think.
13:30:32 <Shivakumar> we backported native thread and enhanced rpc patches, able to get good scale results upto 1000VMs
13:31:17 <sagar_nikam> claudiub: we hit 2 issues in scale tests, will mention once we are done with the networking topic
13:31:28 <claudiub> Shivakumar: ah, I see. I suppose that it hangs during agent's startup?
13:31:40 <Shivakumar> but, sometimes hyperv agent hangs while making an rpc call, and fails to report again
13:32:23 <Shivakumar> since we have introduced threading for processing new ports, is it due to Logger  ?
13:33:13 <Shivakumar> yes, cladiub ..during agent startup
13:33:49 <Shivakumar> when there are no VM ports to be processed, but if an update comes from other HyperV hypervisor
13:33:58 <claudiub> Shivakumar: yeah, it feels like you're missing something.
13:34:22 <claudiub> Shivakumar: I think that the main thread is locked. I don't know exactly what you've backported.
13:35:21 <Shivakumar> i have backported native thread and enhanced rpc only
13:35:23 <claudiub> Shivakumar: I suppose you've also backported the patch with the vNIC event listeners, right?
13:35:41 <Shivakumar> no, we left that
13:36:04 <Shivakumar> have a doubt
13:36:15 <Shivakumar> do we need to add this content https://github.com/openstack/networking-hyperv/blob/master/hyperv/neutron/__init__.py
13:36:31 <Shivakumar> means, add  monkey patch ?
13:37:00 <claudiub> honestly, some logs could be useful, as well as a list of cherry-pick you've done. otherwise I'm just doing guess work.
13:37:14 <Shivakumar> https://review.openstack.org/#/c/263865/
13:37:38 <Shivakumar> https://review.openstack.org/#/c/264235/
13:38:31 <claudiub> Shivakumar: hm, in liberty, the neutron-hyperv-agent was still being loaded from neutron
13:38:46 <Shivakumar> ok,
13:38:52 <claudiub> Shivakumar: so, monkey patching was already being done there.
13:39:24 <Shivakumar> i think here also we have a separate repo for hyperv
13:40:41 <Shivakumar> so, if hyperv is separate, does it mean we need monkey patching ?
13:40:58 <Shivakumar> separately ?
13:41:00 <claudiub> Shivakumar: what do you mean?
13:41:35 <claudiub> Shivakumar: also, logs pls. :)
13:41:46 <Shivakumar> sure,
13:41:52 <Shivakumar> 2016-05-06 18:16:23.986 1780 INFO neutron.agent.securitygroups_rpc [req-6f83c39f-7370-43bb-af2b-3c3c0d9f2ff1 - - - - -] Provider rule updated 2016-05-06 18:16:23.986 1780 INFO neutron.agent.securitygroups_rpc [req-6f83c39f-7370-43bb-af2b-3c3c0d9f2ff1 - - - - -] Refresh firewall rules 2016-05-06 18:16:23.986 1780 DEBUG neutron.api.rpc.handlers.securitygroups_rpc [req-6f83c39f-7370-43bb-af2b-3c3c0d9f2ff1 - - - - -] Get
13:42:07 <claudiub> put them here: http://paste.openstack.org/
13:42:22 <claudiub> and post a link.
13:42:39 <Shivakumar> http://paste.openstack.org/show/496733/
13:43:16 <claudiub> can you put the entire log?
13:43:28 <claudiub> or at least from when the agent starts?
13:46:51 <claudiub> Shivakumar: also, there's another patch that could be useful: https://review.openstack.org/#/c/297773/
13:47:52 <claudiub> anyways... since we have a few other things to discuss...
13:47:57 <claudiub> #topic open discussion
13:48:17 <sagar_nikam> claudiub: 2 bugs we found in our scale tests
13:48:25 <Shivakumar> i have pasted complete log http://paste.openstack.org/show/496736/
13:48:26 <sagar_nikam> https://bugs.launchpad.net/nova/+bug/1473291
13:48:27 <openstack> Launchpad bug 1473291 in OpenStack Compute (nova) "nova compute on hyperv don't wait for vif plugged event from neutron" [Medium,Confirmed]
13:48:33 <sagar_nikam> https://bugs.launchpad.net/nova/+bug/1580161
13:48:34 <openstack> Launchpad bug 1580161 in OpenStack Compute (nova) "hyper-v nova compute go down with error Error writing vm console log file from serial console pipe. Error: [Errno 2] No such file or directory" [Undecided,Confirmed]
13:49:01 <sagar_nikam> fix for bug 1580161, lpetrut: already provided
13:49:09 <sagar_nikam> we will use it
13:49:09 <claudiub> sagar_nikam: cool, for the second one, lpetrut already sent up a patch.
