| *** hongbin has joined #openstack-kuryr | 02:02 | |
| *** gkadam has joined #openstack-kuryr | 02:45 | |
| openstackgerrit | Ghanshyam Mann proposed openstack/kuryr-libnetwork master: DNM: Testing all legacy jobs on bionic https://review.openstack.org/642321 | 03:34 |
|---|---|---|
| *** gkadam has quit IRC | 03:47 | |
| *** hongbin has quit IRC | 03:58 | |
| *** janki has joined #openstack-kuryr | 04:46 | |
| *** ccamposr has joined #openstack-kuryr | 06:51 | |
| *** pcaruana has joined #openstack-kuryr | 07:00 | |
| *** ltomasbo has joined #openstack-kuryr | 07:07 | |
| *** gcheresh_ has joined #openstack-kuryr | 07:40 | |
| *** dmellado has quit IRC | 08:20 | |
| *** maysams has joined #openstack-kuryr | 08:22 | |
| *** janki has quit IRC | 09:30 | |
| *** jchhatbar has joined #openstack-kuryr | 09:30 | |
| *** alisanhaji has joined #openstack-kuryr | 09:55 | |
| *** dmellado has joined #openstack-kuryr | 10:25 | |
| dmellado | hiya folks! | 10:27 |
| dulek | dmellado: Whooo! | 10:29 |
| dulek | dmellado: Hi! | 10:29 |
| dmellado | dulek, it's not me, it's just a zombie carcass what remains | 10:30 |
| dmellado | sleep? what's that anymore? xD | 10:30 |
| dulek | dmellado: Wonderful, you'll be Kuryr's team US ambassador. | 10:31 |
| dmellado | how are things going? | 10:32 |
| dmellado | dulek, the good thing is that I don't think that jet lag would affect me at all | 10:32 |
| maysams | Hey dmellado, welcome back :-) | 10:37 |
| dulek | dmellado: Things are quite okay. ltomasbo and maysams are resolving some bugs related to NP. | 10:37 |
| dulek | dmellado: I'm working on releasing on OCP side (hey, it's still ongoing, but we're really close now). | 10:37 |
| dulek | dmellado: Also at the moment I'm trying to run upstream NP tests on the gate. My offline run found us one major bug. | 10:38 |
| dmellado | hi maysams ! | 10:38 |
| dmellado | a-ha, so NP is still giving off some trouble | 10:38 |
| dulek | dmellado: Besides that you're probably aware of teams reorganization, right? | 10:38 |
| dmellado | dulek, overall yeah | 10:38 |
| dulek | dmellado: Actually it's much better than I expected. | 10:39 |
| dmellado | but you folks wouldn't believe the amount of emails that I got in my inbox | 10:39 |
| dmellado | xD | 10:39 |
| dmellado | so maysams dulek ltomasbo, what's the overall status in terms of NP | 10:39 |
| dulek | dmellado: I thought it'll explode on upstream tests, but implementation seems pretty solid. Some issues with missing features and maybe orphaned resources, sometimes. | 10:39 |
| dmellado | please, please, please tell me that it didn't did boom | 10:39 |
| dmellado | oh, awesome! | 10:39 |
| dmellado | xD | 10:39 |
| dulek | dmellado: We're in testing/bugfixing phase IMO. | 10:40 |
| ltomasbo | dmellado, yep! there was a lot more use cases | 10:40 |
| dmellado | gcheresh_, any news in terms of that testing? | 10:40 |
| dulek | dmellado: Only 3 tests out of 8 exploded. 1 because of my env's issue and 2 because we're missing a feature to use port names instead of port numbers. | 10:40 |
| ltomasbo | dmellado, and we discover several services/lbaas related functionality missing too | 10:40 |
| dmellado | -now is when you reply 'what testing' and I take a flight to TLV just to have a few words with you xD xD xD | 10:40 |
| dmellado | ltomasbo, again Octavia? | 10:41 |
| ltomasbo | dmellado, no, not octavia, I mean the svc (kuryr) implementation was missing a lot of use cases | 10:42 |
| ltomasbo | dmellado, like user editing the service definition, using named ports instead of numbers, ... | 10:42 |
| dmellado | oh, I see | 10:42 |
| dmellado | in any case it's great that this showed up | 10:42 |
