15:00:47 <bswartz> #startmeeting manila
15:00:47 <openstack> Meeting started Thu Nov 13 15:00:47 2014 UTC and is due to finish in 60 minutes.  The chair is bswartz. Information about MeetBot at http://wiki.debian.org/MeetBot.
15:00:48 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
15:00:50 <openstack> The meeting name has been set to 'manila'
15:01:14 <bswartz> hello all
15:01:28 <jasonsb> hi
15:01:32 <xyang1> hi
15:01:36 <csaba> hi
15:01:40 <bswartz> #agenda https://wiki.openstack.org/wiki/Manila/Meetings
15:01:48 <echoingu_> hi
15:01:56 <bswartz> looks like a small group today
15:02:12 <bswartz> maybe people in the US forgot about the time change? ;-)
15:02:20 <jasonsb> yup
15:02:35 <xyang1> ya, I forgot about it yesterday
15:02:43 <jasonsb> alot were canceled this week too
15:02:45 <bswartz> anyways, the agenda is pretty small today
15:03:19 <bswartz> well we could cancel, but I'd rather have a small meeting in case anyone wants to raise an issue or has questions about last week
15:03:29 <bswartz> I know some people weren't able to attend paris in person
15:03:39 <bswartz> #topic paris review
15:04:25 <bswartz> so anyone who missed the design summit sessions, there should be notes added to the etherpads
15:04:29 <bswartz> https://wiki.openstack.org/wiki/Summit/Kilo/Etherpads#Manila
15:04:49 <bswartz> #link https://etherpad.openstack.org/p/kilo-manila-meetup
15:05:11 <bswartz> Personally, I was very happy with how the design summit went
15:05:23 <bswartz> thanks to those who came to the sessions/meetup and contributed
15:05:37 <bswartz> We're already starting to work on the network changes that we talked about
15:06:21 <bswartz> if there are no questions about that
15:06:34 <jasonsb> i have a small question about netowrking
15:06:58 <bswartz> go ahead
15:07:08 <bswartz> vponomaryov doesn't seem to be here for the next topic
15:07:19 <jasonsb> question is about present and new networking
15:08:16 <bswartz> yes?
15:08:30 <jasonsb> i am trying to understand which network configs don't put pressure on east-west routing
15:08:54 <jasonsb> or, if there is pressure, you would use DVR to fix
15:09:14 <bswartz> maybe I'm not enough of a network buff to know what those terms mean
15:09:16 <jasonsb> in my case i think i a single share network might be easisest
15:09:19 <bswartz> east-west routing?
15:09:32 <bswartz> DVR?
15:09:33 <jasonsb> layer 3 between tenant networks
15:09:50 <jasonsb> where the traffic goes through a single neutron l3 agent
15:10:13 <jasonsb> DVR is the new distributed virtual router stuff in juno
15:10:21 <bswartz> so the new networking options that we're creating both rely on the ability of the tenants to route to the share server through their "public" network
15:10:53 <bswartz> in both the single_svm driver mode and the flat_multi_svm driver mode, Manila doesn't manipulate routing tables at all
15:11:03 <jasonsb> nod
15:11:25 <bswartz> so we rely on routing being setup in advance (usually statically)
15:11:58 <bswartz> basically it's a way for us to punt on hard problems, so the administrator can do what he wants
15:12:16 <bswartz> OR so the administrator can adapt to existing configurations that he can't change
15:13:00 <csaba> bswartz: cold
15:13:02 <jasonsb> how about segmented multi SVM
15:13:18 <csaba> sorry, typo
15:13:37 <bswartz> the managed_multi_svm (previously segmented) does what we do today, where we create interfaces on private networks and sometimes create whole new networks and routers
15:14:25 <bswartz> it includes the possbility to put share servers in the same L2 subnet as the tenants that use the shares
15:14:56 <jasonsb> do any drivers use that besides netapp?
15:15:07 <bswartz> jasonb: the generic driver can do that
15:15:25 <bswartz> however I've come to feel over the last several months that ^ is a bad idea
15:15:50 <bswartz> because rarely do you have the case where the share server is *phsyically* close to the VMs that are using it
15:16:34 <jasonsb> in my case my share server is physically close to VM's, but i agree it seems a little bit not good idea
15:16:38 <bswartz> and if they're physically far apart, then extending the L2 broadcast domain across the network creates a lot of unneccesary traffic on top of rack switches
15:17:21 <bswartz> but my hope is that we can preserve both modes of operation and let the admin choose
15:17:59 <bswartz> for a sensible default I think I would choose separate networks with routers in between the client and server
15:18:09 <bswartz> that answer your qusetion?
