12:59:37 <claudiub> #startmeeting hyper-v
12:59:38 <openstack> Meeting started Wed Aug  3 12:59:37 2016 UTC and is due to finish in 60 minutes.  The chair is claudiub. Information about MeetBot at http://wiki.debian.org/MeetBot.
12:59:39 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
12:59:42 <openstack> The meeting name has been set to 'hyper_v'
12:59:46 <claudiub> hello. :)
13:00:11 <sagar_nikam> hello
13:00:34 <atuvenie_> hey all
13:00:35 <atuvenie_> \o
13:01:01 <claudiub> ok, let's start.
13:01:13 <claudiub> #topic os-brick patch status
13:01:20 <claudiub> ok, so starting with some good news. :)
13:01:45 <claudiub> the microsoft osbrick CI is fully up and running, for iscsi, smb, and fibre channel.
13:01:57 <sagar_nikam> nice....
13:01:58 <claudiub> and the 2 remaining patches on os-brick each have a +2
13:02:25 <claudiub> #link os-brick FC patch: https://review.openstack.org/#/c/323780/
13:02:44 <claudiub> #link os-brick SMB patch: https://review.openstack.org/#/c/323781/
13:03:09 <claudiub> so, all that's left is to ping hemna to review them. already did, but he's not yet online.
13:03:29 <claudiub> questions?
13:03:54 <sagar_nikam> no... just waiting for these patches to be merged along with nova
13:03:56 <abalutoiu> hello
13:04:01 <sagar_nikam> we want to use it
13:04:04 <itoader> hi
13:04:11 <sagar_nikam> in our environment
13:04:22 <claudiub> cool, moving on.
13:04:30 <claudiub> #topic designate patch status
13:04:53 <claudiub> abalutoiu: has been replying to comments on it.
13:05:09 <claudiub> abalutoiu: he's been asked to upload some tempest test results. currently in progress.
13:05:58 <claudiub> other than that, no new news.
13:06:17 <claudiub> #topic nova ovs-vif plug driver
13:06:55 <claudiub> this is the only thing that's not frozen on nova at the moment and it can still merge in newton
13:07:07 <claudiub> so, we'll have to give it some special attention. :)
13:07:42 <sonu> yes. I agree. we need OVS in Newton
13:07:46 <claudiub> waiting for this patch to be rebased / merged, as we'll depend on it.
13:07:59 <claudiub> #link os-vif in nova: https://review.openstack.org/#/c/269672/
13:08:30 <claudiub> until then, we have to merge this in os-vif:
13:08:43 <claudiub> #link os-vif OVSPlugin Windows support: https://review.openstack.org/#/c/344433/
13:09:12 <claudiub> still have to run some more tests for it though.
13:09:28 <claudiub> questions?
13:10:27 <claudiub> the https://review.openstack.org/#/c/269672/ previously had some blockers in oslo.privsep / oslo.rootwrap. they were fixed and merged. still has some comments to address though.
13:11:16 <claudiub> ok, moving on.
13:11:27 <claudiub> #topic hyper-v cluster
13:11:45 <claudiub> #link Hyper-V Cluster spec for Ocata: https://review.openstack.org/#/c/269672/
13:12:00 <claudiub> so, I had to repropose the spec for Ocata.
13:12:11 <sagar_nikam> checking
13:12:12 <abalutoiu> claudiub: wrong link
13:12:22 <claudiub> #undo
13:12:23 <openstack> Removing item from minutes: <ircmeeting.items.Link object at 0x7f5bbffecd50>
13:12:37 <claudiub> #link Hyper-V Cluster spec for Ocata: https://review.openstack.org/#/c/350559/
13:12:41 <claudiub> abalutoiu: thanks. :)
13:13:13 <claudiub> so, we'll have to bring it to a state in which the nova core "feels happy" with it. :)
13:14:00 <claudiub> they'll most probably complain about things, so any help in getting it approved again would be greatly appreciated. :)
13:14:33 <sagar_nikam> claudiub: the spec does not address the concern expressed last time
13:14:56 <sagar_nikam> for the failover handled through nova
13:15:14 <sagar_nikam> we discussed multiple options in a previous IRC
13:15:28 <claudiub> sagar_nikam: there is the "Failover migrations" section in there.
