| *** hongbin has joined #openstack-kuryr | 01:19 | |
| *** aojea has joined #openstack-kuryr | 01:23 | |
| *** aojea has quit IRC | 01:27 | |
| *** openstackgerrit has joined #openstack-kuryr | 01:31 | |
| openstackgerrit | Merged openstack/kuryr-kubernetes master: Drop unnecessary import https://review.openstack.org/605418 | 01:31 |
|---|---|---|
| *** hongbin has quit IRC | 03:47 | |
| openstackgerrit | Luis Tomas Bolivar proposed openstack/kuryr-kubernetes master: Add support for OVN provider in Endpoint LBaaS driver https://review.openstack.org/603163 | 06:32 |
| openstackgerrit | Luis Tomas Bolivar proposed openstack/kuryr-kubernetes master: svc namespace isolation support for different octavia drivers https://review.openstack.org/604717 | 06:32 |
| openstackgerrit | Luis Tomas Bolivar proposed openstack/kuryr-kubernetes master: Add gate for Octavia provider OVN https://review.openstack.org/604036 | 06:32 |
| *** pcaruana has joined #openstack-kuryr | 06:33 | |
| ltomasbo | dmellado, https://review.openstack.org/#/c/605068/ | 06:40 |
| ltomasbo | dmellado, also: https://review.openstack.org/#/c/604036/ https://review.openstack.org/#/c/603163/ https://review.openstack.org/#/c/604717/ | 06:41 |
| ltomasbo | and https://review.openstack.org/#/c/605072/ | 06:42 |
| *** celebdor has joined #openstack-kuryr | 07:32 | |
| *** aperevalov has joined #openstack-kuryr | 07:32 | |
| celebdor | ltomasbo: dulek: In regards to https://review.openstack.org/#/c/604717/5/kuryr_kubernetes/config.py | 07:36 |
| *** gkadam has joined #openstack-kuryr | 07:36 | |
| celebdor | shouldn't we choose the mode depending on the provider with which we create the LB? | 07:36 |
| ltomasbo | celebdor, yep | 07:36 |
| celebdor | I am not convinced this is configuration stuff | 07:37 |
| ltomasbo | well, yes and no | 07:37 |
| celebdor | why not? | 07:37 |
| ltomasbo | for now, yes, as ovn-provider will need create since they are not creating any SG | 07:37 |
| ltomasbo | but when they fix it, it could be update or create | 07:38 |
| celebdor | shouldn't then there be a configuration and default per provider? | 07:38 |
| dulek | celebdor: I guess if your Kuryr is running on admin/service tenant, then it could be update for Octavia as well? | 07:38 |
| dulek | celebdor: BTW - I kind of liked that reraise decorator. ;) | 07:38 |
| celebdor | dulek: I didn't send the patch to remove it yet, did I? | 07:39 |
| celebdor | I have the patch more or less ready | 07:39 |
| celebdor | we use it in two places | 07:39 |
| dulek | celebdor: Nah, haven't seen it yet. If it isn't necessary it's okay to remove it of course. | 07:39 |
| celebdor | in one it doesn't really give any advantage | 07:40 |
| dulek | celebdor: Okay, that settles it. :) | 07:40 |
| ltomasbo | yep, as of today, ovn-provider needs create (as they are using the default tenant SG) and amphora driver will require update and admin tenant | 07:40 |
| celebdor | in the other case, in retry.py, removing it makes things a bit awkward (although more explicit) | 07:40 |
| celebdor | I will not remove it from retry until I find a nicer way to put it | 07:40 |
| ltomasbo | dulek, celebdor: what patch are you talking about now? | 07:41 |
| ltomasbo | (seems interesting but I'm not following, I guess I was not present on the previous conversation) | 07:41 |
| celebdor | ltomasbo: My stated intention to drop oslo.utils | 07:42 |
| ltomasbo | celebdor, ok! | 07:42 |
| celebdor | I said it yesterday, that I want to drop deps right and left | 07:42 |
| ltomasbo | yep yep, and this is about dropping the retry decorator... | 07:42 |
| celebdor | ltomasbo: | 07:45 |
| celebdor | ltomasbo: no | 07:45 |
| celebdor | about dropping the reraise context manager from the retry handler | 07:45 |
