09:00:33 <gsagie> #startmeeting dragonflow
09:00:34 <openstack> Meeting started Mon Dec 28 09:00:33 2015 UTC and is due to finish in 60 minutes.  The chair is gsagie. Information about MeetBot at http://wiki.debian.org/MeetBot.
09:00:35 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
09:00:37 <openstack> The meeting name has been set to 'dragonflow'
09:00:50 <gsagie> Anyone else for the dragonflow meeting beside nick-ma and Shlomo?
09:00:59 <nick-ma> hi, all.
09:01:46 <gsagie> Shlomo_N: please tell Yuli to join as well
09:02:04 <gsagie> #topic security groups
09:02:23 <gsagie> ok so we dont have alot of people here but lets try to go over the important points
09:02:35 <Shlomo_N> ok
09:02:50 <gsagie> I have proposed a new security group design which combine some ideas we brought up and combine it with dingboopt design
09:03:08 <gsagie> #link security group new design https://review.openstack.org/#/c/261903/
09:03:39 <gsagie> nick-ma: basically it has some nice ideas how to implement security groups very efficiently both in the data plane but also in the control plane, no need to sync and manage all these ipsets
09:04:02 <gsagie> so i believe this is going to be a good solution, but please review and let me know your comments/ideas
09:04:07 <gsagie> maybe i missed something
09:04:24 <nick-ma> yes, i'm reviewing.
09:04:43 <gsagie> ok great, feel free to ask in #openstack-dragonflow or email if you have any questions
09:04:58 <gsagie> dingboopt is here?
09:05:04 <dingboopt> yes
09:05:07 <yuli_s> I am here
09:05:28 <gsagie> dingboopt: anything you would like to update regarding the security groups work?
09:06:16 <dingboopt> currently no
09:06:20 <gsagie> do you feel you can start working on the code yet? i think that regardless of which design, we still need to populate the DF DB with all the security group configuration from the Neutron plugin
09:06:30 <gsagie> would you like to start working on that as we review the design?
09:06:36 <dingboopt> yes
09:06:52 <dingboopt> I can start working on the code
09:07:29 <shz> o/
09:07:33 <gsagie> #action dingboopt start working on DF neutron plugin to configure security groups data to DF DB
09:07:38 <gsagie> Hi shz o/
09:08:00 <gsagie> #info shz, dinboopt, nick-ma, yuli_s, Shlomo_N in dragonflow meeting
09:08:48 <gsagie> ok, so we can start with that, this is work that needs to be done anyway and we can iterate on the security group spec in the mean time
09:08:58 <gsagie> hi BillyXu, thanks for joining
09:09:24 <gsagie> anyone would like to add anything else regarding security groups? or have any question?
09:09:37 <shz> hi gsagie, I think sg is need for dragonflow, but it shoulb be done at current time?
09:09:49 <yuli_s> gsagie: I submited by comments
09:09:54 <gsagie> shz: what do you mean?
09:10:41 <gsagie> shz: we must work on security groups its a needed feature, and i think most of the parts we need for it are there
09:10:50 <shz> now I think we should focus on making l2/l3 stable
09:11:00 <BillyXu> yuli_s: yes, see your comments
09:11:00 <shz> gsagie, I agree
09:11:04 <gsagie> shz: yes i agree, we do this in parallel
09:11:19 <gsagie> shz: our next topic is testing :)
09:11:27 <shz> ok :)
09:12:05 <gsagie> #action gsagie, dingboopt  continue and iterate on the security group design
09:12:09 <gsagie> #topic testing
09:12:41 <gsagie> ok so regarding testing, we have some good patch from yuli_s that adds a nice idea to fullstack tests
09:12:41 <shz> we need unit terst :)
09:13:08 <yuli_s> Yes, it is here
09:13:08 <gsagie> #info fullstack tests - https://review.openstack.org/261001
09:13:10 <yuli_s> https://review.openstack.org/261001/
09:13:32 <gsagie> shz: i agree, we need to start adding unit tests (we have fullstack/tempest and rally integration)
09:13:45 <gsagie> anyone can work on unit tests? shz any chance?
09:14:05 <gsagie> even just few examples so we can continue and iterate on it
09:15:05 <shz> do ohter have any idea?
09:15:06 <gsagie> Shlomo_N: you were suppose to start working on it, are you close to finish with RethinkDB?
09:15:41 <Shlomo_N> I think I am, still working on it.
