16:03:54 <dougwig> #startmeeting kosmos
16:03:55 <openstack> Meeting started Tue Sep 20 16:03:54 2016 UTC and is due to finish in 60 minutes.  The chair is dougwig. Information about MeetBot at http://wiki.debian.org/MeetBot.
16:03:56 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
16:03:59 <openstack> The meeting name has been set to 'kosmos'
16:04:07 <dougwig> #topic Summit
16:04:16 <dougwig> will anyone here be at barcelona?  should we do a quick meetup?
16:04:38 <RamT> I won't be able to make it
16:05:22 <RamT> is mugsie going to be there?
16:05:32 <RamT> you can probably plan with him
16:05:35 <RamT> are you going?
16:06:28 <dougwig> i'll be there.
16:06:49 <dougwig> i might be alone, talking to myself. but then, what else is new?
16:06:57 <dougwig> #topic Brief status
16:07:05 <RamT> we have made some progress with conductor service
16:07:12 <RamT> db module
16:07:12 <dougwig> i've been consumed by newton/neutron, so nothing from me.
16:07:24 <RamT> integrated alembic migrations
16:07:59 <RamT> we should start submitting reviews once we test the code at our end
16:08:21 <RamT> we have made few changes to the db schema
16:08:31 <RamT> I'll be submitting the specs for that
16:08:47 <RamT> basically created loadbalancer_pools and loadbalancer_parameters
16:09:04 <dougwig> ok, i'll wait for specs on those.
16:09:15 <RamT> please review it and let us know your commends
16:09:20 <RamT> comments*
16:09:49 <RamT> just to confirm once again as per the design
16:09:50 <dougwig> good to see some movement.
16:09:53 <dougwig> #topic Open discussion
16:10:01 <RamT> backend will have two drivers
16:10:06 <RamT> DesignateDriver
16:10:09 <RamT> LBaaSDriver
16:10:24 <RamT> LBaaSDriver is for publishing the pools, poolmembers and monitors
16:10:45 <RamT> DesignateDriver will publish the DNS name for the IP right?
16:10:58 <mugsie> dougwig: I will be there
16:11:25 <RamT> LBaaSDriver is to create the whole LB Policy? Including the VIP? or just pools and monitors?
16:12:22 <dougwig> there is no VIP.  just DNS lookups.
16:12:31 <dougwig> so, just pools and monitors.
16:12:59 <RamT> ok
16:13:16 <RamT> so how would monitors send healthchecks to kosmos?
16:13:22 <RamT> for the pool members?
16:13:46 <mugsie> they should send it over HTTP to the kosmos API
16:13:59 <RamT> or AMQP and conductor?
16:14:06 <mugsie> the healthchecks should be run around the world, by the deployer
16:14:15 <mugsie> definitly not AMQP
16:14:18 <RamT> what is the role of conductor apart from being a db service?
16:14:25 <mugsie> a DB service
16:14:52 <dougwig> for the initial "use lbaas as monitors" implementation, you setup the monitors on lbaas, which will be region specific, and then kosmos will have to poll them.  in parallel, we can see about native monitors or putting call-outs on the lbaas monitors to make them more efficient.
16:14:55 <mugsie> it means that there is much less DB access from other components, which *will* casue issues
16:15:42 <RamT> ok
16:16:10 <RamT> oi
16:16:11 <RamT> ok
16:16:22 <RamT> so it will be pure http/https communication
16:16:33 <RamT> or polling by kosmos it self
16:16:33 <RamT> ok
16:17:07 <RamT> kosmos conductor will be polling or should we have another scheduler for kosmos which will do the polling?
16:18:03 <dougwig> i think the conductor can handle it, at least at first.
16:18:08 <RamT> ok
16:18:27 <RamT> got it
16:19:08 <RamT> I have nothing else.. should be able to move forward
16:19:48 <RamT> any questions for me
16:19:49 <RamT> ?
16:20:38 <RamT> in that case I'll ask few more questions on the functionality and vision of kosmos
16:21:26 <RamT> does anyone need distributed load balancing because there is no packet routing involved with kosmos?
16:21:40 <RamT> or is it just a good to have feature that we are building?
16:22:11 <RamT> I am asking this questions because it's only lookups and we have TTL values associated with DNS records
16:22:29 <RamT> taking those into consideration how can we make the solution work better?
16:24:46 <RamT> anyone there?
16:24:58 <RamT> mugsie? dougwig?
16:26:10 <dougwig> sorry.  traditionally it's low TTL dns for global, and load-balancing is per region.  spreading actual L3 around would have other issues.  kosmos is currently just focused on the DNS angle, though we can discuss others.
16:26:27 <mugsie> RamT: sorry am doing double meetings :)
16:26:43 <dougwig> i have no good reason, i just spaced out. it's early.
16:26:44 <RamT> should be have any upper limit on the TTL?
16:26:53 <mugsie> do distributed l3 is hard :)
16:27:13 <mugsie> RamT: yes - usally approx 5-10 seconds
16:27:18 <RamT> ok
16:27:19 <dougwig> RamT: certainly we should have a low default and low recommendations.
16:27:28 <RamT> ok
16:27:54 <dougwig> anything else today?
16:28:23 <RamT> nopes.. next week I'll have some questions on geographical / proximity GSLB
16:28:28 <dougwig> ok, let's blast out of here. I look forward to the spec and code reviews from RamT.
16:28:39 <mugsie> ++
16:28:40 <RamT> and how we envision to achieve them with kosmos
16:28:54 <RamT> Thank you folks
16:29:00 <RamT> have a great one
16:29:45 <dougwig> #endmeeting