| openstackgerrit | Antoni Segura Puimedon proposed openstack/kuryr-kubernetes master: drop reraise from the controller https://review.openstack.org/605605 | 07:45 |
| celebdor | ltomasbo: dulek: This is getting it out of the lbaas code ^^ | 07:46 |
| celebdor | the retry one I'll look a bit more | 07:46 |
| celebdor | cause I don't like how I replaced it | 07:46 |
| dulek | celebdor: Retry in watcher? | 07:46 |
| ltomasbo | in a related noted (now that I see the cleanup_bogus_pool) when should we remove lbaasv2 related code? | 07:48 |
| ltomasbo | (as well as non-daemonized cni) | 07:49 |
| celebdor | ltomasbo: not sure I follow | 07:49 |
| ltomasbo | this cleanup_bogus_pool function was just there for the lbaasv2 driver | 07:49 |
| ltomasbo | or it was not? | 07:50 |
| celebdor | don't recall | 07:51 |
| ltomasbo | def _cleanup_bogus_pool(self, lbaas, pool, lb_algorithm): | 07:51 |
| ltomasbo | # REVISIT(ivc): LBaaSv2 creates pool object despite raising an | 07:51 |
| ltomasbo | # exception. The created pool is not bound to listener, but | 07:51 |
| ltomasbo | # it is bound to loadbalancer and will cause an error on | 07:51 |
| ltomasbo | # 'release_loadbalancer'. | 07:51 |
| celebdor | dulek: this http://paste.openstack.org/show/730988/ | 07:51 |
| celebdor | but I think it should be simpler | 07:51 |
| ltomasbo | ohh, now if I paste 5 lines, I received a sigyn alert! | 07:51 |
| celebdor | cause now it's blindly just doing practically what the context manager was doing | 07:52 |
| ltomasbo | about stopping now or I'll be kick out! xD | 07:52 |
| celebdor | ltomasbo: wow | 07:52 |
| celebdor | That's cool | 07:52 |
| dulek | celebdor: Hey, is it used anywhere? I think we have a few dead handler classes? | 07:52 |
| celebdor | dulek: where, the retry handler? | 07:53 |
| ltomasbo | celebdor, yep, it even says, this time is ok because I've been for a while in the channel... | 07:53 |
| dulek | celebdor: I'm just not sure. | 07:53 |
| celebdor | dulek: it is in the pipeline | 07:53 |
| dulek | Damn. | 07:53 |
| celebdor | dulek: look for h_retry.Retry | 07:54 |
| celebdor | it's the one responsible for retrying things when we do resource not ready | 07:54 |
| celebdor | ltomasbo: you should add that to the downstream channel | 07:54 |
| celebdor | I've seen people paste big stuff there | 07:54 |
| celebdor | xD | 07:54 |
| ltomasbo | xD | 07:55 |
| *** garyloug has joined #openstack-kuryr | 08:23 | |
| openstackgerrit | Michał Dulko proposed openstack/kuryr-kubernetes master: Add HA gate https://review.openstack.org/588223 | 08:24 |
| *** janki has joined #openstack-kuryr | 08:27 | |
| celebdor | ltomasbo: dulek: I think https://github.com/openstack/kuryr-kubernetes/blob/master/kuryr_kubernetes/handlers/retry.py#L59-L63 is incorrect | 08:42 |
| celebdor | if for some reason we get an exception in the sleep code | 08:42 |
| *** Sigyn has quit IRC | 08:44 | |
| dulek | celebdor: I'll need a coffee to analyze this. That stuff is overly complicated as well. | 08:45 |
| celebdor | it will never set the handler health | 08:45 |
| *** Sigyn has joined #openstack-kuryr | 08:45 | |
| *** Sigyn has quit IRC | 08:45 | |
| celebdor | dulek: that's the problem with reraise | 08:45 |
| ltomasbo | celebdor, that code is like black magic to me, only look at it when there is no other option... | 08:45 |
| celebdor | it makes it looks simple | 08:45 |
| celebdor | xD | 08:45 |
| *** Sigyn has joined #openstack-kuryr | 08:45 | |
| *** Sigyn has quit IRC | 08:45 | |
| celebdor | ltomasbo: dulek let me paste you what I think is the right flow | 08:45 |
| *** Sigyn has joined #openstack-kuryr | 08:48 | |
| *** Sigyn has quit IRC | 08:48 | |
| openstackgerrit | Antoni Segura Puimedon proposed openstack/kuryr-kubernetes master: Set handler health even in _sleep exception https://review.openstack.org/605620 | 08:49 |