09:15:46 <gsagie> yuli_s: you need to add more tests for port creation/deletion as well
09:15:57 <gsagie> #action yuli_s add tests for VM port creation/deletion to fullstack tests
09:16:12 <yuli_s> Ok,
09:16:25 <gsagie> ok, i will start writing some unit tests examples and Shlomo_N can continue when he finish and then anyone thats adding new feature will write there own tests
09:16:35 <gsagie> #action gsagie start writing some unit tests examples for Dragonflow
09:16:53 <Shlomo_N> great :)
09:17:03 <gsagie> We also need to investigate some tempest tests that we exclude
09:17:15 <gsagie> there arent many but we need to make sure no one that relates to us fails
09:17:22 <gsagie> Anyone would like to check it out?
09:18:01 <gsagie> i will put it on me for now, but if anyone wants to take some work please let me know by next meeting
09:18:11 <gsagie> #action gsagie fix failing tempest tests
09:18:36 <shz> gsagie, thx for your hard work
09:18:51 <nick-ma> do you have the blacklist of tempest tests that fail with df?
09:19:20 <gsagie> Regarding scale we still need to test both data path and control path with projects like Shaker/VMTP and scale environments
09:19:45 <gsagie> nick-ma: the excluded tests are in devstackgaterc file (in devstack folder)
09:20:20 <gsagie> nick-ma: but many of the excluded tests are either not relevant for networking or not related to Dragonflow, so the first task will be to find which tests we actually need to solve
09:21:16 <gsagie> nick-ma: any update on the scale env? or its too early?
09:21:17 <nick-ma> ok. i will review it when i finish zookeeper support. maybe before next meeting, i'm not sure.
09:21:31 <gsagie> nick-ma: thats great, thank you
09:21:57 <gsagie> #action nick-ma help review failing tempest tests
09:22:46 <nick-ma> kun_huang is preparing, afaik.
09:22:59 <gsagie> kun_huang: here by any chance?
09:23:20 <nick-ma> i think it takes time.
09:23:34 <gsagie> nick-ma: yeah, for sure
09:23:41 <gsagie> not easy to bring these enviorments
09:24:04 <gsagie> #action gsagie verify with kun_huang progress on scale env
09:24:22 <gsagie> #action gsagie Write test plan for scale and performance testing on control and data path
09:24:52 <gsagie> nick-ma: there are some tests that we are planning to do which are not only related for scale, L3 performance with projects like VMTP/Shaker can already be tested
09:24:57 <gsagie> and we can start tunning the pipeline
09:25:29 <gsagie> Also we are planning to create a small automated framework to test various DBs behavior and performance
09:25:47 <gsagie> #action gsagie work with Aharon to define test framework for DB drivers and servers
09:26:02 <gsagie> ok, before we move to the next topic, anyone has anything else for testing?
09:26:54 <gsagie> #topic DB
09:27:20 <gsagie> ok, noticed some great work from nick-ma on adding zookeeper driver
09:27:49 <gsagie> #info nick-ma adding Zookeeper driver, review patch - https://review.openstack.org/261731
09:27:56 <gsagie> nick-ma: want to update us how thats going?
09:28:14 <gsagie> the path i put above is just first part to make clustering configurable
09:28:15 <nick-ma> i've finished the demo today, but it still needs testing locally.
09:28:31 <gsagie> nick-ma: cool, very interested to see it :)
09:28:37 <gsagie> thanks for working on that!
09:28:49 <nick-ma> i just sent out an email discussing the cluster configuration. feel free to reply, thanks.
09:29:23 <gsagie> nick-ma: ok havent seen yet, i will look
09:29:38 <gsagie> #action nick-ma working on zookeeper DB driver
09:29:47 <gsagie> Shlomo_N: any update on RethinkDB patch?
09:30:31 <Shlomo_N> I found the problem, probaly will upload a patch today.
09:30:51 <Shlomo_N> This will allow using rejoin-stack for rethinkDB
09:30:53 <gsagie> ok, once that is done you can start with the unit tests part
09:31:00 <gsagie> great, good job Shlomo
09:31:01 <Shlomo_N> I know
09:31:10 <Shlomo_N> thank you :-)
09:31:15 <gsagie> #info Shlomo_N upload fix for rethinkDB rejoin stack
09:31:23 <gsagie> #info RethinkDB to merge by next meeting
09:31:45 <gsagie> Anything else on the DB?
