16:03:35 <mugsie> #startmeeting Kosmos
16:03:36 <openstack> Meeting started Tue Aug 18 16:03:35 2015 UTC and is due to finish in 60 minutes.  The chair is mugsie. Information about MeetBot at http://wiki.debian.org/MeetBot.
16:03:37 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
16:03:40 <openstack> The meeting name has been set to 'kosmos'
16:04:03 <mugsie> #topic Polaris Code https://github.com/polaris-gslb/polaris-core
16:04:08 <mugsie> #link https://github.com/polaris-gslb/polaris-core
16:04:19 <mugsie> did anyone get a chance to look at ^?
16:04:49 <KunalGandhi> i looked at briefly and had a few questions
16:05:06 <johnsom> I also had a brief look
16:05:09 <mugsie> (also, as a note - I am at the Designate mid-cycle, so I may get distracted for a few minutes from time to time)
16:05:39 <KunalGandhi> does this code include a monitoring piece for to monitor the regional VIP's ?
16:06:03 <mugsie> I think it looks interesting - I think it might present issues for dynamic reloading of end points to test
16:06:12 <mugsie> KunalGandhi: yup - https://github.com/polaris-gslb/polaris-core/tree/master/polaris/health/core
16:06:27 <mugsie> https://github.com/polaris-gslb/polaris-core/tree/master/polaris/health is the health checker
16:06:41 <KunalGandhi> @mugsie .. got it
16:06:56 <ag__> what kinda of an issue you see for dynamic reloading
16:07:44 <mugsie> it loks like it reads from a config file on boot for the list of endpoints?
16:07:47 <KunalGandhi> seems like it supports http and tcp both, which is good
16:07:58 <mugsie> we need to add in endpoints without any restart
16:08:04 <elarson_> o/
16:08:08 * elarson_ is lurking ;)
16:08:10 <johnsom> I worry about the shared memory usage and HA/horizontal scale.
16:08:45 <mugsie> ag__: what is your feeling about us usign segments of code?
16:09:07 <ag__> yea, it does need a restart to reload endpoints, state is cached though and is unaffected while restart is taking place
16:09:17 <johnsom> Have we looked at the health checking in Consul?
16:09:44 <ag__> mugsie, no special feelings, it's open code
16:09:59 <KunalGandhi> @ag__ .. does the gslb store the monitors, regional vip info in the database ?
16:10:35 <mugsie> KunalGandhi: this one reads from a YAML file on disk
16:10:36 <ag__> i woould like to re-do it to make config changes dynamicm it'll need some redesign though
16:11:01 <mugsie> ag__: what licence is this code again?
16:11:13 <ag__> well, it's yaml serialized, can be any key/value storage, e.g. json
16:11:17 <KunalGandhi> ok..
16:11:19 <ag__> bsd-3
16:11:50 <mugsie> i think thats OK.. will need to check
16:12:32 <mugsie> OK, any other questions on this?
16:13:02 <ag__> immediate plan is to collapse monitors to just http and tcp with bunch of parameters
16:13:07 <KunalGandhi> nope.
16:13:10 <johnsom> It's pretty slick for a single instance solution.
16:13:42 <KunalGandhi> @ag__ ... can we run a cluster of these in active/active mode ?
16:14:39 <ag__> you mean health tracker or soltuion as a whole?
16:14:48 <dougwig> Sorry, just got to ops summit. Bubble level traffic here.
16:14:54 <KunalGandhi> the solution as a whole for HA
16:15:20 <ag__> it's just an authoritative dns server
16:15:24 <mugsie> dougwig: np - I got a hotel right beside the designate one to avoid the traffic :)
16:15:42 <KunalGandhi> @ag__ .. ok
16:15:44 <mugsie> KunalGandhi: I dont think that overall we can just use this code
16:15:44 <ag__> so any HA considerations apply
16:16:12 <mugsie> we will need to make changes to move it to an Openstack style API, and add in other things like plugins
16:16:27 <mugsie> for vendor GSLB appliances
16:17:03 <ag__> yea, like we discussed it's probably a matter of writing a small converters health state -> vendor specific format
16:18:09 <johnsom> mugsie what are your thoughts on the shared memory for health state?  I don't want to see a situation where we have 20-50 health checks beating on member nodes.
