| *** kiennt has joined #openstack-kuryr | 00:36 | |
| *** kiennt_ has joined #openstack-kuryr | 00:58 | |
| *** kiennt_ has quit IRC | 00:59 | |
| *** kiennt has quit IRC | 00:59 | |
| *** kiennt has joined #openstack-kuryr | 00:59 | |
| *** kiennt_ has joined #openstack-kuryr | 01:01 | |
| *** kiennt_ has quit IRC | 01:01 | |
| *** kiennt has quit IRC | 01:01 | |
| *** kiennt has joined #openstack-kuryr | 01:01 | |
| *** kiennt has quit IRC | 01:05 | |
| *** kiennt has joined #openstack-kuryr | 01:05 | |
| *** kiennt26 has joined #openstack-kuryr | 01:06 | |
| *** kiennt has quit IRC | 01:06 | |
| *** kiennt26 has quit IRC | 01:19 | |
| *** kiennt has joined #openstack-kuryr | 01:19 | |
| *** hongbin has joined #openstack-kuryr | 01:22 | |
| *** gouthamr has quit IRC | 01:35 | |
| *** kiennt is now known as kiennt_AWAY | 01:36 | |
| *** gouthamr has joined #openstack-kuryr | 01:36 | |
| *** caowei has joined #openstack-kuryr | 01:39 | |
| *** hongbin_ has joined #openstack-kuryr | 01:45 | |
| *** hongbin has quit IRC | 01:47 | |
| *** wangbo has joined #openstack-kuryr | 02:13 | |
| *** kiennt_AWAY is now known as kiennt | 02:20 | |
| *** c00281451_ has quit IRC | 03:06 | |
| *** edisonxiang_ is now known as edisonxiang | 03:06 | |
| *** gouthamr has quit IRC | 03:10 | |
| *** caowei has quit IRC | 03:47 | |
| *** caowei has joined #openstack-kuryr | 03:48 | |
| *** hongbin_ has quit IRC | 03:58 | |
| *** wangbo has quit IRC | 04:27 | |
| *** janki has joined #openstack-kuryr | 04:47 | |
| *** kiennt has quit IRC | 04:53 | |
| *** wangbo has joined #openstack-kuryr | 05:23 | |
| *** aojea has joined #openstack-kuryr | 05:44 | |
| *** aojea has quit IRC | 05:46 | |
| *** aojea has joined #openstack-kuryr | 05:46 | |
| *** wangbo has quit IRC | 06:10 | |
| *** c00281451 has joined #openstack-kuryr | 06:17 | |
| *** c00281451 is now known as zengchen | 06:17 | |
| *** yamamoto has quit IRC | 06:41 | |
| *** yamamoto has joined #openstack-kuryr | 06:45 | |
| *** pcaruana has joined #openstack-kuryr | 07:34 | |
| *** egonzalez has joined #openstack-kuryr | 07:44 | |
| *** zengchen has quit IRC | 08:02 | |
| *** zengchen has joined #openstack-kuryr | 08:02 | |
| *** ashishb_ has quit IRC | 08:05 | |
| *** wangbo has joined #openstack-kuryr | 08:33 | |
| *** vikasc has quit IRC | 09:01 | |
| *** livelace has joined #openstack-kuryr | 09:03 | |
| livelace | apuimedo, ltomasbo Hello guys, were there any cases when you connected _physical_ nodes to the trunk ports ? I'm trying to combine physical and virtual nodes | 09:07 |
|---|---|---|
| ltomasbo | livelace, I'm not sure I got the question | 09:08 |
| ltomasbo | you mean with trunks with ironic? or what? | 09:09 |
| ltomasbo | I've only used for VMs, where the VM is connected to the trunk, and then attaching subports to the trunk for the nested containers | 09:10 |
| livelace | ltomasbo, In default case, your kubernetes/openshift nodes are virtual machines, which connected to a virtual network with OVS, but I have several physical openshift nodes which I want to connect to that virtual network | 09:10 |
| ltomasbo | but then you don't need trunk ports | 09:11 |
| ltomasbo | that is the kuryr default (baremetal) case | 09:11 |
| livelace | Hm | 09:11 |
| ltomasbo | where you need to install neutron agents into the openshift physical nodes | 09:11 |
| ltomasbo | and kuryr-cni | 09:11 |
| livelace | Hm | 09:12 |
| livelace | It's interesting | 09:12 |
