19:00:10 <oanson_> #startmeeting Dragonflow
19:00:11 <openstack> Meeting started Mon Jan 29 19:00:10 2018 UTC and is due to finish in 60 minutes.  The chair is oanson_. Information about MeetBot at http://wiki.debian.org/MeetBot.
19:00:12 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
19:00:14 <openstack> The meeting name has been set to 'dragonflow'
19:00:22 <oanson_> All right. Who's here for the meeting?
19:00:51 <leyal> Hi
19:01:01 <oanson_> Hi
19:01:07 <oanson_> I'm not alone! :)
19:01:35 <oanson_> All right. Let's do this quick and painless
19:01:37 <oanson_> #topic roadmap
19:01:47 <leyal> yep, this hour is problematic :/
19:01:50 <oanson_> DNS spec was merged
19:02:01 <leyal> whohoo :)
19:02:10 <oanson_> leyal, yes. Next week we'll vote if we bother to keep it. I don't see a point of doing this alone every two weeks
19:02:26 <oanson_> Upgrades spec was merged. Started working on the gate
19:02:38 <leyal> yep , make sense
19:02:41 <oanson_> #link Upgrade gate https://review.openstack.org/#/c/496837/
19:03:05 <oanson_> Would have worked better if zuul cooperated :)
19:03:27 <leyal> cool . is it's plan to run on every commit ?
19:03:42 <oanson_> Kuryr PoC is coming along. There's now the question of IPAM and distributed lock.
19:03:54 <oanson_> #link distributed lock spec https://review.openstack.org/#/c/533711/
19:04:11 <oanson_> leyal, I believe we agreed to have a pluggable mechanism, like db driver and pubsub driver?
19:04:42 <leyal> yep , indeed all relevant  people  was in the offline talk
19:04:49 <oanson_> Great!
19:04:52 <oanson_> deployment - I don't have a status on OSA.
19:05:13 <leyal> i will update the spec -
19:05:18 <oanson_> RPM - jsonmodels is finally in RDO. now it's failing on a docs error in Pike. I'll cherry pick the relevant patch, re-pip, and try again
19:05:24 <oanson_> leyal, great. Thanks!
19:05:39 <oanson_> I don't have a status on sync with Neutron DB
19:05:54 <oanson_> Gates: started working on grenade gate (see above)
19:05:57 <leyal> what is RDO ?
19:06:13 <oanson_> The Red Hat Openstack deployment thing.
19:06:25 <leyal> cool
19:06:37 <oanson_> As far as I'm concerned, it's the 'where we keep RPMs that didn't make it to Centos/RHEL' :)
19:06:56 <oanson_> Troubleshooting -
19:07:08 <oanson_> I don't remember if this was mentioned last week, but OSProfiler support was merged
19:07:17 <oanson_> Skydive work is moving ahead nicely.
19:07:32 <oanson_> #link Skydive patch https://review.openstack.org/#/c/518670/
19:07:37 <leyal> yep , i saw snapiri patch - looks cool
19:07:55 <oanson_> Yes. He says there's still much to do, e.g. cleanup of deleted objects, and testing
19:08:45 <oanson_> Lastly, the Kuryr PoC is coming along nicely (I think this was mentioned). Hopefully we can get it done in March, and then backport it to Queens including IPAM, distributed lock, and all
19:08:54 <oanson_> Hopefully. Not a must :)
19:09:02 <oanson_> Anything else for roadmap?
19:09:22 <oanson_> #topic Bugs
19:09:46 <oanson_> I went over our bug list. I don't recall anything too serious that shouldn't be handle-able
19:10:11 <oanson_> Except https://bugs.launchpad.net/dragonflow/+bug/1740739
19:10:12 <openstack> Launchpad bug 1740739 in DragonFlow "Security group mismatch for floating IP" [High,New]
19:10:22 <oanson_> Which we have to see what we do about
19:10:29 <oanson_> Anything else for bugs?
19:10:36 <leyal> not from my side
19:10:48 <oanson_> #topic Open Discussion
19:11:02 <oanson_> I want to take a second to discuss selective-proactive. We discussed it offline today
19:11:14 <oanson_> The tenant/project based solution today seems... not to work
19:11:29 <oanson_> With the recommendation of msimonin, I want to take a more fine-grained approach
19:11:37 <leyal> what is the " tenant/project based solution" ?
19:11:39 <oanson_> Each compute node registers to the objects that interest it
19:12:13 <oanson_> Currently, selective-proactive assigns a 'topic' to each object. That topic is the tenant or project (depends on the OpenStack source) to which the object belongs)
19:12:39 <leyal> The idea is to save bandwith ?
19:12:40 <oanson_> The issue we ran into with this is that for e.g. a shared network, we never get events, even if it is used
19:12:44 <oanson_> leyal, yes
19:13:16 <oanson_> So now we want each node to register to the IDs of the objects it needs. That will be our 'topic'.