13:49:15 <claudiub> #link https://review.openstack.org/#/c/314590/
13:49:20 <sagar_nikam> any plans for the first bug ?
13:49:37 <claudiub> currently the gate is down, so I'll +2 it when it's ok.
13:49:44 <sagar_nikam> we had hit that bug in juno and now in liberty
13:49:59 <claudiub> as for the 1st one, we've already fixed it compute_hyperv in mitaka.
13:50:09 <claudiub> sagar_nikam: it is marked as bp though.
13:50:29 <claudiub> #link https://blueprints.launchpad.net/nova/+spec/hyper-v-spawn-on-neutron-event
13:50:41 <sagar_nikam> claudiub: plans to upstream it in newton ?
13:51:23 <claudiub> sagar_nikam: yup. there are going to be a few changes to it though.. there is some duplicate code related to this sort of mechanism in all drivers. I want to refactor it.
13:51:35 <sagar_nikam> ok
13:52:02 <sagar_nikam> it would be good if it gets merged in newton, we have seen that issue since juno
13:52:11 <claudiub> here's the link to the implementation in compute_hyperv:
13:52:27 <claudiub> #link https://review.openstack.org/#/c/292615/
13:52:59 <sagar_nikam> regarding the first bug for which lpetrut: submitted a patch in stable/liberty
13:53:09 <sagar_nikam> you will +2 it ?
13:53:09 <claudiub> sagar_nikam: this issue doesn't exist anymore in mitaka, since port binding is being done a lot faster
13:53:18 <claudiub> even without this bp.
13:53:45 <claudiub> sagar_nikam: will do as soon as Jenkins gives its +1. Hyper-V CI already passed on the patch, which was needed, so it's good.
13:53:51 <sagar_nikam> claudiub: should i wait for it to be merged in stable/liberty
13:54:21 <sagar_nikam> claudiub: ok thanks
13:54:54 <claudiub> sagar_nikam: yeah, you can.
13:55:16 <sagar_nikam> claudiub: alexpilotti: not in meeting?
13:55:24 <sagar_nikam> had some updates on freerdp
13:55:36 <claudiub> or you can cherry-pick it in your downstream repos, if you have any. :)  but it probably won't take long to be merged.
13:55:53 <claudiub> alex is currently away on some bussiness trip
13:56:02 <claudiub> c64cosmin: ^
13:56:03 <c64cosmin> I'm here for any RDP info though
13:56:07 <sagar_nikam> claudiub: since it will not take long to merge, i will probably wait
13:56:23 <claudiub> c64cosmin: welcome. :)
13:56:45 <sagar_nikam> c64cosmin: sometimes the freerdp service hangs
13:56:52 <claudiub> c64cosmin: ?????????
13:57:08 <sagar_nikam> and restart also hangs... only way to recover is kill wsgate.exe
13:57:40 <c64cosmin> sagar_nikam: that happens with the installed version right?
13:58:03 <sagar_nikam> c64cosmin: yes, installer taken from cloudbase
13:58:52 <c64cosmin> sagar_nikam: we also have a beta version, the fact that the service hangs if well known and had been taken care of
13:59:17 <c64cosmin> though, the beta version is also unstable
13:59:34 <sagar_nikam> c64cosmin: i used the stable version, if the issue is fixed in beta... when will come in stable ?
14:00:34 <c64cosmin> sagar_nikam: cannot approximate for sure, but that is main priority right now on this project
14:00:48 <sagar_nikam> sure... will wait
14:00:57 <sagar_nikam> let me know when you have a good stable build
14:01:56 <sagar_nikam> we are almost out of time... we will discuss in detail in next meeting
14:01:57 <c64cosmin> sagar_nikam: I sure will :)
14:02:25 <claudiub> well... this was a.. short meeting. :)
14:02:31 <claudiub> thanks all for attending. :)
14:02:35 <sagar_nikam> thanks all
14:02:36 <claudiub> see you next time!
14:02:40 <claudiub> #endmeeting