09:00:26 <oanson> #startmeeting Dragonflow
09:00:27 <openstack> Meeting started Mon May 22 09:00:26 2017 UTC and is due to finish in 60 minutes.  The chair is oanson. Information about MeetBot at http://wiki.debian.org/MeetBot.
09:00:28 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
09:00:30 <openstack> The meeting name has been set to 'dragonflow'
09:00:31 <dimak_> Hey
09:00:34 <oanson> Hello. Who's here for the Dragonflow weekly?
09:00:36 <lihi> Hi
09:00:57 <oanson> Tough croud
09:01:04 <oanson> crowd*
09:01:49 <oanson> nick-ma, hi
09:02:11 <nick-ma> hi
09:02:14 <oanson> #info lihi dimak_ nick-ma in meeting
09:02:17 <oanson> Let's begin
09:02:22 <oanson> #topic Roadmap
09:02:29 <oanson> IPv6 - lihi ?
09:03:24 <lihi> I planned to update a patch by now, but I'm still rebasing it. Will upload today
09:04:17 <oanson> All right. Just reminder - portsec is the last patch for the bp, right?
09:04:23 <oanson> https://review.openstack.org/#/c/464569/
09:04:25 <lihi> Yes
09:04:32 <oanson> Great! Thanks!
09:04:46 <oanson> Chassis health - patch was merged!
09:04:56 <dimak_> 🎊
09:04:57 <lihi> Woohoo
09:05:11 <itamaro> :)
09:05:34 <oanson> Northbound API - lport finally passes gate
09:05:48 <dimak_> 🎊🎊🎊
09:06:07 <oanson> Also most votes are positive. I hope it will make it in soon
09:06:19 <oanson> I hope this will be the last time I have a ~1400 line patch :)
09:06:39 <oanson> lport migration patch: https://review.openstack.org/#/c/447366/
09:06:42 <oanson> #link lport migration patch https://review.openstack.org/#/c/447366/
09:07:06 <oanson> I am now working on VLAN aware VMs on top of it. I hope to have it done by Wednesday.
09:07:13 <oanson> That leaves us with SFC - dimak_ ?
09:07:47 <dimak_> I've rebased it on top of the passing lport patch and bringing up a machine to see if it still works :P
09:08:01 <dimak_> I still have some -1 from you on a patch with
09:08:12 <dimak_> "Requires further discussion"
09:08:19 <dimak_> The further discussion didn't follow
09:08:31 <oanson> dimak_, which patch? I don't remember what needs to be discussed
09:08:44 <dimak_> It's the one that split several OVS tables
09:08:52 <dimak_> To allow classification
09:09:07 <dimak_> We can do it with regs, just need to decide how we do it
09:09:08 <oanson> This one: [04/xx] Split L2 and egress tables ( https://review.openstack.org/#/c/429666 ) ?
09:09:13 <dimak_> Yeah
09:10:03 <oanson> All right - the discussion is that this patch splits the L2 table into two tables
09:10:15 <oanson> To allow SFC to inject flows in the middle
09:10:50 <oanson> I think this is redundant - that SFC can inject its flows in the L2 table with a higher priority. dimak_ disagrees. dimak_ - does this sum it up correctly?
09:11:02 <dimak_> Yes
09:11:09 <dimak_> Once the packet is out of SFC
09:11:14 <dimak_> it goes again to L2 table
09:11:20 <dimak_> and gets classified again
09:11:47 <dimak_> We have to mark it in some way that it did the SFC part already
09:12:06 <dimak_> (Splitting tables was your suggestion btw :P)
09:12:26 <oanson> What's the reg6 value (source lport)? The first time round, is it the source VM? The second time round, what is its value?
09:12:54 <oanson> (Many of my solutions appear to be... bad... Apparently (:. Sorry for the red herring)
09:13:18 <dimak_> It depends whether we intercept on source port or on dest port
09:13:43 <dimak_> but if I am restoring a source port classification, I restore reg6 aswell
09:13:54 <dimak_> as well
09:14:09 <oanson> And if you don't, reg6 is empty? Or the SF's lport?
09:15:10 <dimak_> Could be empty, could be SF
09:15:20 <dimak_> In a multinode env its a problem
09:15:35 <dimak_> Last SF can be on other machine than dest
09:16:04 <oanson> But then L2 is done on the other machine, no?
09:16:23 <oanson> On the SF's node, that is
09:16:25 <oanson> ?
09:16:39 <dimak_> We can decide we terminate the SFC on the machine it was classified at
09:17:03 <dimak_> It's a bit redundant but that way you can know for certain to not classify it again
09:18:15 <dimak_> Lets discuss it on the patch
09:18:20 <oanson> Wait, how do you know now that the chain has finished and to proceed to destination? What if the chain is longer?
09:18:25 <oanson> Sure
09:18:54 <oanson> That's actually it for roadmap
09:19:15 <oanson> I hope that next week we will have made some headway in VLAN aware VMs and SFC, and we can discuss future directions
09:19:30 <dimak_> +1
09:19:34 <oanson> Anything else for roadmap?
09:20:06 <oanson> #topic Bugs
09:20:26 <oanson> I want to stop here and give a big thank you for nick-ma for the excellent work during the bug smash last week!
09:20:51 <nick-ma> thanks your assignment. :-)
09:21:36 <oanson> I want to discuss bug 1691904 for a minute.
09:21:38 <openstack> bug 1691904 in DragonFlow "an object is deleted twice that causes object not found" [Critical,New] https://launchpad.net/bugs/1691904
09:21:48 <nick-ma> yes, i found it in the ci.
09:22:04 <oanson> As I wrote on the bug, I think https://review.openstack.org/#/c/466149 should close this bug.
09:22:31 <oanson> Some investigation shows that Topology sends the same events as received from the neutron server
09:22:52 <nick-ma> ok. if it is the case, i think we just ignore it.
09:23:03 <oanson> Sounds good.
09:23:11 <nick-ma> i will update the commit msg.
09:23:14 <nick-ma> to close it.
09:23:21 <oanson> Great. Thanks again!
09:23:32 <oanson> Anything else for bugs?
09:24:04 <oanson> #topic Open Discussion
09:24:32 <oanson> The floor is for the taking
09:25:11 <oanson> All right then.
09:25:20 <oanson> Thanks everyone for coming.
09:25:26 <dimak_> Bye
09:25:34 <oanson> #endmeeting