| dmellado | but that also means that we do need to improve our test plans quite a lot | 10:42 |
| openstackgerrit | Michał Dulko proposed openstack/kuryr-kubernetes master: Add option to deploy coredns https://review.openstack.org/642072 | 10:53 |
| *** dmellado_ has joined #openstack-kuryr | 11:04 | |
| *** dmellado has quit IRC | 11:05 | |
| *** dmellado_ is now known as dmellado | 11:05 | |
| *** dmellado_ has joined #openstack-kuryr | 11:07 | |
| dmellado | maysams: ltomasbo all right, irc seems to be back for me | 11:14 |
| dmellado | regarding NP | 11:14 |
| dmellado | any outstanding bugs? | 11:14 |
| dmellado | otoh dulek, what's the status regarding py37 on the gates? | 11:14 |
| * dmellado reading too many emails xD | 11:14 | |
| dulek | dmellado: https://bugs.launchpad.net/kuryr-kubernetes/+bug/1818983 | 11:14 |
| openstack | Launchpad bug 1818983 in kuryr-kubernetes "NetworkPolicy doesn't support named ports" [High,Confirmed] - Assigned to Maysa de Macedo Souza (maysa) | 11:14 |
| dulek | dmellado: Uh, riiiiight… | 11:14 |
| dulek | dmellado: We have that commit that runs 3.7 unit tests. | 11:15 |
| dulek | dmellado: And that's passing. | 11:15 |
| dulek | dmellado: But I've tried updating our Python 3 container to Fedora 29 with 3.7 and that failed miserably due to eventlet issues. | 11:15 |
| dulek | dmellado: I don't know if that's resolved. | 11:15 |
| dulek | dmellado: I can try iterating my commit again to see. | 11:15 |
| dmellado | huh, I see, there was some issue with eventlet | 11:16 |
| dmellado | dulek: yeah, please give it a retry and let's see what happens | 11:16 |
| dulek | dmellado: Take back your +W on https://review.openstack.org/#/c/633744, just to be sure it won't get merged by mistake. :P | 11:17 |
| dmellado | dulek: done xD | 11:18 |
| dmellado | I've also noticed some migration on the git repos upstream | 11:18 |
| dmellado | naming, so we'd need to change our readme's | 11:18 |
| dmellado | I'll submit a few patches later | 11:18 |
| dulek | dmellado: BTW - my efforts to run CoreDNS on gates is related to upstream K8s NP tests are calling pods using svc-name.namespace instead of IP's. | 11:23 |
| dmellado | oh, I see | 11:24 |
| dmellado | that kinda recalls me a little bit about the workaround we did do with unbound | 11:24 |
| dulek | dmellado: Do you happen to know if I can get the upstream DNS on the gate somehow? Proxying requests to 127.0.0.1 gives me loop, obviously. | 11:24 |
| dmellado | so are you trying to use coreDNS in such way? | 11:24 |
| dmellado | dulek: not sure as of now, but will check with the infra folks and get back to you | 11:24 |
| dulek | dmellado: I've asked on #openstack-infra already. ;) | 11:24 |
| dmellado | dulek: what's the outcome? who did you speak with? | 11:25 |
| dmellado | clarkb? | 11:25 |
| dmellado | dulek: how about setting a local dns, I guess that's your idea with CoreDNS | 11:26 |
| dmellado | like i.e we'd use the internal DNS for the internal domain and then forward the remaining queries to the upstream one | 11:26 |
| dulek | dmellado: I've just put the question on the channel. | 11:27 |
| dulek | dmellado: Yep, that's my idea. Maybe I could use hostNetworking, bind to 127.0.0.53 and forward to 127.0.0.1, that also seems like an option. | 11:28 |
| dulek | dmellado: At the moment I'm just trying 8.8.8.8 as the upstream one, let's see if it'll work. ;) | 11:28 |
| dmellado | I've seen times when on some specific clouds queries to 8.8.8.8 were throttled | 11:29 |
| dmellado | just letting you know in case you start missing some dns queries... | 11:29 |
| dulek | dmellado: Ah, interesting, let's see. | 11:43 |
| *** jchhatbar has quit IRC | 11:58 | |