| ltomasbo | then, your containers will be plugged into ovs by kuryr-cni | 09:12 |
| livelace | Do you mean I need to install only neutron agents ... it's interesting ... | 09:12 |
| ltomasbo | you also need ovs ofc | 09:12 |
| livelace | yes, I understood | 09:13 |
| *** vikasc has joined #openstack-kuryr | 09:13 | |
| ltomasbo | and in the openshift case | 09:13 |
| livelace | Does it work for you ? | 09:13 |
| ltomasbo | you will need to force it to use kuryr-cni as sdn, instead of openshift-sdn | 09:13 |
| livelace | Is there any installation ? | 09:13 |
| livelace | ltomasbo, Yes, I know | 09:13 |
| livelace | Is there any installation with baremetal nodes ? Do you use it anywhere ? | 09:14 |
| ltomasbo | actually, the default devstack installation is for that case (with kubernetes instead of openshifT) | 09:14 |
| ltomasbo | you deploy a kubernetes cluster in the local node (usually VM) and the neutron server/agents | 09:14 |
| livelace | ltomasbo, Any documentation ? | 09:14 |
| ltomasbo | and the containers are connected directly to the ovs br-int | 09:15 |
| ltomasbo | livelace, yes | 09:15 |
| livelace | Cool, where ? :) | 09:15 |
| ltomasbo | https://github.com/openstack/kuryr-kubernetes/blob/master/doc/source/installation/manual.rst | 09:16 |
| ltomasbo | https://ltomasbo.wordpress.com/2017/01/29/side-by-side-and-nested-kubernetes-and-openstack-deployment-with-kuryr/ | 09:16 |
| livelace | ok, saw it before | 09:17 |
| ltomasbo | livelace, perhaps this one also helps: https://docs.google.com/presentation/d/1V4s2JVYLLhaZ873jLOca-30qCcOEoLqdfMUsDs_cRw8/edit#slide=id.p | 09:17 |
| livelace | anyway, thanks for the direction | 09:17 |
| livelace | Cool | 09:18 |
| livelace | I saw almost everything about kuryr :) | 09:18 |
| ltomasbo | :D | 09:18 |
| livelace | Cool, guys, I thought that I can attach my nodes to the flat network and work with trunks, but now I see, that case with neutron agents more preferable | 09:19 |
| livelace | Thanks again :) | 09:19 |
| *** livelace has quit IRC | 09:19 | |
| ltomasbo | also, the official documentation may cover some of your questions/issues: https://docs.openstack.org/kuryr-kubernetes/latest/# | 09:20 |
| *** yamamoto has quit IRC | 09:41 | |
| *** yamamoto has joined #openstack-kuryr | 09:42 | |
| openstackgerrit | Merged openstack/kuryr-kubernetes master: [Trivial] Doc fix K8S/K8s -> Kubernetes https://review.openstack.org/491666 | 09:49 |
| *** yamamoto has quit IRC | 09:50 | |
| apuimedo | ltomasbo: ping | 09:59 |
| ltomasbo | apuimedo, pong | 10:00 |
| apuimedo | https://blueprints.launchpad.net/kuryr-kubernetes/+spec/port-creation-control-plane-perf | 10:00 |
| apuimedo | you can put that https://review.openstack.org/#/c/504915/ implements https://blueprints.launchpad.net/kuryr-kubernetes/+spec/port-creation-control-plane-perf | 10:00 |
| apuimedo | or partially rather | 10:00 |
| ltomasbo | sure | 10:01 |
| ltomasbo | should I add this too: | 10:01 |
| ltomasbo | https://review.openstack.org/#/c/505605/ | 10:01 |
| apuimedo | let's start getting things ready for merging :-) | 10:01 |
| apuimedo | you did nice work there ;-) | 10:01 |
| ltomasbo | thanks! | 10:02 |
| ltomasbo | let me update the commit messages | 10:02 |
| apuimedo | ltomasbo: this is deletion flow though | 10:02 |
| apuimedo | so I'll make another blueprint | 10:02 |
| ltomasbo | ahh, true | 10:02 |
| ltomasbo | then only the first one | 10:02 |
