| *** yamamoto has joined #openstack-kuryr | 00:03 | |
| *** s1061123_ has quit IRC | 00:06 | |
| *** yamamoto has quit IRC | 00:08 | |
| *** s1061123 has joined #openstack-kuryr | 00:23 | |
| *** mattmceuen has quit IRC | 00:27 | |
| *** kiennt has joined #openstack-kuryr | 00:57 | |
| *** yedongcan has joined #openstack-kuryr | 01:04 | |
| *** rwallner has joined #openstack-kuryr | 01:15 | |
| *** mattmceuen has joined #openstack-kuryr | 01:18 | |
| *** rwallner has quit IRC | 01:20 | |
| *** caowei has quit IRC | 01:50 | |
| *** caowei has joined #openstack-kuryr | 02:16 | |
| *** yamamoto has joined #openstack-kuryr | 02:36 | |
| *** gouthamr has quit IRC | 02:40 | |
| *** yamamoto has quit IRC | 03:08 | |
| *** caowei has quit IRC | 03:15 | |
| *** gouthamr has joined #openstack-kuryr | 04:05 | |
| *** openstackgerrit has joined #openstack-kuryr | 04:37 | |
| openstackgerrit | OpenStack Proposal Bot proposed openstack/kuryr-libnetwork master: Updated from global requirements https://review.openstack.org/484555 | 04:37 |
|---|---|---|
| *** yamamoto has joined #openstack-kuryr | 04:39 | |
| *** yamamoto has quit IRC | 04:48 | |
| *** yamamoto has joined #openstack-kuryr | 04:51 | |
| *** yedongcan1 has joined #openstack-kuryr | 04:57 | |
| *** yedongcan has quit IRC | 04:59 | |
| *** yamamoto has quit IRC | 05:15 | |
| *** kiennt has quit IRC | 05:44 | |
| *** kiennt has joined #openstack-kuryr | 06:03 | |
| *** caowei has joined #openstack-kuryr | 06:22 | |
| *** s1061123_ has joined #openstack-kuryr | 06:22 | |
| *** s1061123 has quit IRC | 06:22 | |
| *** gouthamr has quit IRC | 06:36 | |
| openstackgerrit | Liang Cui proposed openstack/kuryr-kubernetes master: Update URLs in documents according to document migration https://review.openstack.org/485995 | 06:47 |
| *** pc_m has quit IRC | 06:58 | |
| *** pc_m has joined #openstack-kuryr | 07:00 | |
| *** pc_m has quit IRC | 07:07 | |
| *** pc_m has joined #openstack-kuryr | 07:08 | |
| *** aojea has joined #openstack-kuryr | 07:16 | |
| *** yamamoto has joined #openstack-kuryr | 07:20 | |
| *** yamamoto has quit IRC | 07:42 | |
| apuimedo | janonymous: I had an idea | 08:02 |
| *** neiljerram has joined #openstack-kuryr | 08:02 | |
| apuimedo | mmmm | 08:02 |
| apuimedo | nevermind | 08:02 |
| apuimedo | let me think about it a bit longer | 08:03 |
| apuimedo | vikasc: ping | 08:03 |
| vikasc | apuimedo, pong | 08:03 |
| *** neiljerram has quit IRC | 08:07 | |
| *** dims has quit IRC | 08:09 | |
| *** dims has joined #openstack-kuryr | 08:11 | |
| *** vikasc has quit IRC | 08:17 | |
| janonymous | apuimedo: :) | 08:18 |
| apuimedo | janonymous: well, the idea is to run pyroute under the privsep context | 08:18 |
| apuimedo | it's the lowest cost solution | 08:18 |
| apuimedo | this way the eventlet stuff won't mess with it | 08:18 |
| apuimedo | I wonder what kind of perf impact will it have | 08:19 |
| apuimedo | IIUC, if I send the patch to force privsep into fork mode, we'll only fork once | 08:19 |
| apuimedo | and then the price will be low | 08:19 |
| janonymous | apuimedo: something like own process space | 08:20 |
| janonymous | for pyroute part too | 08:20 |
| *** egonzalez has joined #openstack-kuryr | 08:21 | |
| *** dims has quit IRC | 08:21 | |
| apuimedo | janonymous: right | 08:22 |
| apuimedo | let me show you | 08:22 |
| janonymous | apuimedo: one random thought, how about using greenpool for eventlet for pyroute | 08:23 |
| apuimedo | https://github.com/openstack/os-vif/blob/8b1e906f2c356959adfc05ad7a7153dbb68ace9d/vif_plug_linux_bridge/privsep.py | 08:23 |
| *** dims has joined #openstack-kuryr | 08:23 | |
| apuimedo | janonymous: green pool? | 08:24 |
| apuimedo | I'm not familiar with it | 08:24 |
| janonymous | apuimedo: ohkay | 08:24 |