| *** rh-jelabarre has joined #openstack-kuryr | 12:02 | |
| maysams | dmellado: I was having lunch, sry, but dulek already shared the NP bug with you | 12:11 |
| dmellado | maysams: yep, thanks! I'm trying to get back on current status | 12:12 |
| maysams | dmellado: There is also this one under review https://review.openstack.org/#/c/640216/ | 12:12 |
| dmellado | maysams: sounds like besides checking experimental is almost good to go | 12:12 |
| dmellado | but I'll be taking a look as well | 12:12 |
| dmellado | maysams: ltomasbo thanks for handling this while I was away folks | 12:13 |
| maysams | dmellado: Np! and thanks for reviews | 12:14 |
| *** mrostecki has quit IRC | 12:28 | |
| *** mrostecki has joined #openstack-kuryr | 12:31 | |
| dmellado | maysams: dulek ltomasbo for the record I might no be able to attend today's upstream meeting | 12:33 |
| dmellado | need to take the kid to the doctor so he gets some vaccines | 12:33 |
| dmellado | s/no/not | 12:33 |
| dmellado | I should be fine next one on | 12:33 |
| dulek | dmellado: With all this rise of anti-vaccines movements, I guess doing the right thing is totally a proper excuse. :D | 12:35 |
| *** janki has joined #openstack-kuryr | 12:35 | |
| dmellado | dulek: lol, I'm just waiting for these crazy folks to become so high in numbers so they realize it was a mistake | 12:35 |
| dmellado | xD | 12:35 |
| *** gcheresh has joined #openstack-kuryr | 13:05 | |
| *** gcheresh_ has quit IRC | 13:06 | |
| *** gcheresh has quit IRC | 13:11 | |
| *** gcheresh has joined #openstack-kuryr | 13:11 | |
| dulek | dmellado: Surprisingly I've got 8.8.8.8 blocked on one RAX run and working fine on another on RAX. | 13:18 |
| dulek | dmellado: Anyway you were right that doesn't always work. | 13:18 |
| dmellado | dulek: crap, I wish I would've been wrong on that xD | 13:18 |
| dmellado | in any case there should be an openstack infra dns which we should be able to use | 13:19 |
| dulek | dmellado: I think it might be relative to the cloud the job is running on. | 13:20 |
| dmellado | dulek: yeah, that's my point | 13:20 |
| dmellado | the whole openstack infra should've an internal dns *all* the clouds could speak to | 13:20 |
| dulek | dmellado: So my best bet at the moment is to run coredns with host networking making it bind to 127.0.0.<anything> and redirecting to 127.0.0.1 as upstream DNS. | 13:20 |
| dmellado | I bet a six-pack on it | 13:20 |
| dmellado | sup, that was my idea as well | 13:20 |
| dulek | Will that work? I think celebdor did something similar for openshift-dns. | 13:20 |
| dmellado | dulek: something really close to that, but using unbound, IIRC | 13:21 |
| dmellado | but that shouldn't matter | 13:21 |
| dmellado | depending on how coredns can be configured | 13:21 |
| dulek | dmellado: unbound is on the gates by default, IIRC, he didn't need to enable it. | 13:21 |
| dmellado | you might even be able to just add upstream dns as a forwarding one | 13:21 |
| dmellado | I mean, he added an additional unbound instance on the nodes | 13:22 |
| dulek | dmellado: That would be best, but I can't figure out the address, as /etc/resolv.conf points me to 127.0.0.1:53, which is unbound instance IIUC. | 13:22 |
| dmellado | dulek: https://coredns.io/explugins/forward/ | 13:22 |
| dmellado | hmmmm could totally be | 13:23 |
| dmellado | I mean 53 is for DNS | 13:23 |
| dmellado | but I'm not sure on the internal dns configuration within the infra | 13:24 |
| dmellado | I mean, localhost:53 would mean there's some dns or dns forwarding running on the gate node directly | 13:24 |
| dmellado | either that, or some kind of dnsmasq running | 13:25 |