| ltomasbo | apuimedo, it seems you didn't complete the last sentence of the blueprint | 10:03 |
| apuimedo | ltomasbo: https://blueprints.launchpad.net/kuryr-kubernetes/+spec/port-deletion-control-plane-perf | 10:05 |
| apuimedo | ltomasbo: good catch | 10:06 |
| apuimedo | fixed now | 10:06 |
| openstackgerrit | Luis Tomas Bolivar proposed openstack/kuryr-kubernetes master: Avoid port update neutron call during pods boot up https://review.openstack.org/504915 | 10:06 |
| apuimedo | ltomasbo: I can't wait for the outreachy project to add the readiness http endpoint when using pools | 10:08 |
| apuimedo | that will only mark kuryr-controller ready once teh pool is loaded | 10:08 |
| apuimedo | ltomasbo: dmellado: did any of you delete ports in the scale env? | 10:08 |
| apuimedo | there used to be >2k and now there's 1448 | 10:09 |
| ltomasbo | I did not | 10:09 |
| apuimedo | or maybe I misunderstood the working of the pool settings | 10:09 |
| apuimedo | and it got deleted by the pool drvier?! | 10:10 |
| *** wangbo has quit IRC | 10:10 | |
| apuimedo | ltomasbo: I know I asked this already | 10:10 |
| apuimedo | but let me go through this again | 10:10 |
| apuimedo | (maybe it should have its own chapter in the docs | 10:11 |
| apuimedo | [vif_pool] | 10:11 |
| apuimedo | ports_pool_max = 500 | 10:11 |
| apuimedo | ports_pool_min = 10 | 10:11 |
| apuimedo | ports_pool_batch = 10 | 10:11 |
| apuimedo | ports_pool_update_frequency = 180 | 10:11 |
| apuimedo | this means that each worker node pool will be at most 500 port wide | 10:11 |
| apuimedo | and at minimum 10 port wide | 10:11 |
| apuimedo | and that everytime that a pool for a worker node has less than 10 ports, it will to a batch creation | 10:11 |
| apuimedo | is that right? | 10:11 |
| ltomasbo | let me see | 10:12 |
| ltomasbo | there will be no more than 500 per pool | 10:12 |
| ltomasbo | and if a pod gets created taking a port from a pool and the pool has less than 10 ports, another batch (of 10) will be created | 10:13 |
| ltomasbo | unless another batch was triggered in the last 180 seconds | 10:13 |
| openstackgerrit | Luis Tomas Bolivar proposed openstack/kuryr-kubernetes master: Optimize return_to_pool actions at Pool drivers https://review.openstack.org/505605 | 10:14 |
| apuimedo | aha | 10:14 |
| apuimedo | ltomasbo: I know that in the code or docs there was a warning against overlapping actions | 10:14 |
| apuimedo | can you clarify what is the issue with that? | 10:14 |
| ltomasbo | ahh, I forgot to add the word blueprint in the first patch, let me do it again | 10:14 |
| apuimedo | let's say that 30 pods are scheduled | 10:14 |
| apuimedo | and the batch is 10 and the minimum is 20 | 10:15 |
| apuimedo | why shouldn't we be doing to batch at the same time? | 10:15 |
| apuimedo | is it to protect Neutron | 10:15 |
| apuimedo | ? | 10:15 |
| apuimedo | (rate limitting) | 10:15 |
| openstackgerrit | Luis Tomas Bolivar proposed openstack/kuryr-kubernetes master: Avoid port update neutron call during pods boot up https://review.openstack.org/504915 | 10:15 |
| apuimedo | also, the 180s are per worker pool, right? | 10:16 |
| ltomasbo | per worker pool, right | 10:16 |
| ltomasbo | and the problem was that, as the population take some time | 10:16 |
| ltomasbo | it may happend that, for example, 5 pods get ports from a pool | 10:16 |
| ltomasbo | when it is at size 10 (in your example) | 10:17 |
| ltomasbo | so the first one will take one (9 remaining) and a population of the pool is trigger | 10:17 |
