20:01:13 <sdake> #startmeeting kolla
20:01:14 <openstack> Meeting started Mon Apr 13 20:01:13 2015 UTC and is due to finish in 60 minutes.  The chair is sdake. Information about MeetBot at http://wiki.debian.org/MeetBot.
20:01:15 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
20:01:17 <openstack> The meeting name has been set to 'kolla'
20:01:21 <sdake> #topic rollcall
20:01:23 <sdake> hey folks :)
20:01:29 <rhallisey> hello
20:01:32 <daneyon> hi
20:01:55 <sdake> jpeeler slower samyaple around?
20:01:58 <Slower> yea
20:02:00 <Slower> o/
20:02:08 <jpeeler> hi
20:02:35 <sdake> #topic kilo 4 planning
20:03:06 <sdake> for kilo 4 (April 25th), our goals are all the basic services + ha
20:03:17 <sdake> I'm unclear if we will actually be able to deliver all that :)
20:03:36 <sdake> lets talk about the basic services first
20:03:36 <daneyon> i think it will be a strech
20:03:37 <sdake> neutron
20:03:53 <sdake> daneyon can you give us  a status update
20:04:08 <daneyon> neutron server and agents have been merged
20:04:14 <daneyon> agents run in a fat container
20:04:32 <daneyon> fat = multiple agents in a single container being managed by supervidord
20:04:45 <sdake> the issue with thin containers was name spaces needed sharing?
20:05:09 <daneyon> you need to have multiple NIC's on your vm/machine to run neutron
20:05:35 <daneyon> by default, the agents container will use eth1 as the physcial network to bridge to
20:05:35 <sdake> you need that with nova networking
20:05:58 <daneyon> this is the neutron flat network used to assign floating_ip's to instances
20:06:41 <daneyon> even without using eth1, you should be able to start neutron-server container, neutron agents-container and spin-up vm's, attach to the linux bridge, get a dhcp address, etc..
20:07:06 <sdake> cool nice work
20:07:13 <sdake> I want to understand the thin contaniers issue
20:07:21 <sdake> is it that they need to share namespaces?
20:07:57 <daneyon> sdake that is correct, however nova-networking uses the flat interface not for floating-ip's, but to have instances communicate on the tenant network acorss compute nodes. Since nova-net use the flat interface for a different purpose, I created a diff param for the neutron flat interface
20:08:44 <sdake> which containers need to share namespaces?
20:09:11 <daneyon> you can set neutron l3/dhcp agents to use net namespaces. when netns is true, each agent with create a netns for every router and tenant network you create... this allows for overlapping ip's and such
20:09:51 <sdake> so l3 and dhcp need to share their network namespaces?
20:10:19 <daneyon> to further complicate things, the metadata agent creates a metadata-agent-prxy in the dhcp namespace
20:10:25 <daneyon> they have their own
20:10:51 <sdake> when samyaple is around could you have a dsicussion around which needs sharing and which doesn't
20:10:57 <sdake> he is of the opinion nothing needs sharing
20:11:09 <sdake> his objection is that he can't attach tcpdump to the netns of a vm
20:11:22 <daneyon> the dhcp ns is shared with the metadata-agent-proxy. the metadata-agent creates a metadata-agent-proxy per dhcp ns.
20:11:24 <sdake> other then that, he says thin containers work
20:11:30 <daneyon> sure
20:11:41 <sdake> so dhcp and metadata-agent-proxy must share a container
20:11:49 <sdake> another option is --net=container
20:11:55 <daneyon> yes
20:12:04 <sdake> this works like a data container
20:12:14 <sdake> but seems hacky :)
20:12:28 <sdake> ok cool thanks
20:12:38 <sdake> rhallisey cinder progress?