| openstackgerrit | Antoni Segura Puimedon proposed openstack/kuryr-kubernetes master: Set handler health even in _sleep exception https://review.openstack.org/605620 | 08:49 |
| celebdor | ltomasbo: dulek: ^^ | 08:49 |
| *** Sigyn has joined #openstack-kuryr | 08:50 | |
| *** Sigyn has quit IRC | 08:51 | |
| celebdor | ltomasbo: dulek: only StopWatch remaining from oslo.utils | 08:51 |
| *** gkadam is now known as gkadam-afk | 08:52 | |
| ltomasbo | celebdor, man you are fast... | 08:52 |
| celebdor | which has a 1y7m message from ivc_ | 08:52 |
| celebdor | stating that it should probably be integrated with retry | 08:52 |
| celebdor | and it maybe should | 08:52 |
| dulek | celebdor: xD, consider integration. :D | 08:52 |
| celebdor | cause it is a different source of exponential backoffs | 08:53 |
| celebdor | :/ | 08:53 |
| celebdor | does anybody volunteer? | 08:53 |
| celebdor | xD | 08:53 |
| *** Sigyn has joined #openstack-kuryr | 08:56 | |
| celebdor | heh, there's these few places with untouched ivc_ code that are scary | 08:57 |
| celebdor | xD | 08:57 |
| *** gkadam-afk is now known as gkadam | 08:59 | |
| ltomasbo | yep, most of the ivc_ notes look scary... | 09:01 |
| *** janki has quit IRC | 09:13 | |
| celebdor | ltomasbo: so not volunteering, eh? | 09:28 |
| celebdor | xD | 09:28 |
| ltomasbo | xD | 09:29 |
| openstackgerrit | Antoni Segura Puimedon proposed openstack/kuryr-kubernetes master: Set handler health even in _sleep exception https://review.openstack.org/605620 | 09:53 |
| *** garyloug has quit IRC | 11:01 | |
| openstackgerrit | Antoni Segura Puimedon proposed openstack/kuryr-kubernetes master: Add support for pre-commit-hooks https://review.openstack.org/605649 | 11:07 |
| celebdor | dmellado: dulek: ltomasbo: be good to forgetful people like me and accept ^^ | 11:08 |
| celebdor | :-) | 11:08 |
| ltomasbo | celebdor, nice! | 11:09 |
| celebdor | :-) | 11:13 |
| celebdor | thanks | 11:13 |
| celebdor | We could add more nice hooks | 11:13 |
| celebdor | I'm thinking of auto-formatting | 11:13 |
| celebdor | like go-fmt with ambv black | 11:13 |
| *** pcaruana has quit IRC | 11:50 | |
| dmellado | lol | 12:23 |
| dmellado | I'll take a look | 12:23 |
| *** pcaruana has joined #openstack-kuryr | 12:39 | |
| *** rh-jelabarre has joined #openstack-kuryr | 13:43 | |
| *** maysams has joined #openstack-kuryr | 13:50 | |
| openstackgerrit | Antoni Segura Puimedon proposed openstack/kuryr-kubernetes master: drop reraise from the controller https://review.openstack.org/605605 | 14:12 |
| openstackgerrit | Antoni Segura Puimedon proposed openstack/kuryr-kubernetes master: Set handler health even in _sleep exception https://review.openstack.org/605620 | 14:15 |
| *** jistr is now known as jistr|call | 14:31 | |
| openstackgerrit | Luis Tomas Bolivar proposed openstack/kuryr-kubernetes master: [WIP] Move VIF pod annotation to CRDs https://review.openstack.org/599320 | 14:47 |
| celebdor | ltomasbo: does ^^ work? | 14:50 |
| ltomasbo | celebdor, I don't think so, I pushed some changes I had from my last time testing it | 14:51 |
| ltomasbo | celebdor, I hit one problem when trying to make the cni watching on the CRDs | 14:51 |
| ltomasbo | and is that watching CRDs is not supported... | 14:51 |
| ltomasbo | so, I'm thinking now of annotation pods with something like | 14:51 |
| ltomasbo | eth0: {port: crd-port-selflink, status: ACTIVE|...} | 14:52 |
| ltomasbo | but I need to test it | 14:52 |
| dulek | ltomasbo: Hey, I don't think I have had issues with watching a CRD. | 14:54 |
| ltomasbo | celebdor, also, it is still no polished at all... so if you want to take a quick look is fine, but don't spend time making a in-depth review | 14:55 |
| ltomasbo | dulek, really? | 14:56 |
| ltomasbo | dulek, I'm even almost sure that I read that only a couple of fields were supported... | 14:56 |