| apuimedo | janonymous: can you give me more details about greenpool? | 08:25 |
| janonymous | apuimedo: it is like thread pool with own context | 08:25 |
| janonymous | apuimedo: so if using nested eventlet threads, problem of switching might occur | 08:25 |
| apuimedo | janonymous: does any openstack project use it | 08:26 |
| janonymous | apuimedo: swift used that last i remember.. | 08:26 |
| apuimedo | I'll check it | 08:28 |
| * apuimedo currently fighting octavia devstack | 08:28 | |
| janonymous | apuimedo: But not exactly sure on that too..but privsep looks good :) | 08:28 |
| apuimedo | janonymous: can you check with privsep, then? | 08:28 |
| janonymous | apuimedo: sure, i will give a try :) | 08:29 |
| janonymous | apuimedo: till you get free :D | 08:30 |
| *** vikasc has joined #openstack-kuryr | 08:30 | |
| apuimedo | thanks | 08:31 |
| *** kural has joined #openstack-kuryr | 08:39 | |
| *** kural_ has joined #openstack-kuryr | 08:39 | |
| *** garyloug has joined #openstack-kuryr | 08:43 | |
| *** garyloug has joined #openstack-kuryr | 08:47 | |
| *** neiljerram has joined #openstack-kuryr | 09:05 | |
| *** yedongcan1 has left #openstack-kuryr | 09:43 | |
| *** kiennt has quit IRC | 09:53 | |
| *** zengchen has quit IRC | 10:07 | |
| *** vikasc has quit IRC | 10:23 | |
| *** pc_m has quit IRC | 10:41 | |
| *** vikasc has joined #openstack-kuryr | 10:42 | |
| *** yamamoto_ has joined #openstack-kuryr | 10:48 | |
| *** pc_m has joined #openstack-kuryr | 10:50 | |
| *** rwallner has joined #openstack-kuryr | 11:08 | |
| *** yboaron__ has joined #openstack-kuryr | 11:09 | |
| *** aojea has quit IRC | 11:22 | |
| *** aojea has joined #openstack-kuryr | 11:30 | |
| *** yamamoto_ has quit IRC | 11:32 | |
| *** aojea has quit IRC | 11:34 | |
| *** aojea has joined #openstack-kuryr | 11:39 | |
| *** aojea has quit IRC | 11:43 | |
| *** kzaitsev_ws has quit IRC | 11:48 | |
| *** aojea has joined #openstack-kuryr | 11:48 | |
| *** yamamoto has joined #openstack-kuryr | 11:52 | |
| *** aojea has quit IRC | 11:53 | |
| *** yamamoto has quit IRC | 11:56 | |
| *** aojea has joined #openstack-kuryr | 11:57 | |
| *** aojea has quit IRC | 12:01 | |
| *** kzaitsev_ws has joined #openstack-kuryr | 12:06 | |
| *** aojea has joined #openstack-kuryr | 12:06 | |
| *** aojea has quit IRC | 12:09 | |
| *** aojea has joined #openstack-kuryr | 12:09 | |
| *** yboaron__ has quit IRC | 12:20 | |
| *** atoth is now known as atoth_guinea_pig | 12:37 | |
| *** atoth_guinea_pig is now known as atoth | 12:37 | |
| *** yamamoto has joined #openstack-kuryr | 12:42 | |
| openstackgerrit | Antoni Segura Puimedon proposed openstack/kuryr-kubernetes master: Use a separate network for services subnet https://review.openstack.org/486134 | 12:50 |
| *** mattmceuen has quit IRC | 12:51 | |
| *** yamamoto has quit IRC | 12:52 | |
| *** yamamoto has joined #openstack-kuryr | 12:52 | |
| *** caowei has quit IRC | 13:07 | |
| garyloug | Hi all, we have some general questions regarding the kuryr_conf option in kuryr cni (as passed from /opt/stack/cni/conf/10-kuryr.conf). What is the significance of this variable to the controller and the cni? | 13:23 |
| apuimedo | none to the controller | 13:27 |
| apuimedo | to the cni it is just to specify where to find the kuryr config file | 13:27 |
| *** yamamoto_ has joined #openstack-kuryr | 13:28 | |
| *** yamamoto_ has quit IRC | 13:28 | |
| *** yamamoto has quit IRC | 13:28 | |
| garyloug | Ok, for example, if we want to run 2 kuryr controllers e.g one with nested macvlan and another with our DPDK PoC plugin, is there a way we can configure kuryr cni to use different config files? | 13:31 |
| apuimedo | garyloug: with or without multus? | 13:32 |
| garyloug | :-) for a PoC we are working with Multus for now | 13:32 |
