14:00:27 <dmellado> #startmeeting kuryr
14:00:28 <openstack> Meeting started Mon Jul 16 14:00:27 2018 UTC and is due to finish in 60 minutes.  The chair is dmellado. Information about MeetBot at http://wiki.debian.org/MeetBot.
14:00:29 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
14:00:31 <openstack> The meeting name has been set to 'kuryr'
14:00:33 <dmellado> o/ Hi fellow kuryrs
14:00:39 <dmellado> who's here around today and not on PTO? xD
14:00:58 <dmellado> #chair dulek apuimedo
14:00:59 <openstack> Current chairs: apuimedo dmellado dulek
14:01:11 <apuimedo> I am
14:01:11 <dulek> I'm here!
14:01:20 <dmellado> I suspected the two of you :D
14:01:26 <dmellado> so, let's do a quick roll call
14:01:39 <dmellado> gcheresh_: juriarte itzikb QA Team say hi if you're around!
14:02:11 <dmellado> #topic kuryr-kubernetes
14:02:36 <dmellado> All right, I'll be polite and let you folks start if you've anything to share. Otherwise I'll go myself ;)
14:02:46 <dulek> Okay, I can start.
14:02:51 <dulek> Lemme dig the links. :)
14:03:15 <dulek> Easy one - unblocking changes making modifications in requirements.txt: https://review.openstack.org/#/c/581373/
14:03:21 <dulek> apuimedo: Please take a look.
14:03:38 <dmellado> heh, dulek awesome
14:03:44 <dmellado> let's unblock this, I also had that under my scope
14:03:52 * apuimedo looking
14:04:06 <dmellado> apuimedo: either you +A it or I'll let you be alone in the worst ever possible neighborhood in Madrid
14:04:08 <dmellado> up to you!
14:04:14 <dulek> I've found the issue that prevented usage of K8s LB from the host.
14:04:24 <dulek> Basically here's what happened:
14:04:30 <dulek> 1. My patch was working fine.
14:04:41 <dulek> 2. Neutron switched to openvswitch SG driver.
14:04:46 <dulek> 3. My patch stopped to work fine.
14:04:47 <dulek> :)
14:04:53 <dmellado> dulek: heh xD
14:05:02 <apuimedo> dmellado: going to Madrid is such a drag
14:05:07 <apuimedo> >37°C
14:05:10 <dmellado> yeah, they are now using ovs-firewall by default
14:05:12 <apuimedo> not looking forward
14:05:14 <dmellado> apuimedo: it's terrible, I know!
14:05:16 <dmellado> xD
14:05:22 <dulek> https://review.openstack.org/#/c/582517/ - this is adding correct SGs to proper VIFs. It'll need apuimedo's review.
14:05:42 <dulek> Then there's https://review.openstack.org/#/c/575188/ and https://review.openstack.org/#/c/569095/ to be merged.
14:05:44 <apuimedo> dulek: that's because we were bypassing the SGs for the kubelet
14:05:46 <apuimedo> xD
14:06:19 <dulek> apuimedo: I just drew it on the whiteboard and figured out correct SGs on the first try. ;)
14:06:53 <apuimedo> dulek: I don't recall I needed to do that
14:07:03 <apuimedo> but I'll take a look
14:07:14 <dulek> Also there's patch enabling externalIPs in OpenShift gates: https://review.openstack.org/#/c/580105/. Nothing urgent, it just simplifies the logs a bit.
14:07:24 <dulek> apuimedo: Sorry, to do what? Drawing? xD
14:07:45 <apuimedo> dulek: no, do this extra SG
14:08:04 <dmellado> now we can also unlock the stestr patch
14:08:12 <dmellado> which I was lazy enough not to do myself xD
14:08:16 <dulek> apuimedo: Ah. I've tested that on infra VM. It was required, but maybe I'm missing a simpler solution.
14:08:38 <dulek> Also I'd like to raise https://review.openstack.org/#/c/570553/ - updating pyroute2 version.
14:08:42 <apuimedo> dulek: did we merge my project fix?
14:08:47 <dulek> apuimedo: Yup!
14:08:57 <apuimedo> was it for octavia or only for lbaasv2?