15:18:15 <jasonsb> yes indeed thank you
15:18:21 <bswartz> cool
15:18:31 <bswartz> I think we'll skip the next topic
15:18:40 <bswartz> #topic open discussion
15:18:51 <bswartz> anyone have anything else for this week?
15:19:05 <jasonsb> i have 2 topics
15:19:11 <bswartz> ok
15:19:23 <jasonsb> go?
15:19:26 <bswartz> sure
15:19:31 <bswartz> it's a slow day
15:19:33 <jasonsb> topic 1: quick comment
15:19:35 <jasonsb> :)
15:20:01 <jasonsb> it occurred to me that the share image use case you described kind of matches google use case
15:20:20 <bswartz> you mean manila-image?
15:20:23 <bswartz> what's  share image?
15:20:26 <jasonsb> i'll reach out and see if maybe we can borrow some stuff they use for their little debian root
15:20:35 <jasonsb> the devstack gate image
15:20:39 <bswartz> ok
15:20:57 <jasonsb> maybe it has the size you are looking for
15:21:05 <bswartz> do you have a link?
15:21:11 <jasonsb> no
15:21:27 <jasonsb> but marc merlin does it and he has talked about it at conferences before
15:21:32 <jasonsb> i was just going to ask him
15:21:40 <bswartz> okay I'd like to hear more
15:21:49 <bswartz> I'm googling now and not finding anything
15:21:58 <jasonsb> i'm not sure what they call it
15:22:15 <jasonsb> its just enough to bootstrap into omega
15:22:38 <jasonsb> topic 2: interested to hear your thoughts on heat integration sometime
15:23:14 <jasonsb> if you have any strong opinions etc
15:23:42 <bswartz> re: heat, I don't have much details, other than that integration would enable a lot of interesting stuff, and it probably won't happen without one of us driving it
15:24:20 <jasonsb> i know somebody who might be interested
15:24:28 <jasonsb> but i wasn't able to describe it very well
15:24:39 <bswartz> someone at HDS?
15:24:57 <jasonsb> yes
15:25:11 <bswartz> from what little I know, it's just a matter of writing a new plugin to heat that provides a mapping from the manila API to whatever heat understands
15:26:05 <bswartz> there's someone at netapp who will probably get around to doing that work sooner or later
15:26:26 <bswartz> I don't control other people's priorities though :-/
15:26:32 <jasonsb> is it actively being worked on?
15:26:39 <bswartz> not yet and not soon
15:26:57 <bswartz> if you decide to start something please let us know so we can avoid duplicating work
15:27:06 <jasonsb> oh ok
15:27:10 <bswartz> probably send a note to the ML
15:27:12 <jasonsb> will do
15:27:20 <bswartz> ML is about as public as we can get
15:27:27 <bswartz> thanks
15:27:44 <jasonsb> any contacts in heat who might assist?
15:27:51 <bswartz> okay it sounds like nobody has anything else so we'll end this meeting
15:27:53 <jasonsb> or just reach out to them?
15:28:06 <bswartz> jasonsb: grab me in the manila channel
15:28:07 <xyang1> emails to ML can get missed too.  probably good to mention it at IRC meeting too
15:28:26 <bswartz> xyang1: I'll see a ML message and make sure others know about it
15:28:33 <xyang1> ok
15:28:40 <bswartz> ML is archived unlike IRC
15:29:03 <xyang1> the volume is too high.  with [manila] in the subject will help
15:29:10 <bswartz> see you all next week, when you're all no jetlagged anymore
15:29:17 <xyang1> bye
15:29:19 <jasonsb> sounds good
15:29:22 <bswartz> oh yea, you MUST out [manila] in your ML posts
15:29:24 <jasonsb> bye
15:29:28 <rushil> bye
15:29:29 <bswartz> otherwise I don't read it ;-)
15:29:37 <bswartz> must *put [manila]
15:29:41 <bswartz> #endmeeting