| ltomasbo | which will take a few seconds | 10:17 |
| ltomasbo | then, in the mean time, the second pod sees the pool as 8, and would trigger yet another re-population action | 10:17 |
| ltomasbo | and you may end up with 5 instead of 1 | 10:18 |
| dmellado | apuimedo: nope | 10:18 |
| dmellado | dont' tell me they magically deleted | 10:18 |
| dmellado | on its own | 10:18 |
| ltomasbo | so, the idea is that the update_frequency should be more or less proportional to the time to re-populate one pool | 10:18 |
| apuimedo | ltomasbo: IIUC, then, if you have ports_pool_max | 10:18 |
| apuimedo | it doesn't really matter | 10:18 |
| apuimedo | does it? | 10:18 |
| apuimedo | (if not for rate limitting) | 10:18 |
| ltomasbo | well, it does | 10:18 |
| ltomasbo | because it will create them anyway (as they are no in the pool until they are ready) | 10:19 |
| ltomasbo | the only different is that then, when pods are deleted, their associated ports will be deleted instead of recovered | 10:19 |
| ltomasbo | so, it will lead to a lot more neutron calls | 10:19 |
| apuimedo | ok | 10:19 |
| ltomasbo | apuimedo, note that ports are not delete from the pools when the pools are long, but only when there are ports to be recicled (due to pods being deleted) | 10:20 |
| ltomasbo | in that moment is when we check if the pool is long enough so that we don't want to keep it | 10:20 |
| apuimedo | ltomasbo: we do a lot of wasteful recycling though | 10:20 |
| apuimedo | most ports won't need any recycling now that we do not change names | 10:21 |
| ltomasbo | ohh, yes, but this patch fixes that: https://review.openstack.org/#/c/505605/ | 10:21 |
| apuimedo | ltomasbo: this patch should probably depend on the port debug one | 10:22 |
| ltomasbo | and ups... I need to check if port_debug is enabled to skip that call | 10:22 |
| apuimedo | because if port debug is enabled it should rename the prots | 10:22 |
| apuimedo | ;-) | 10:22 |
| ltomasbo | yep, it is actually dependent on the previous one | 10:22 |
| ltomasbo | but forgot that part... | 10:23 |
| ltomasbo | I'll add it right away! | 10:23 |
| apuimedo | :-) | 10:23 |
| Guest20940 | apuimedo , Itomasbo trying the project , can you provide some suggested reading or any suggestion , for outreachy project ?trying to learn from discussions here :) | 10:27 |
| Guest20940 | s/trying the/learning chunks about | 10:29 |
| *** Guest20940 is now known as deepika | 10:31 | |
| *** deepika is now known as Guest71207 | 10:32 | |
| *** caowei has quit IRC | 10:36 | |
| *** caowei has joined #openstack-kuryr | 10:36 | |
| openstackgerrit | Luis Tomas Bolivar proposed openstack/kuryr-kubernetes master: Optimize return_to_pool actions at Pool drivers https://review.openstack.org/505605 | 10:39 |
| dmellado | Guest71207: I really do advice that you first try to deploy a devstack with kuryr-k8s | 10:41 |
| dmellado | you can check the config at https://github.com/openstack/kuryr-kubernetes/blob/master/devstack/local.conf.sample | 10:42 |
| dmellado | if you're also not familiar with devstack, you can check the docs here: https://docs.openstack.org/devstack/latest/guides/single-vm.html | 10:42 |
| Guest71207 | dmellado sure thing it's in progress ,I am trying with conjure up now :) | 10:44 |
| dmellado | ;) | 10:44 |
| *** caowei has quit IRC | 10:45 | |
| Guest71207 | get back after it's success ,hope this time everything work fine ... | 10:47 |