13:15:33 <sagar_nikam> most options we discussed had lot of issues
13:15:50 <sagar_nikam> claudiub: ok checking
13:17:06 <sagar_nikam> option2- it means VMs can be down
13:17:28 <sagar_nikam> with nova triggering it... failover is actually not failover
13:17:52 <sagar_nikam> since the host could have gone down by the time failover is triggered
13:18:31 <claudiub> yep
13:19:45 <sagar_nikam> do we have any other options
13:19:50 <sagar_nikam> can we make it a conf entry
13:20:01 <sagar_nikam> where in the cloud-admin can decide
13:20:16 <sagar_nikam> if auto-failover is required or nova triggered failover
13:20:20 <claudiub> the thing is, they have to agree on this. :)
13:20:42 <claudiub> i'm ok with your suggestion, but I'm not the one approving the spec. :)
13:20:51 <sagar_nikam> conf entry is not a good option ?
13:21:04 <sagar_nikam> agree... we need to check with nova cores
13:21:21 <sagar_nikam> one question
13:21:30 <sagar_nikam> when does mscluster send the failover event
13:21:53 <sagar_nikam> does it send when the host is about to go down or when it is already down
13:22:50 <sonu> IMO with clustering nova live migrate option should not be used
13:23:27 <claudiub> soo, the events are caught when something in the cluster changes (states, objects, hosts, etc.)
13:23:40 <sagar_nikam> sonu: agree... nova live-migrate and cluster failover dont go well with each other
13:23:40 <claudiub> sonu: what do you mean?
13:23:54 <claudiub> they are not the same thing
13:24:03 <claudiub> you can still live-migrate your instances in a cluster.
13:24:10 <sonu> deprecate the nova live migrate option when cluster driver is in use
13:24:19 <sagar_nikam> claudiub: it is like this... we have 4 hosts ... 2 standalone and 2 in cluster
13:24:34 <sagar_nikam> now we can do nova live-migrate between standalone hosts
13:24:49 <sagar_nikam> we cant do between a standalone host and a cluster host
13:25:09 <sagar_nikam> for example the VMs VHDs files will be in CSV for cluster
13:25:21 <sagar_nikam> but for standalone on the filesystem
13:25:29 <sagar_nikam> they dont go well
13:26:53 <claudiub> i see no reason why it wouldn't work. the files will be copied to the csv and it will be fine. if this really was a problem, then live-migration wouldn't have worked at all with standalone compute nodes, since they use their own storage.
13:27:15 <claudiub> and yet, the storage moves to the destination node on live-migration.
13:28:03 <sagar_nikam> so in case of migrate from cluster to standalone
13:28:13 <claudiub> plus, live-migration can be useful to make unclustered instances clustered and vice-versa
13:28:16 <sagar_nikam> they will get copied from CSV to filesystem
13:28:25 <claudiub> yea
13:28:27 <sagar_nikam> i thought for live migration to work
13:28:40 <sagar_nikam> we need the instance path to be same on all computes
13:29:12 <sagar_nikam> is that correct ? instance path same on all computes
13:29:13 <sagar_nikam> ?
13:30:31 <claudiub> the instance path will have to be the same for the clustered compute nodes
13:30:47 <claudiub> or at least, point to the same share
13:31:04 <sagar_nikam> i think for shared nothing live migration to work, the instance path needs to be same,
13:31:10 <sagar_nikam> will check and confirm
13:31:36 <claudiub> ok. :)
13:32:05 <claudiub> ok, anything else?
13:32:35 <sonu> I see challenges when I consider OVS based neutron with clustering
13:32:38 <sagar_nikam> not much as of now for this topic
13:32:47 <claudiub> sonu: why?
