17:20:44 <mmichelson> #startmeeting ovn-community-development-discussion
17:20:45 <openstack> Meeting started Thu Apr 16 17:20:44 2020 UTC and is due to finish in 60 minutes.  The chair is mmichelson. Information about MeetBot at http://wiki.debian.org/MeetBot.
17:20:46 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
17:20:48 <openstack> The meeting name has been set to 'ovn_community_development_discussion'
17:21:06 <numans> cool got it
17:21:09 <mmichelson> Turns out if I press up enough times I can get back to the previous time I started the meeting
17:21:19 <mmichelson> Anyways, it's meeting time.
17:21:21 <flaviof> lol. nice
17:21:32 <mmichelson> So who'd like to go first this week?
17:22:23 <flaviof> I suppose I can go
17:23:04 <flaviof> I spent some time looking at the changes that Ankur did for NAT external port range and proposed some follow ups
17:23:37 <flaviof> #link https://patchwork.ozlabs.org/project/openvswitch/cover/20200415221006.22628-1-flavio@flaviof.com/  port range folow up in need of review
17:24:16 <flaviof> My goal was actually to see if we could leverage this functionality to do something a little different, which in openstack is called port forwarding.
17:24:45 <flaviof> In that case the same one external_ip can be mapped to multiple internal ips, based on the (non-overlapping) port range
17:25:08 <flaviof> It seems initially that is not possoble bc there can only be one port mapping per external_ip.
17:25:30 <numans> flaviof, Thanks for the patch. I'll take a look.
17:25:39 <flaviof> I would like to see if changes on ovn schema can be done so that the external ip is not the index, but the tuple external_ip and port_range are
17:25:44 <numans> Regarding port forwarding can't we use ovn load balancer feature ?
17:25:55 <flaviof> numans: np!
17:26:02 <flaviof> that is all from me
17:26:05 <numans> I think probably it can be
17:26:20 <flaviof> numans: oh that would be awesome!
17:26:49 <numans> I'd suggest to explore if its possible
17:26:59 <mmichelson> I'll have a look at your patch, too flaviof
17:27:06 <mmichelson> So who'd like to go next?
17:27:29 <numans> I can go next
17:27:33 <flaviof> numans: yes, ovn load balancer feature could be used.
17:27:43 <flaviof> but we were hoping for something simpler
17:27:52 <flaviof> anywyas, we can talk more on that later. pls proceed
17:28:03 <numans> flaviof, sure. Thansk
17:28:08 <numans> I've been working on the I-P enhancement patches
17:28:31 <numans> I have now the patches ready which handles runtime data changes. And the initial tests show promising improvements
17:28:47 <numans> I'm hoping to submit the patches for review tomorrow
17:29:06 <numans> I also did a couple of reviews
17:29:49 <zhouhan> numans: looking forward to it!
17:29:50 <numans> I've the patches here at the moment - https://github.com/numansiddique/ovn/commits/IP_FullIP/v3/p10
17:30:13 <numans> That's it from me.
17:30:47 <mmichelson> OK cool. Who'd like to go next?
17:30:49 <numans> Forgot to mention that Anil worked with me on this.
17:31:03 <flaviof> avishnoi?
17:31:20 <numans> flaviof, No. Venkata Anil
17:31:43 <flaviof> ack. Both of them are awesome. ;)
17:31:52 <zhouhan> I can go next
17:32:40 <zhouhan> I did some review for ovn-scale-test. Had some comment on dceara's PR
17:33:58 <zhouhan> We deployed full interconnection in prod between 3 AZs in ebay. No major issues so far, but I found debugging ECMP is kind of tricky because of the "group" action
17:34:07 <zhouhan> that's it from me
17:34:26 <numans> zhouhan, that's good to hear
17:34:53 <flaviof> +1
17:35:06 <zhouhan> The route advertisement feature was really helpful, otherwise I would have to manually configure thoussands of routes
17:35:07 <mmichelson> zhouhan, about how many nodes in each AZ?
17:35:34 <_lore_> zhouhan: regarding ovn-scale-test I added something similar to openshift networkpolicy
17:35:49 <zhouhan> mmichelson: the number of nodes is around 1k, as we shared before
17:35:54 <zhouhan> for each AZ
17:36:20 <zhouhan> _lore_: oh, I merged your last PR, thanks! Is this a new one?
17:37:18 <mmichelson> zhouhan, OK
17:37:35 <mmichelson> OK, so who'd like to go next?