14:09:02 <apuimedo> your patch
14:09:08 <dulek> apuimedo: LBaaS v2
14:09:39 <dmellado> dulek: on that patch, did you see what was going on with multinode?
14:09:48 <apuimedo> ok
14:09:54 <dulek> dmellado: It's from the time multinode wasn't working. Let me recheck. :)
14:10:06 <dmellado> dulek: please do ;)
14:10:10 <dulek> I remember folks reporting more stability with newer pyroute2, so I believe it's worth to update it.
14:10:33 <dmellado> dulek: in any case +2 to that
14:10:33 <dulek> And probably a question to dmellado - is updating lower-constraints.txt enough to get a newer version?
14:10:36 <apuimedo> dulek: I have it working on my machine without this SG stuff
14:10:40 <dmellado> dulek: it should
14:10:45 <apuimedo> and the driver is openvswitch
14:10:51 <apuimedo> what did extactly fail?
14:11:06 <dmellado> apuimedo: maybe some evil rules on upstream infra? xD
14:11:14 <dulek> apuimedo: For some reason it was working on my local env as well. But not on infra frozen VM.
14:11:51 <dulek> apuimedo: And what was failing is connectivity to the LB. What I've did:
14:12:24 <apuimedo> dulek: in my machine the problem I observed was docker messing with iptables
14:12:38 <dulek> apuimedo: I've added the first rule (to the LB VIF) and tried curling the LB while tcpdumping LB namespace.
14:12:49 <dulek> apuimedo: I've noticed that traffic gets in but never gets back.
14:12:54 <apuimedo> dulek: https://review.openstack.org/#/c/580198/ is not merged yet
14:12:58 <dulek> apuimedo: So I've added another rule to kubelet VIF.
14:13:24 <dulek> apuimedo: Hm! I really thought it's merged…
14:13:43 <apuimedo> with it merged you shouldn't have had the bug :-)
14:14:24 <dulek> apuimedo: Okay, I'll try that.
14:14:45 <dulek> apuimedo: Though I don't understand how changing a project would help, really.
14:14:51 <dulek> apuimedo: Ah, it'll get correct default SG?
14:15:03 <dmellado> I think so
14:15:13 <dmellado> so dulek, please try depending on that patch with yours
14:15:20 <dmellado> just in case adding new SG isn't really needed
14:15:26 <dulek> Okay, cool.
14:15:48 <pliu> https://review.openstack.org/#/c/574131/ Could it be merged or is therer still concerns on it?
14:15:51 <apuimedo> dulek: exactly
14:15:53 <apuimedo> xD
14:15:54 <dulek> So back to pyroute2 upgrade. Are you okay with that?
14:16:01 <dmellado> dulek: +1 on that, no worries
14:16:06 <apuimedo> welcome Peng Liu!
14:16:27 <pliu> hi apuimedo
14:16:29 <apuimedo> thanks for the reminder
14:16:30 <dmellado> lol pliu
14:16:32 <apuimedo> I'm merging it now
14:16:36 <dmellado> seems that we forgot to workflow it
14:16:39 <dmellado> thanks for pinging us ;)
14:17:04 <dmellado> dulek: anything else from your side?
14:17:20 <dulek> dmellado: pyroute2 upgrade. Are we okay with that?
14:17:30 <dulek> Given that it passes gates, of course.
14:17:33 <dmellado> dulek: I replied jsut like, 2 lines before?
14:17:43 <dmellado> didn't it get through?
14:17:48 * dmellado having also isp issues today...
14:17:54 <dmellado> in any case dulek
14:17:58 <dmellado> +2 once it passes gates
14:18:00 <dmellado> yeah
14:18:04 <dulek> Okay, awesome!
14:18:18 <dmellado> apuimedo: you wanna go next?
14:18:44 <apuimedo> let me see what I have
14:20:41 <apuimedo> no. I've got nothing
14:20:46 <snapiri> dmellado: I finally got https://review.openstack.org/#/c/579599/ in. Unfortunately I will not be available to work on Kuryr more in the near future.,,
14:20:52 <apuimedo> just need to address comments to my patches
14:21:01 <apuimedo> so we can finally have the registry working
14:21:19 <apuimedo> and it would be nice if somebody added skydns support also for kubernetes devstack
14:21:23 <dmellado> snapiri: thanks! and sorry to hear that... Hope things can get back there in the future!