09:32:13 <Shlomo_N> Any1 tested RAMcloud for rejoin?
09:32:51 <gsagie> Shlomo_N : RAMCloud is a little bit trickier as it saves everything in RAM, we decided previous week that we will wait with it
09:32:58 <gsagie> I will talk with Aharon about it
09:33:14 <gsagie> #action gsagie Investigate RAMCloud rejoin stack options
09:33:27 <Shlomo_N> anyway it should allow running rejoin
09:33:34 <gsagie> #topic publish-subscribe
09:34:01 <gsagie> Shlomo_N: yes, but testing is more urgent at this point, RAMCloud must always have a server that is online
09:34:09 <gsagie> otherwise all data is lost as its not writing it to disk
09:34:39 <gsagie> we will talk about it in the next meeting, i think that tests are more important at this point in time
09:35:08 <gsagie> just to update, if you noticed i have started adding some pub-sub code, its still WIP
09:35:46 <gsagie> i will write a detail design about it soon and upload, the idea is to abstract between the DB and the publish-subscribe mechanism when the end goal is to be able to only notify relevant information per compute node and not sync all
09:35:49 <nick-ma> ok. i noticed. if you can provide some spec to review? it should be awesome :-)
09:35:51 <gsagie> the configuration to all the nodes
09:36:01 <gsagie> nick-ma: yeah i will write it by next meeting
09:36:12 <gsagie> #action gsagie write spec about publish-subscribe abstraction
09:36:41 <gsagie> I will describe the advantages with this approach, but anyway its going to be optional, so you dont have to use it if you dont need it
09:37:13 <gsagie> but it has some advantages, first to DB's that dont have buildin publish-subscribe and it will be easier to optimize and sync the nodes only with the needed configuration
09:37:18 <gsagie> i will write all about it
09:37:40 <gsagie> #action Shlomo_N upload fix for RethinkDB rejoin_stack
09:37:55 <gsagie> #action Shlomo_N start working on unit tests
09:38:07 <gsagie> #topic Kuryr and Dragonflow integration
09:38:11 <gsagie> kexiaodong: here ?
09:38:47 <gsagie> ok, guess he is not here will check next meeting
09:39:14 <gsagie> #info kexiaodong sent OVS unbind to Kuryr, need to continue with Dragonflow integration
09:39:17 <gsagie> #topic bugs
09:40:24 <gsagie> #info shihanzhang fixing sync bug with etcd - please review patch https://review.openstack.org/261589
09:40:29 <gsagie> shz: thanks for working on that
09:40:39 <gsagie> gampel: when you have time, please review it looks good to me
09:40:57 <gsagie> anyone have anything else to add for Bugs?
09:41:37 <nick-ma> i'd like to share a bug, https://bugs.launchpad.net/dragonflow/+bug/1529326
09:41:38 <openstack> Launchpad bug 1529326 in DragonFlow "network creation error but it is recorded at neutron DB side." [Undecided,New]
09:42:35 <gsagie> #info https://bugs.launchpad.net/dragonflow/+bug/1529326
09:42:41 <gsagie> nick-ma: thanks, will take a look at it
09:43:20 <Shlomo_N> Most (if not all) of the bug I opened aren't prioritized, I think we have to set priorities to bugs.
09:43:22 <gsagie> we do have a bug for rollbacks, but will take a look at this
09:43:33 <gsagie> Shlomo_N: do you have access to do it?
09:43:47 <Shlomo_N> no
09:43:47 <nick-ma> yes, i think so.
09:44:12 <gsagie> ok, i will prioritize and clean all bugs from launchpad, i will see how we can give you access for that Shlomo
09:44:25 <gsagie> #action gsagie prioritize all launchpad bugs and clean old bugs-blueprints
09:44:37 <Shlomo_N> great, 10x
09:44:38 <gsagie> #action gsagie look at https://bugs.launchpad.net/dragonflow/+bug/1529326
09:44:40 <openstack> Launchpad bug 1529326 in DragonFlow "network creation error but it is recorded at neutron DB side." [Undecided,New]
09:45:52 <gsagie> #topic open discussion
09:46:10 <gsagie> Ok everyone, our time is almost up, anyone have anything else they would like to talk about?
09:46:53 <gsagie> Thanks everyone for joining! i appreciate your time
09:47:06 <nick-ma> thanks. bye.
09:47:15 <Shlomo_N> thank you, bye
09:47:18 <gsagie> #endmeeting