16:18:11 <ag__> johnsom main synchronization process doesn't do much, so it should scale to some degree on a single core... but ofcourse has limits
16:18:13 <KunalGandhi> @mugsie .. ok.. i think for now i dont have any questions.. maybe down the line once we start integrating it with Kosmos
16:19:22 <mugsie> johnsom: yeah... I am not sure where the best place to store that is yet...
16:19:59 <mugsie> OK - next topic ...
16:20:14 <mugsie> #topic Arch Review
16:20:24 <mugsie> #link https://github.com/gslb/gslb-specs/pull/4
16:20:33 <mugsie> have people looked at this yet?
16:21:26 <KunalGandhi> i have given a very high level arch diagram, hoping to get a lot of feedback and more details to be added
16:21:33 <mugsie> I have a few initial pieces of feedback
16:21:41 <KunalGandhi> ok
16:21:52 <mugsie> designate is a driver for this#
16:21:57 <mugsie> so it should be inline
16:22:10 <mugsie> we also need the status checker to be represented
16:22:39 <KunalGandhi> the status checker for the healths status of the local VIP's ?
16:22:43 <mugsie> and I would drop the global_ prefix on the ER diagrm
16:22:48 <mugsie> yes
16:23:09 <johnsom> Yeah, we should include the end user in the diagram to highlight the touch points
16:23:15 <KunalGandhi> the health checker will be in the GSLB right ?
16:23:29 <mugsie> KunalGandhi: well, it might be
16:23:37 <mugsie> but it could also be in Kosmos
16:23:48 <mugsie> and for the default implementation it will be
16:24:08 <KunalGandhi> so are thinking of using Kosmos as a full GSLB solution ?
16:24:25 <KunalGandhi> including data and control plane ?
16:25:01 <KunalGandhi> @mugsie .. ok i will drop the global_ prefix from the er diagram
16:25:17 <mugsie> KunalGandhi: yeah, as a full solution
16:25:24 <KunalGandhi> since it is going to be it might be in its own database
16:25:37 <mugsie> that can be driven by vendor drivers if needed
16:25:47 <johnsom> +1 for having it's own DB
16:25:50 <KunalGandhi> @mugsie ... ok..
16:26:17 <mugsie> we sould be aim by default to be using Neutron LBaaS in the Regions, and Designate to move traffic by changing DNS entries
16:26:49 <mugsie> KunalGandhi: an actuall write up (using the sample rst file would be good as well)
16:26:56 <KunalGandhi> @mugsie .. makes sense
16:27:08 <mugsie> I am going to try and get this repo moved to infra later on today
16:27:35 <mugsie> so you may need to use gerrit
16:27:41 <mugsie> but I will tell you when it happens
16:28:02 <KunalGandhi> so the default gslb-plugin-driver is where we can have other vendor drivers
16:28:22 <mugsie> yup
16:28:26 <KunalGandhi> and default driver will use neutron lbaas and designate
16:28:55 <mugsie> ++
16:29:11 <mugsie> Any other comments?
16:29:13 <KunalGandhi> @mugsie .. ok understood
16:29:23 <KunalGandhi> i will update the arch diagram
16:29:32 <mugsie> ty
16:29:34 <KunalGandhi> any other feedback in terms of the format
16:29:43 <KunalGandhi> of the doc
16:29:57 <mugsie> I would prefer source files
16:29:57 <KunalGandhi> i will include a write up and an updated arch diagram
16:30:09 <mugsie> not just jpgs, if possible
16:30:22 <mugsie> it is a mistake we made in designate
16:30:34 <KunalGandhi> source files as in folders (modules) as place holders
16:30:43 <mugsie> and all our arch diagrams have to be updated by one person now
16:31:17 <mugsie> #topic Open Discussion
16:31:24 <mugsie> any off agenda items/
16:31:26 <mugsie> ?*
16:33:50 <mugsie> going once ...
16:33:53 <xgerman> guess we are done :-)
16:33:58 <dougwig> seconded
16:34:03 <mugsie> OK, thanks everyone!
16:34:10 <xgerman> bye
16:34:11 <mugsie> #endmeeting