| apuimedo | garyloug: what kind of config does the DPDK plugin use? | 13:37 |
| * apuimedo wondering if we could do with just a single cni conf | 13:37 | |
| apuimedo | since the vif info will come from the controllers | 13:37 |
| apuimedo | in the future I want it to be possible with a single controller as well | 13:37 |
| apuimedo | once we have multi network support | 13:38 |
| kural | DPDK is only do rx/tx packets, for kubernetes we have to send back the IPAM information, we planning to have kuryr nested MACVLAN as the control/Mgnt plane and nested-DPDK as the data plane | 13:40 |
| apuimedo | kural: yeah, yeah. I get that | 13:40 |
| apuimedo | what's the timeline to this PoC? | 13:40 |
| kural | For the 1st phase, we planning to use Multus to configure with nested MACVLAN and nested DPDK.. | 13:40 |
| kural | We will contribute to the Multi-network PoC as well as. We about to finish the kubernetes network native support..But Multi-network support in Kuryr we need modification in Kuryr controller. For the time being, we planning to run two controller and configure the pod network them using Multus.. | 13:43 |
| kzaitsev_ws | garyloug: kural: if you guys are using multus you'll be able to use different kuryr.conf files. so I don't really get the question =) | 13:43 |
| kzaitsev_ws | what would be the difference between them btw? | 13:43 |
| kzaitsev_ws | kural: have you looked at my patches, that add multi-VIF support btw? | 13:44 |
| garyloug | We tried to use 2 different config files with Multus - 1 uses the nested-macvlan plugin, the other uses the DPDK PoC. However, it looks like Kuryr uses the usual /etc/kuryr/kuryr.conf in the controller no matter what config Multus passes to it. | 13:46 |
| kural | <kzaitsev_ws> : I yet to check it:( | 13:46 |
| *** yamamoto has joined #openstack-kuryr | 13:47 | |
| kzaitsev_ws | kural: I would really appreciate it. and also it's kind of already working | 13:47 |
| apuimedo | kzaitsev_ws: yeah, I thought the same | 13:48 |
| kural | Ok cool, we will apply your patches that support Multi-VIF in the controller.. | 13:48 |
| apuimedo | garyloug: what were you passing with multus? the path to different /etc/kuryr/kuryr-x.conf or directly kuryr config ini content? | 13:49 |
| garyloug | the path to a new conf file with a different pod_vif_driver | 13:51 |
| *** gouthamr has joined #openstack-kuryr | 13:51 | |
| garyloug | You think this should work? | 13:51 |
| apuimedo | it should | 13:52 |
| apuimedo | let me confirm | 13:52 |
| openstackgerrit | Antoni Segura Puimedon proposed openstack/kuryr-kubernetes master: Use a separate network for services subnet https://review.openstack.org/486134 | 13:53 |
| garyloug | ok, thanks :-) | 13:53 |
| apuimedo | garyloug: it should work | 13:54 |
| apuimedo | in the openshift integration what I do is: | 13:54 |
| apuimedo | /etc/kuryr/kuryr-controller.conf for the controller | 13:54 |
| apuimedo | and /etc/kuryr/cni.conf for the CNI part | 13:54 |
| apuimedo | and in /opt/stack/cni/conf/10-kuryr.conf | 13:55 |
| apuimedo | the kuryr_conf parameter | 13:55 |
| apuimedo | points to the /etc/kuryr/cni.conf | 13:55 |
| apuimedo | and it is confirmed to work | 13:55 |
| *** kzaitsev_ws has quit IRC | 13:56 | |
| garyloug | did you manually create the controller and cni conf files? | 13:57 |
| apuimedo | garyloug: I have a jinja template in my ansible | 13:58 |
| apuimedo | http://paste.openstack.org/show/616170 | 13:58 |
| kural | apuimedo: I am new to Kuryr, we like to understand the difference between the the /etc/kuryr/kuryr-controller.conf and /etc/kuryr/cni.conf. We running kuryr-kubernetes --config-file /etc/kuryr/kuryr-controller.conf | 14:03 |
| kural | with pod_vif_driver nested-macvlan | 14:04 |