13:33:10 <claudiub> as far as I know, atuvenie_ tested the hyper-v cluster driver with ovs
13:33:18 <sonu> security group rules would remain in WMI/MI?
13:33:40 <sonu> we are considering using WMI firewall driver, then fine
13:34:11 <sonu> I was hopeful to get conntrack with 2.6 and use community's OVS firewall
13:34:29 <atuvenie_> sonu: yes, tested cluster with ovs, when a failover is detected the port will be disconnected from the source node ( if it's still up) and added in the bridge at the destination node
13:35:01 <atuvenie_> sonu: sec groups remain with WMI
13:35:05 <sonu> atuvenie_: Thank you.
13:35:44 <claudiub> ok, moving on
13:36:04 <claudiub> #topic Barcelona OpenStack Summit presentations
13:36:33 <claudiub> this is just a reminder: pls vote the Windows related presentations for the next summit. :)
13:37:02 <claudiub> I would add the links here, but they don't work, they'll just redirect to the 1st presentation
13:37:16 <sagar_nikam> sure
13:37:47 <claudiub> just search for: Sonu, Alessandro, Vladu, Samfira, Ciuhandu
13:37:57 <claudiub> note: last week I forgot Ciuhandu. :)
13:38:20 <claudiub> should be 8 presentations. :)
13:38:27 <claudiub> moving on.
13:38:34 <claudiub> #topic open discussion
13:38:55 <claudiub> anything to talk about here?
13:40:00 <sagar_nikam> we have started using Mitaka recently
13:40:08 <sagar_nikam> and everything is working fine
13:40:18 <claudiub> nice. :)
13:40:20 <sagar_nikam> including CHAP support
13:40:30 <sagar_nikam> we are getting it for first time for hyperv
13:40:34 <claudiub> you mean for iscsi?
13:40:39 <sagar_nikam> yes
13:40:46 <claudiub> i think we introduced support for it in Kilo
13:40:59 <sagar_nikam> liberty and earlier releases CHAP did not work well with 3par
13:41:06 <claudiub> oh, i see.
13:41:14 <sagar_nikam> it worked only with windows storage server
13:41:29 <sagar_nikam> now we have tested with 3par and VSA
13:41:34 <sagar_nikam> both work well
13:41:43 <claudiub> sounds great. :)
13:41:50 <sagar_nikam> VSA is another storage solution from HPE
13:42:18 <sagar_nikam> claudiub: how is monasca and freezer going ?
13:42:44 <claudiub> for freezer lpetrut had to post-pone it, as he had to prioritize os-brick ci.
13:42:55 <sagar_nikam> ok
13:43:04 <claudiub> he's been working on the os-win part though.
13:43:14 <claudiub> didn't submit the code yet.
13:43:25 <claudiub> as for monasca, I didn't get any reviews lately.
13:43:45 <claudiub> will have to ping Roland
13:43:52 <sagar_nikam> can you provide the list of patches
13:44:00 <sagar_nikam> i will try to get it reviewed as well
13:44:26 <claudiub> #link monasca windows support patches: https://review.openstack.org/#/q/status:open+project:openstack/monasca-agent+branch:master+topic:bp/add-windows-support
13:44:41 <claudiub> note, some of them are in merge conflict, because the libvirt checker changes often.
13:44:51 <claudiub> so, i always have to rebase that.
13:45:17 <sagar_nikam> ok
13:45:23 <claudiub> sagar_nikam: also, I've been thinking if you folks can test something, if you have some available resources for this
13:45:28 <sagar_nikam> is it correctly rebased now ?
13:45:44 <sagar_nikam> sure... i think we should be able to test it soon
13:45:50 <sagar_nikam> does it work on mitaka ?
13:45:51 <claudiub> sagar_nikam: not at the moment, will have it done by tomorrow, because its merge conflicts are ugly.
13:46:00 <sagar_nikam> or we need newton ?