| dmellado | Guest71207: feel free to ping us around in case you get stuck | 10:47 |
| *** yamamoto has joined #openstack-kuryr | 10:50 | |
| Guest71207 | :)) | 10:50 |
| *** yamamoto has quit IRC | 10:57 | |
| *** vikasc has quit IRC | 11:02 | |
| *** wangbo has joined #openstack-kuryr | 11:11 | |
| *** vikasc has joined #openstack-kuryr | 11:20 | |
| *** Suramya_ has joined #openstack-kuryr | 11:21 | |
| *** pgadige has joined #openstack-kuryr | 11:50 | |
| *** yamamoto has joined #openstack-kuryr | 11:56 | |
| *** wangbo has quit IRC | 11:57 | |
| *** pgadige has quit IRC | 12:09 | |
| *** apuimedo has quit IRC | 12:46 | |
| openstackgerrit | OpenStack Proposal Bot proposed openstack/kuryr master: Updated from global requirements https://review.openstack.org/496048 | 12:52 |
| openstackgerrit | OpenStack Proposal Bot proposed openstack/kuryr-libnetwork master: Updated from global requirements https://review.openstack.org/502663 | 12:52 |
| openstackgerrit | OpenStack Proposal Bot proposed openstack/kuryr-kubernetes master: Updated from global requirements https://review.openstack.org/500060 | 12:52 |
| *** Suramya_ has quit IRC | 12:58 | |
| *** apuimedo has joined #openstack-kuryr | 13:11 | |
| *** gouthamr has joined #openstack-kuryr | 13:24 | |
| *** janki has quit IRC | 13:27 | |
| *** Suramya has joined #openstack-kuryr | 13:40 | |
| *** Suramya has quit IRC | 13:45 | |
| *** hongbin has joined #openstack-kuryr | 14:18 | |
| openstackgerrit | Luis Tomas Bolivar proposed openstack/kuryr-kubernetes master: Add Pool Manager to handle subports https://review.openstack.org/498698 | 14:42 |
| *** yamamoto has quit IRC | 15:06 | |
| *** yamamoto has joined #openstack-kuryr | 15:06 | |
| *** yamamoto has quit IRC | 15:07 | |
| *** yamamoto has joined #openstack-kuryr | 15:07 | |
| *** yamamoto has quit IRC | 15:23 | |
| *** egonzalez has quit IRC | 15:25 | |
| openstackgerrit | Luis Tomas Bolivar proposed openstack/kuryr-kubernetes master: Optimize return_to_pool actions at Pool drivers https://review.openstack.org/505605 | 15:54 |
| *** vikasc has quit IRC | 16:07 | |
| *** yamamoto has joined #openstack-kuryr | 16:23 | |
| *** yamamoto has quit IRC | 16:30 | |
| *** robust has joined #openstack-kuryr | 17:34 | |
| *** hongbin has quit IRC | 18:11 | |
| *** hongbin has joined #openstack-kuryr | 18:56 | |
| *** pcaruana has quit IRC | 18:56 | |
| *** hongbin_ has joined #openstack-kuryr | 19:55 | |
| *** hongbin has quit IRC | 19:56 | |
| *** ltomasbo has quit IRC | 20:18 | |
| *** ltomasbo has joined #openstack-kuryr | 20:19 | |
| *** hongbin_ has quit IRC | 20:31 | |
| *** hongbin has joined #openstack-kuryr | 20:32 | |
| *** hongbin_ has joined #openstack-kuryr | 20:36 | |
| *** hongbin has quit IRC | 20:38 | |
| *** reedip has quit IRC | 20:41 | |
| *** reedip has joined #openstack-kuryr | 20:52 | |
| *** gouthamr has quit IRC | 20:58 | |
| *** hongbin_ has quit IRC | 21:02 | |
| *** hongbin has joined #openstack-kuryr | 21:02 | |
| *** aojea has quit IRC | 21:42 | |
| *** robust has quit IRC | 21:59 | |
| *** gouthamr has joined #openstack-kuryr | 22:16 | |
| *** hongbin_ has joined #openstack-kuryr | 22:32 | |
| *** hongbin has quit IRC | 22:34 | |
| *** gouthamr has quit IRC | 23:05 | |
| *** hongbin_ has quit IRC | 23:12 | |
| *** hongbin has joined #openstack-kuryr | 23:12 | |
| *** hongbin has quit IRC | 23:16 | |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!