20:12:45 <daneyon> i have only tested all in one so far
20:12:59 <rhallisey> I have iscsi running in a container
20:13:10 <daneyon> next step for me will be to run nova-compute with the neutron agents on a seperate node from the controller containers
20:13:11 <rhallisey> which was a decent hurdle to get over
20:13:17 <rhallisey> the api works
20:13:22 <sdake> ncie
20:13:27 <rhallisey> have issues with cinder-volume
20:13:34 <sdake> which sort of issues
20:13:37 <rhallisey> I think it's a config thing
20:13:56 <rhallisey> I need to create the cinder-volume group
20:14:05 <rhallisey> or something.. I forgot what the error was
20:14:07 <sdake> ya lvm access
20:14:23 <sdake> you can bindmount /dev/lvm
20:14:36 <sdake> or it shoudl come in i think automatically
20:14:40 <rhallisey> ok
20:14:47 <sdake> docker 1.6.0 iirc can bindmount /dev correctly, dwalsh knows more
20:14:51 <rhallisey> actually I might be doing that
20:15:12 <rhallisey> ya I am mounting /dev/lvm though I don't think I'm using 1.6
20:15:14 <sdake> is it a new config option?
20:15:21 <rhallisey> so that could be the issue
20:15:49 <sdake> so that leaves swift and ceilometer
20:16:15 <sdake> Ive been working on a magnum blueprint, but we recently declared it not feasible
20:16:18 <sdake> (this morning)
20:16:24 <sdake> so I'm back to kolla full time until end of cycle
20:16:38 <sdake> I am working on making local registry work well
20:16:46 <sdake> so people don't have to spend hours uploading/downloding images
20:17:04 <sdake> a dl of the full iamge set takes me 2 minutes 30 seonds
20:17:04 <sdake> (tools/pull)
20:17:39 <sdake> next I'll take charle's work with ceilometer nand get that going
20:17:52 <sdake> that will probably take to end of cycle to  get reliable
20:18:15 <sdake> I think for neutron, there is some remaining work
20:18:39 <sdake> like tools/start/stop/pull and some kind of conf-neutronnet script like my conf-novanet script
20:18:46 <daneyon> it would be good to get people to test the currect neutron code
20:18:53 <sdake> daneyon I am testing
20:18:54 <daneyon> current
20:19:08 <sdake> all day and all tomorrow ;)
20:19:19 <sdake> I'll help pick out the holes in the imp :)
20:19:36 <sdake> that leaves HA and swift
20:19:47 <rhallisey> got nothing for HA thus far
20:19:51 <rhallisey> still messing with cinder
20:19:57 <sdake> ya finish the job
20:20:19 <sdake> rhallisey if you can get some of your container work up in reviews that would be good
20:20:22 <sdake> the ones that work :)
20:20:37 <rhallisey> I have a WIP for cinder up
20:20:46 <rhallisey> I'm just updating that on the fly
20:20:46 <daneyon> sdake i'll start creating the ha spec later tonight or tomorrow
20:20:51 <rhallisey> anyone can comment on that
20:20:53 <sdake> sounds good
20:21:04 <sdake> so lets try to get the spec wrapped up before the 25th
20:21:19 <daneyon> OK
20:21:23 <sdake> lets make sure its solid
20:21:27 <sdake> so lets take our time
20:21:31 <sdake> and we can crank it out in l1
20:22:04 <sdake> we have everything covered but swift
20:22:05 <sdake> red headed step child i guess ;-)
20:22:21 <sdake> so sounds like k4 is:
20:22:52 <sdake> neutron, ceilometer, cinder, support for local registry
20:23:05 <sdake> with all tools/* working with neutron and cinder and ceilometer
20:23:16 <sdake> sounds good?
20:23:26 <daneyon> sounds good
20:23:55 <sdake> #topic open discussion
20:24:18 <sdake> so, the kolla talk is #6 on the popuarity contest of ods :)
20:24:34 <rhallisey> cool
20:24:59 <rhallisey> Slower and I will be there
20:25:05 <sdake> jpeeler?
20:25:24 <jpeeler> i won't be :/
20:25:58 <sdake> bummer
20:26:06 <sdake> well rhallisey slower make sure to hit the room early
20:27:39 <sdake> if folks could jam through the review queue today that would be great
20:27:45 <sdake> anyone else have anything to discuss?
20:27:57 <Slower> sdake: ok
20:28:00 <rhallisey> nada
20:28:20 <sdake> ok thanks for coming
20:28:22 <sdake> #endmeeting