| dulek | ltomasbo: https://review.openstack.org/#/c/527243/11/kuryr_kubernetes/cni/plugins/k8s_cni_registry.py@335 | 14:56 |
| dulek | couple of fields? | 14:56 |
| dulek | ltomasbo: I suspect you have one mistake there, let me check… | 14:57 |
| ltomasbo | dulek, I think I'm refering to this: | 14:57 |
| ltomasbo | # FIXME(dulek): fieldSelector is not available for CRD's until K8s 1.9. | 14:57 |
| dulek | ltomasbo: :) | 14:57 |
| dulek | ltomasbo: But you can filter on Kuryr side for 1.8. | 14:58 |
| dulek | ltomasbo: So it's not that bad even if suboptimal. | 14:58 |
| ltomasbo | dulek, this was what it was not working for me actually; | 14:59 |
| ltomasbo | : | 14:59 |
| ltomasbo | self.watcher.add( | 14:59 |
| ltomasbo | + "%(base)s/kuryrports" % { | 14:59 |
| ltomasbo | + 'base': k_const.K8S_API_CRD}) | 14:59 |
| dulek | ltomasbo: Yeah, I thought about it. | 14:59 |
| dulek | ltomasbo: You haven't provided thread pool to daemon's Watcher instance, right? | 15:00 |
| ltomasbo | umm | 15:00 |
| ltomasbo | I was adding that to cni/daemon/service.py | 15:00 |
| dulek | ltomasbo: https://review.openstack.org/#/c/527243/11/kuryr_kubernetes/cni/daemon/service.py@186 | 15:01 |
| ltomasbo | instead of wathcing on hostname pods | 15:01 |
| dulek | ltomasbo: Without this Watcher is able only to watch one path. | 15:01 |
| dulek | ltomasbo: Ah, so you're adding it *instead*. Okay. | 15:01 |
| ltomasbo | yep (I was...) | 15:01 |
| dulek | ltomasbo: Okay, so I'm not sure what was wrong then, but I definitely have watched for CRDs with success. | 15:02 |
| ltomasbo | ok, probably I missed something... | 15:03 |
| ltomasbo | dulek, this was what I tested... http://paste.openstack.org/show/731023/ | 15:04 |
| ltomasbo | it hang waiting for the port information | 15:05 |
| dulek | ltomasbo: https://github.com/openstack/kuryr-kubernetes/blob/0478a47502beed7efece43a55a39cdd17ee2e96c/kuryr_kubernetes/cni/handlers.py#L34 | 15:05 |
| ltomasbo | ohh, f***ck | 15:06 |
| ltomasbo | see, I was missing something... | 15:06 |
| dulek | ltomasbo: Hey, at least we've found it! | 15:06 |
| ltomasbo | we --> you... | 15:06 |
| dulek | ltomasbo: Oh come on, it's teamwork. :P | 15:15 |
| ltomasbo | :D | 15:15 |
| ltomasbo | going to give it a try! | 15:16 |
| *** jistr|call is now known as jistr | 15:46 | |
| openstackgerrit | Daniel Mellado proposed openstack/kuryr-kubernetes master: [WIP] Support network policy update https://review.openstack.org/605150 | 16:01 |
| celebdor | ltomasbo: dulek: Folks we're on k8s 1.11 | 16:03 |
| celebdor | it means I should probably update the damned devstack | 16:03 |
| openstackgerrit | Merged openstack/kuryr-kubernetes master: Add support for pre-commit-hooks https://review.openstack.org/605649 | 16:17 |
| *** gkadam has quit IRC | 17:20 | |
| *** dims_ is now known as dims | 18:05 | |
| *** jistr has quit IRC | 18:47 | |
| *** jistr has joined #openstack-kuryr | 18:49 | |
| *** maysams has quit IRC | 18:55 | |
| *** jistr has quit IRC | 19:08 | |
| *** jistr has joined #openstack-kuryr | 19:08 | |
| *** jistr has quit IRC | 19:23 | |
| *** jistr has joined #openstack-kuryr | 19:26 | |
| *** jistr has quit IRC | 19:28 | |
| *** jistr has joined #openstack-kuryr | 19:29 | |
| *** jistr has quit IRC | 19:34 | |
| *** jistr has joined #openstack-kuryr | 19:37 | |
| *** jistr has quit IRC | 19:39 | |
| *** jistr has joined #openstack-kuryr | 19:49 | |
| *** pcaruana has quit IRC | 19:54 | |
| *** aojea has joined #openstack-kuryr | 19:57 | |
| *** aojea has quit IRC | 20:32 | |
| *** aojea has joined #openstack-kuryr | 20:33 | |
| *** rh-jelabarre has quit IRC | 21:10 | |
| *** aojea has quit IRC | 22:38 | |
| *** celebdor has quit IRC | 23:10 | |
| *** maysams has joined #openstack-kuryr | 23:59 | |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!