13:46:17 <sagar_nikam> claudiub: let me know once the patches are rebased
13:46:36 <sagar_nikam> i will try to reach to roland and get it reviewed
13:46:46 <claudiub> I've heard at the midcycle that it *might* be possible to switch from networking-hyperv ports to neutron-ovs ports via live-migration
13:46:52 <sonu> I have to log-off now. Thank you for all the time. Appreciated.
13:46:54 <claudiub> sagar_nikam: can you test this?
13:47:01 <sonu> I have to log-off now. Thank you for all the time. Appreciated.
13:47:04 <claudiub> sonu: thanks for joining. :)
13:47:12 <claudiub> sagar_nikam: cool, thanks. :)
13:47:30 <sagar_nikam> claudiub: test which one ? monasca or neutron ?
13:47:38 <claudiub> sagar_nikam: the neutron thingy.
13:48:04 <sagar_nikam> claudiub: ok .. will check with sonu: his team works on neutron
13:48:20 <sagar_nikam> i think they should be able to pick and check it
13:48:24 <sagar_nikam> patch is important
13:48:24 <claudiub> sagar_nikam: basically, you would have a hyper-v host A with networking-hyperv and host B with neutron-ovs. if you migrate / live-migrate from A to B, you should have an ovs port then.
13:48:40 <sagar_nikam> ok
13:48:51 <sagar_nikam> will ask sonu: tomorrow
13:49:04 <claudiub> but then again, they said "might"
13:49:08 <sagar_nikam> my question on monasca - will it work on mitaka ?
13:49:26 <sagar_nikam> sure will check with them
13:49:35 <claudiub> the monasca work will be on newton, as we're proposing the code to newton. we can't backport a blueprint.
13:49:35 <sagar_nikam> they do all network related work for hyperv
13:49:46 <claudiub> sagar_nikam: cool, thanks. :)
13:50:17 <claudiub> sagar_nikam: also, I was thinking about some rally scenarios.
13:50:33 <sagar_nikam> claudiub: no... what i meant, since we have mitaka only as of now, it is easy for us to checrry-pick these patches to mitaka
13:50:45 <claudiub> sagar_nikam: I am curious to see what are the network throughput difference between networking-hyperv and ovs.
13:50:54 <sagar_nikam> ok
13:51:46 <claudiub> sagar_nikam: ah, i see what you mean. well, you can try cherry-picking them, but the code probably changed quite a lot since mitaka, so you'll have quite a few merge conflicts.
13:51:58 <sagar_nikam> ok
13:52:15 <sagar_nikam> then ... i think we will have to wait till we get into newton
13:52:23 <sagar_nikam> couple of months from now
13:52:56 <claudiub> cool. :)
13:53:00 <claudiub> so, anything else?
13:53:15 <sagar_nikam> how is magnum going ?
13:54:11 <claudiub> atuvenie_: ^
13:54:14 <sagar_nikam> have you seen this ? http://blog.kubernetes.io/2016/07/update-on-kubernetes-for-windows-server-containers.html
13:54:44 <atuvenie_> sagar_nikam: yeah, we've been working with the aprenda folks
13:55:12 <sagar_nikam> atuvenie_: cool.... let us know how it goes
13:55:42 <atuvenie_> sagar_nikam: pretty much at this point most of the work is focusing on kubernetes
13:55:51 <sagar_nikam> in a few months from now, when things become stable for windows and containers, we can start trying it
13:55:58 <atuvenie_> sagar_nikam: after that is done, magnum should be pretty easy
13:56:23 <sagar_nikam> atuvenie_: let us know when we can pick some of your work and test it
13:57:11 <claudiub> yeah, we'll let you know. :)
13:57:18 <claudiub> anything else for today?
13:57:41 <sagar_nikam> no ... thanks
13:57:43 <claudiub> if not, we're going to end the meeting. :)
13:58:06 <claudiub> well then, thanks folks for the updates and for joining, see you next week! :D
13:58:29 <claudiub> #endmeeting