| apuimedo | kural: the difference is that the cni part needs to know nothing about the [neutron] section, basically | 14:04 |
| kural | then created the /etc/kuryr/dpdk-cni.conf with pod_vif_driver nested-dpdk and /etc/kuryr/macvlan-cni.conf with pod_vif_driver nested-macvlan | 14:05 |
| kural | using CRD to create network objects and assign them in the pod spec using Multus | 14:05 |
| kural | multus invoke kuryr-cni with different conf file information. | 14:06 |
| kural | in our case only MACVLAN is invoked twice, different pod_vif_driver from cni.conf is ignored by kuryr controller | 14:07 |
| apuimedo | kural: as I said, multus should just invoke kuryr cni with the json conf pointing to different kuryr config files in the kuryr_conf section | 14:08 |
| garyloug | ok, thanks Toni, we will try again with separate configs | 14:09 |
| apuimedo | pod_vif_driver only matters for the controller IIRC | 14:10 |
| apuimedo | the cni part gets the vif kind from the vif annotation | 14:10 |
| kural | Thanks Toni, we will try it out | 14:19 |
| kural | ;) | 14:19 |
| apuimedo | kural: garyloug: report back if you find any issue :-) | 14:19 |
| apuimedo | and if you want to discuss the next phase over bluejeans with Kuryr native multi networks let me know | 14:19 |
| * apuimedo should update the spec | 14:20 | |
| *** mattmceuen has joined #openstack-kuryr | 14:22 | |
| *** atoth has quit IRC | 14:56 | |
| *** aojea has quit IRC | 15:30 | |
| *** aojea has joined #openstack-kuryr | 15:30 | |
| *** aojea has quit IRC | 15:35 | |
| *** yamamoto has quit IRC | 15:44 | |
| *** yamamoto has joined #openstack-kuryr | 15:44 | |
| *** yamamoto has quit IRC | 15:49 | |
| openstackgerrit | Antoni Segura Puimedon proposed openstack/kuryr-kubernetes master: Use a separate network for services subnet https://review.openstack.org/486134 | 15:56 |
| openstackgerrit | Antoni Segura Puimedon proposed openstack/kuryr-kubernetes master: Use a separate network for services subnet https://review.openstack.org/486134 | 16:13 |
| *** egonzalez has quit IRC | 16:16 | |
| *** neiljerram has quit IRC | 16:39 | |
| *** garyloug has quit IRC | 16:46 | |
| *** yamamoto has joined #openstack-kuryr | 16:50 | |
| *** kural_ has quit IRC | 16:54 | |
| *** kural has quit IRC | 16:54 | |
| *** vikasc has quit IRC | 17:03 | |
| *** yamamoto has quit IRC | 17:05 | |
| *** atoth has joined #openstack-kuryr | 17:11 | |
| *** vikasc has joined #openstack-kuryr | 17:16 | |
| openstackgerrit | Antoni Segura Puimedon proposed openstack/kuryr-kubernetes master: Use a separate network for services subnet https://review.openstack.org/486134 | 17:20 |
| *** atoth has quit IRC | 18:07 | |
| *** ajo has quit IRC | 18:18 | |
| *** tonanhngo has joined #openstack-kuryr | 19:07 | |
| *** aojea has joined #openstack-kuryr | 20:01 | |
| *** rwallner has quit IRC | 20:46 | |
| *** rwallner has joined #openstack-kuryr | 20:48 | |
| *** rwallner_ has joined #openstack-kuryr | 20:49 | |
| *** rwallner has quit IRC | 20:53 | |
| *** rwallner_ has quit IRC | 20:54 | |
| *** gouthamr has quit IRC | 21:14 | |
| *** mestery has quit IRC | 21:40 | |
| *** mestery_ has joined #openstack-kuryr | 21:40 | |
| *** mestery_ is now known as mestery | 21:42 | |
| *** kural has joined #openstack-kuryr | 21:44 | |
| *** kural_ has joined #openstack-kuryr | 21:44 | |
| *** kural_ has quit IRC | 22:32 | |
| *** kural has quit IRC | 22:32 | |
| *** mattmceuen has quit IRC | 22:49 | |
| *** aojea has quit IRC | 22:51 | |
| *** aojea has joined #openstack-kuryr | 22:52 | |
| *** aojea has quit IRC | 22:56 | |
| *** gouthamr has joined #openstack-kuryr | 22:59 | |
| *** mattmceuen has joined #openstack-kuryr | 23:42 | |
| *** mattmceuen has quit IRC | 23:52 | |
| *** gouthamr has quit IRC | 23:52 | |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!