| dmellado | dulek: did you try to set that up? to let CoreDNS forward to such address? | 13:25 |
| dulek | dmellado: I'll try in a moment. | 13:26 |
| dmellado | dulek: https://docs.openstack.org/infra/system-config/dns.html | 13:26 |
| dmellado | it'd be worth to check if just | 13:27 |
| dmellado | ns1.opendev.org | 13:27 |
| dmellado | ns2.opendev.org | 13:27 |
| dmellado | work by any chance | 13:27 |
| dulek | dmellado: I've also poked some failures of centos gate. slaweq tracked it down to ancient openvswitch 2.0 on the centos-7 gate. | 13:27 |
| dulek | dmellado: Heh, why not, trying now. | 13:27 |
| dmellado | dulek: huh, no wonder, we have had to hack the centos gate quite a lot in the past, with pip and setuptools | 13:27 |
| dmellado | so centOS starts recalling me at debian stable now and then... | 13:28 |
| dmellado | dulek: heading to the doctor now, let me know later if it gets to work! | 13:30 |
| dulek | dmellado: Hm, from my local network ns1.opendev.org doesn't resolve google.com for me, but let's see how it works in the gate. | 13:31 |
| openstackgerrit | Michał Dulko proposed openstack/kuryr-kubernetes master: Add option to deploy coredns https://review.openstack.org/642072 | 13:33 |
| *** lxkong has quit IRC | 13:42 | |
| dulek | Hey, this actually looks really nice: https://opendev.org/openstack/kuryr-kubernetes | 13:57 |
| dulek | I even see it in Polish, "Gałąź: stable/rocky" is obviously pleasing to my eyes. xD | 13:58 |
| *** alisanhaji has quit IRC | 14:30 | |
| *** FlorianFa has joined #openstack-kuryr | 14:50 | |
| *** FlorianFa has quit IRC | 14:52 | |
| *** FlorianFa has joined #openstack-kuryr | 14:52 | |
| openstackgerrit | Alexey Perevalov proposed openstack/kuryr-kubernetes master: Support sriovdp arbitrary resource names https://review.openstack.org/642491 | 15:11 |
| *** alisanhaji has joined #openstack-kuryr | 15:22 | |
| *** janki has quit IRC | 15:33 | |
| *** gcheresh has quit IRC | 16:12 | |
| *** pcaruana has quit IRC | 16:23 | |
| *** pcaruana has joined #openstack-kuryr | 16:23 | |
| dulek | dmellado, ltomasbo: Okay, so I've had a nice discussion with #openstack-infra. | 16:36 |
| ltomasbo | and...? | 16:36 |
| dulek | So thing is… Some clouds we run on are IPv6-only! | 16:36 |
| dulek | Hence my question - are we even supporting this on Kuryr level? | 16:36 |
| dulek | Because I suppose we don't on DevStack level at all. :D | 16:37 |
| ltomasbo | I don't thing so... at least not for NetworkPolicies... nor for namespace isolation... | 16:37 |
| dulek | ltomasbo: At this point it's only about pods being able to connect to outside via IPv6, not that pods will get IPv6 addresses. | 16:38 |
| ltomasbo | dulek, ahh, ok | 16:38 |
| dulek | But I doubt we set any default IPv6 gateway, or anything. | 16:38 |
| ltomasbo | dulek, well, not sure about the ovn support for ipv6 either... | 16:38 |
| dulek | To be honest I don't really know much about v6. | 16:38 |
| ltomasbo | xD same here... | 16:39 |
| dulek | Okay, giving up here, probably I'll be required to copy what celebdor did for openshift-dns. | 17:24 |
| *** aperevalov has quit IRC | 17:27 | |
| *** maysams has quit IRC | 17:57 | |
| *** ccamposr has quit IRC | 19:01 | |
| dmellado | ltomasbo: dulek so we might need to check whether that would be skip if IPv6 | 19:30 |
| *** lxkong has joined #openstack-kuryr | 20:00 | |
| *** alisanhaji has quit IRC | 21:24 | |
| *** pcaruana has quit IRC | 21:45 | |
| *** threestrands has joined #openstack-kuryr | 22:50 | |
| *** ccamposr has joined #openstack-kuryr | 23:33 | |
| *** ccamposr has quit IRC | 23:40 | |
| *** threestrands has quit IRC | 23:45 | |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!