17:37:37 <ihrachys> i can do next
17:37:41 <mmichelson> sure ihrachys
17:38:03 <ihrachys> i've been working on multiple localnet ports for openstack peeps (their routed network feature)
17:38:06 <_lore_> zhouhan: nope
17:38:09 <ihrachys> #link https://patchwork.ozlabs.org/project/openvswitch/patch/20200415014412.449850-1-ihrachys@redhat.com/
17:38:16 <_lore_> it is another one not posted yet
17:38:35 <ihrachys> happy to report it seems to work, I've asked Maciej from openstack to validate in their env and it works fine.
17:39:06 <_lore_> I was waiting the pending one to be merged
17:39:09 <_lore_> I will post it soon
17:39:13 <mmichelson> ihrachys, that's good news!
17:39:24 <ihrachys> there are still some issues to handle (logs still flooded with messages about missing bridge mappings; plus the e-w routing between vlan tagged networks issue Dumitru mentioned at one point)
17:39:38 <ihrachys> so it's RFC for now in case someone wants to take an early look
17:39:42 <zhouhan> ihrachys: multiple localnet ports for one lswitch? Or something else?
17:40:05 <ihrachys> yes same switch, multiple localnet ports; assumption being a single port is actually mapped per chassis
17:40:16 <ihrachys> then routing between l2 segments is implemented by fabric
17:41:02 <ihrachys> there's some bg at https://mail.openvswitch.org/pipermail/ovs-discuss/2020-March/049876.html
17:41:09 <ihrachys> that's it from me
17:41:42 <flaviof> really cool ihrachys++ I will take a look too
17:41:43 <zhouhan> ihrachys: thanks for the pointer, I will study it.
17:42:35 <mmichelson> OK so anyone else want to give an update?
17:43:06 <_lore_> I can go very fast
17:43:22 <_lore_> this week I worked on iPXE support in OVN
17:43:50 <_lore_> I shared a path to OSP folks and I am waiting for feedbacks, I will post it as soon as I got some news
17:44:02 <_lore_> moreover I continued working on ovn-scale-test
17:45:54 <_lore_> that all from my side
17:46:01 <_lore_> *that's
17:47:45 <mmichelson> OK, anybody else?
17:48:02 <numans> Just a note - Dumitru moved the ovn-scale-testing code to his personal repo - https://github.com/dceara/ovn-heater
17:48:13 <numans> we use this for our internal scale testing.
17:48:41 <numans> in case someone wants to give a try :)
17:48:51 <flaviof> I like the name! ;)
17:48:53 <dceara> numans, thanks for bringing it up :)
17:49:05 <dceara> zhouhan, thanks for the ovn-scale-test reviews!
17:49:11 <numans> dceara, I thought you're not joining today
17:49:30 <dceara> numans, managed to put the kids to bed early ;)
17:50:13 <mmichelson> dceara, did you want to give an update?
17:50:25 <dceara> mmichelson, just a short update
17:50:44 <zhouhan> numans: dceara what does it mean - moved to personal repo? Is it because of my slow code review :( ?
17:51:00 <numans> zhouhan, no no. that's a different project :)
17:51:20 <dceara> regarding the scale testing code, it's just temporarily under my github account until we figure out if it's useful for the community and then we can decide if we move it somewhere else (maybe directly under ovn-org or as part of ovn-scale-test)
17:52:04 <zhouhan> dceara: would hope it is in ovn-scale-test, if it is not totally different
17:52:07 <dceara> zhouhan, no, definitely not. It's just that we need to provision quite a lot of infra to run the scale tests on our physical lab. This is what ovn-heater does: provision & run tests & collect results
17:52:41 <mmichelson> ovn-heater glues ovn-scale-test, rally, browbeat, ovn, ovs, and ovn-fake-multinode together into something useful
17:52:50 <dceara> zhouhan, right now it uses ovn-scale-test. I'm still not sure if it should be part of it or just a separate repo but we can discuss
17:52:53 <mmichelson> And makes it super easy to set up and run scale tests.
17:52:56 <zhouhan> dceara: I think that is the same goal as ovn-scale-test, which of course has lots of space for improvement :)
17:53:25 <flaviof> ovn-heater: the glue gun of ovn scaling tests! ;)
17:53:27 <dceara> zhouhan, that's definitely one option: to integrate it completely in ovn-scale-test
17:54:01 <dceara> zhouhan, that's why I said "temporarily" on my github account. It still needs work and a clear direction.
17:54:08 <zhouhan> awesome!
17:54:31 <dceara> mmichelson, that was all from me today, thanks!
17:54:47 <mmichelson> dceara, excellent
17:55:14 <mmichelson> I don't have an update of my own to give. I've mostly been doing internal things (scripts to mess with bugzilla, bug triage, etc.) so nothing to say from me here.
17:55:28 <mmichelson> Soooo I guess that's it.
17:56:09 <mmichelson> #endmeeting