14:21:41 <dmellado> snapiri: I assume same goes for irenab :\
14:22:17 <snapiri> dmellado: Thanks :) I am not sure, but I believe that if she will do stuff it will be smaller things :-\
14:22:41 <snapiri> Will try to keep track of your work still...
14:23:08 <apuimedo> thanks a lot snapiri!
14:23:08 <dmellado> snapiri: thanks again! plese do let me know if your situation changes
14:23:32 <snapiri> Cheers. Will do xD
14:23:42 <snapiri> apuimedo: thanks :)
14:24:18 <pliu> One more thing regarding to the multi vif implementaion, It's the basis of many features like NPWG Spec, SRIOV and nested DPDK driver. It would be nice to have a consensus on the approach
14:24:40 <dmellado> pliu: did you get to see (and speak with) garyloug on the dpdk approach?
14:25:14 <pliu> dmellado, yes we had a meeting last week
14:25:28 <dmellado> https://review.openstack.org/#/c/559363/
14:25:32 <dmellado> pliu: ^^
14:25:58 <dmellado> pliu: any outcome from that meeting?
14:26:10 <dmellado> please let me know if you get to do a follow-up
14:27:10 <pliu> demllado, I demoed my patch with support of different vif drivers, such as nested dpdk vif driver to him.
14:27:56 <dmellado> pliu: ack on that
14:28:08 <apuimedo> pliu: is there a recording on that?
14:28:13 <dmellado> in any case, please do sync with him on the implementation and devref
14:28:33 <dmellado> apuimedo: I really doubt it. pliu. any chance you could put up together a document
14:28:41 <dmellado> like, a devref with the final implementation thoughts
14:28:43 <dmellado> ?
14:29:04 <apuimedo> dmellado: isn't the spec for the CRD enough?
14:29:32 <dmellado> apuimedo: probably, but I'd like to get the outcome from their meeting in some docs
14:29:48 <dmellado> so that'd be the closest doc that I can think about now
14:30:06 <pliu> I made a new patch set for https://review.openstack.org/#/c/575629/
14:30:46 <pliu> And You can find the patch that integrate nested dpdk driver in https://github.com/pliurh/kuryr-kubernetes/commits/dpdk-nested
14:31:08 <pliu> in https://review.openstack.org/#/c/575629/, I made some documentation update
14:31:20 <apuimedo> cool
14:31:44 <dmellado> thanks for the detailed info pliu
14:32:45 <dmellado> apuimedo: anything else from your side?
14:33:38 <apuimedo> no
14:34:39 <dmellado> ok, I'll go next!
14:34:48 <dmellado> I've been doing some work on the NPS as of lately
14:35:01 <dmellado> handler and driver initial draft is merged
14:35:09 <dmellado> and I've a working patch for creating SG and SGR
14:35:18 <dmellado> that said, I think I'll create CRD also for that
14:35:25 <dmellado> as a way to try not to rely that much on the neutron api
14:35:28 <dmellado> as that will be slow
14:35:55 <dmellado> so if you've got any objections do so when I upload the patch later on today xD
14:36:08 <dmellado> besides that, I also have been drafting some stuff for the Berlin summit
14:36:14 <dmellado> let's go through this
14:36:43 <dmellado> https://etherpad.openstack.org/p/kuryr-kubernetes-berlin-summit
14:37:11 <dmellado> pliu: would you like to submit a session on DPDK?
14:37:19 <dulek> dmellado: Okay, so my abstract's there, who want to copresent that one?
14:37:25 <dmellado> I've added the title and put garyloug as a tentative speaker
14:37:53 <dmellado> apuimedo: would you like to handle that one?
14:38:03 <apuimedo> pliu: handle what?
14:38:15 <apuimedo> sorry
14:38:19 <apuimedo> dmellado: handle what?
14:38:32 <dmellado> apuimedo: dulek's session on How to make OpenStack service a Kubernetes application
14:38:36 <dmellado> as a cospeaker
14:38:44 <dmellado> otherwise I'll take it
14:38:45 <pliu> demllado, I 've submitted a session on the NPWG spec
14:38:53 <apuimedo> that sounds interesting
14:38:56 <pliu> in kuryr-kubernetes
14:38:58 <apuimedo> I'll look at the abstract
14:39:05 <apuimedo> pliu: that's awesome
14:39:07 <dmellado> pliu: oh, awesome! could you add your session to our etherpad so we got that tracked?