19:13:43 <leyal> why we don't get event on shared network?
19:13:46 <oanson_> Only issue is that it might take time to span the entire dependency tree. We may want to push that if we already know it in advance (and we should on the server side)
19:14:00 <oanson_> leyal, shared network may belong to a different tenant, hence different 'topic'
19:14:25 <leyal> it;s something like third party tenant?
19:14:42 <oanson_> Additionally, we may need to map some of the backwards references. e.g. in DNS, the DNS object points to the lport, but selective-proactive wise, we want to receive that object if we have the lport.
19:15:05 <oanson_> This is something we need to take into account - holding a 'backwards reference which works as forwards reference for selective proactive'
19:15:44 <oanson_> I hope to upload a spec explaining all this by next week's meeting
19:15:53 <oanson_> Hopefully then it would be clearer
19:16:02 <oanson_> That's all I have for open discussion
19:16:08 <oanson_> leyal, do you have something?
19:16:16 <leyal> i don't sure that i understood but maybe we will take offline tommarow ..
19:16:29 <leyal> yep, small question -
19:16:44 <oanson_> Sure.
19:16:47 <oanson_> Go ahead
19:17:06 <leyal> the commit that send event of all object tree  -
19:17:28 <oanson_> This one: https://review.openstack.org/#/c/480196/ ?
19:17:49 <leyal> why we need that ? is came from specific issue ?
19:18:03 <leyal> yep
19:18:36 <oanson_> leyal, it tries to solve the selective proactive item mentioned earlier
19:19:02 <leyal> ok , so let's talk about that tommarow :)
19:19:06 <oanson_> e.g. a port needs a shared network. We'll never get that shared network via pub/sub, since we aren't registered to the topic.
19:19:06 <irenab> oanson, is there any overview documentation of the selective proactive?
19:19:08 <oanson_> Sire
19:19:28 <oanson_> Sure*
19:19:32 <oanson_> irenab, one sec. Looking
19:19:55 <irenab> I am more concerned about the requirements we should satisfy
19:20:14 <oanson_> irenab, there is this: https://github.com/openstack/dragonflow/blob/69644670d269ffa1fa5c3a6b8734d9b825ac8d7e/doc/source/pluggable_db.rst#selective-proactive
19:20:35 <oanson_> irenab, what do you mean?
19:20:45 <snapiri-> Sorry, just got to the end of the meeting?
19:20:53 <irenab> oanson, thanks. I would like to go over it, so maybe we can chat about the rest tomorrow
19:20:55 <oanson_> snapiri-, yes, but we're still going :)
19:21:02 <oanson_> irenab, sure
19:21:16 <oanson_> Anything else for open discussion?
19:21:36 <oanson_> snapiri-, looks like I lied. This is the end of the meeting :)
19:21:37 <irenab> maybe one thing
19:21:43 <oanson_> Shoot!
19:22:00 <irenab> we have till Feb 8 to submit talks for OS summit
19:22:06 <oanson_> Yes
19:22:37 <oanson_> We have many interesting topics we can cover. SDN cabling, FEMDC/SD-WAN, application decoupling
19:22:48 <oanson_> Sorry, 1 and 3 are the same thing :)
19:23:08 <oanson_> Also new features and Project update.
19:23:23 <irenab> the topics this time are a bit more generic, so we can also have the CNI, containers integration
19:23:34 <oanson_> Definitely
19:23:59 <irenab> lets try to come up with proposals asap
19:24:08 <oanson_> Sure. We can do that tomorrow as well
19:24:16 <irenab> +1
19:24:29 <snapiri-> I have a minor update about skydive...
19:24:35 <leyal> +1
19:24:47 <oanson_> snapiri-, go ahead
19:25:06 <snapiri-> Today I discovered a bug I introduced to skydive. Fixed it. waiting for a new fixed version to be released
19:25:18 <snapiri-> ITMW, there is a problem with the asyncio
19:25:45 <snapiri-> When I run the skydive on a main() it works
19:26:09 <oanson_> And in the controller it doesn't?
19:26:16 <snapiri-> but in the controller it is run from a thread and then the asyncio loop does not work
19:26:50 <snapiri-> I am not sure what the solution should be, but if there is anyone who is familiar with asyncio I would appreciate help
19:27:18 <oanson_> snapiri-, let's hack at it tomorrow. Debugging it by remote will be difficult
19:27:59 <snapiri-> of course. I was just saying I was looking for someone with asyncio experience :)
19:28:29 <oanson_> Sure
19:28:45 <oanson_> Anything else for open discussion?
19:29:05 <irenab> nope
19:29:10 <snapiri-> not here
19:29:17 <oanson_> leyal, last chance?
19:29:38 <leyal> nothing for me
19:29:47 <oanson_> Great. Thanks everyone for comming!
19:29:49 <oanson_> #endmeeting