14:39:35 <pliu> https://www.openstack.org/summit/berlin-2018/call-for-presentations/manage/22023/summary
14:40:26 <apuimedo> dmellado: hey! You wrote puimedon with a ''
14:40:29 <dulek> pliu: We don't have access to that.
14:40:31 <apuimedo> 'ó'
14:40:36 <apuimedo> it's not spanish, eh?
14:40:38 <apuimedo> xD
14:40:42 <dmellado> apuimedo: because it sounds even more loud
14:40:48 <dmellado> and so it can rhyme!
14:40:52 <apuimedo> dmellado: copón!
14:40:53 <dmellado> puimedón, cabrón! xD
14:40:55 <dmellado> xD
14:41:02 <dulek> apuimedo: That's Polish seconds "u" letter. :D
14:41:20 <apuimedo> dulek: really?
14:41:28 <apuimedo> in czech ó is just a longer o
14:41:30 <dmellado> anyways apuimedo I've seen that you already changed it, so don't complain xD
14:41:31 <apuimedo> like putting it twice
14:41:40 <dulek> apuimedo: Yep, e.g. próba - try. ;)
14:41:40 <apuimedo> dmellado: I couldn't let it stand
14:41:41 <dmellado> kinda like japanese
14:41:54 <apuimedo> dmellado: right
14:42:12 <apuimedo> except in japanese it is 'ō'
14:42:15 <apuimedo> romaji
14:42:17 <apuimedo> rather
14:42:26 <dmellado> yeah, but I was lazy enough not to look how to type that here xD
14:42:34 <dmellado> anyways, apuimedo
14:42:35 <apuimedo> compose key FTW
14:42:38 <dmellado> you take that as co-presenter?
14:42:41 <dmellado> going once
14:42:45 <dmellado> going twice
14:42:47 <dmellado> sold!
14:42:49 <apuimedo> xD
14:43:00 <apuimedo> alright, alright
14:43:09 <dmellado> I'll be adding luis to mine
14:43:12 <dmellado> so we got a few covered
14:43:22 <dmellado> and I already agreed with him, just forgot to update the etherpad
14:43:22 <apuimedo> pliu: could you add the abstract you submitted to the etherpad?
14:43:36 <apuimedo> dmellado: get irenab for the updates
14:43:38 <pliu> apuimedo, will do
14:43:43 <dmellado> apuimedo: yeah, that was my idea
14:43:49 <apuimedo> good
14:43:51 <apuimedo> thanks pliu
14:44:09 <dmellado> apuimedo: would you like to come for the onboarding?
14:44:35 <apuimedo> dmellado: the what?
14:44:37 <dmellado> otherwise maybe we can get maysa again, she should be already around by there
14:44:39 <apuimedo> oh, yes
14:44:40 <dmellado> onboarding session
14:44:41 <apuimedo> of course
14:44:48 <apuimedo> I love to make a fool of myself on those
14:45:02 <dmellado> lol
14:45:04 <dmellado> xD
14:45:09 <dmellado> all right, so settled
14:45:15 <dmellado> I've also another session with some folks from dell
14:45:19 <dmellado> + hogepodge o/
14:45:30 <dmellado> about edge computing, interop and kuryr
14:45:36 <dmellado> so I guess that's settled
14:45:42 <apuimedo> nice
14:46:22 <dmellado> dulek: please add apuimedo as cospeaker and submit your session
14:46:28 <dulek> dmellado: I'll do.
14:46:47 <dmellado> and unless anyone has anything else to add, I'll close today's session
14:46:49 <dmellado> thanks all!
14:46:54 <dulek> I've also had idea about session on Kubernetes Cloud Provider, but I'm sure I don't know enough about it.
14:47:22 <dmellado> dulek: yeah, I was also thinking about our QA folks to try to draft up something
14:47:32 <dmellado> but my roll dice told me better luck next time
14:47:33 <dmellado> xD
14:48:24 <dmellado> anyways
14:48:27 <dmellado> thanks again folks!
14:48:29 <dmellado> #endmeeting kuryr