| *** yuvalb has quit IRC | 00:06 | |
| *** yuvalb has joined #openstack-kuryr | 00:08 | |
| openstackgerrit | Hongbin Lu proposed openstack/kuryr-libnetwork: Add prefix to the specified name of subnetpool https://review.openstack.org/426623 | 00:09 |
|---|---|---|
| openstackgerrit | Hongbin Lu proposed openstack/kuryr-libnetwork: [WIP] Support creating from existing subnetpool https://review.openstack.org/426595 | 00:09 |
| *** limao_ has quit IRC | 00:10 | |
| *** dims_ has joined #openstack-kuryr | 00:10 | |
| *** dims has quit IRC | 00:11 | |
| *** limao has joined #openstack-kuryr | 00:11 | |
| *** hongbin has quit IRC | 00:14 | |
| *** limao_ has joined #openstack-kuryr | 00:21 | |
| *** limao has quit IRC | 00:22 | |
| *** limao_ has quit IRC | 00:26 | |
| *** limao has joined #openstack-kuryr | 00:26 | |
| *** saneax is now known as saneax-_-|AFK | 00:30 | |
| *** limao has quit IRC | 01:08 | |
| *** limao has joined #openstack-kuryr | 01:08 | |
| *** limao has quit IRC | 01:44 | |
| *** limao_ has joined #openstack-kuryr | 01:45 | |
| *** tonanhngo has quit IRC | 01:47 | |
| *** yedongcan has joined #openstack-kuryr | 02:11 | |
| *** tonanhngo has joined #openstack-kuryr | 02:13 | |
| *** yedongcan has quit IRC | 02:15 | |
| *** yedongcan has joined #openstack-kuryr | 02:15 | |
| *** tonanhngo has quit IRC | 02:17 | |
| *** yuanying has quit IRC | 02:21 | |
| janonymous | apuimedo: https://etherpad.openstack.org/p/Dur3bXQTNH , Some pointers | 02:22 |
| *** limao_ has quit IRC | 02:36 | |
| *** limao has joined #openstack-kuryr | 02:52 | |
| openstackgerrit | Dongcan Ye proposed openstack/kuryr: Fix param in get_neutron_subnetpool_name https://review.openstack.org/427533 | 03:00 |
| *** limao has quit IRC | 03:04 | |
| *** yedongcan has left #openstack-kuryr | 03:27 | |
| *** janki has joined #openstack-kuryr | 05:01 | |
| *** tonanhngo has joined #openstack-kuryr | 05:22 | |
| janonymous | apuimedo: https://review.openstack.org/#/c/410609/ | 05:32 |
| janonymous | apuimedo: non-voting jobs pass now | 05:32 |
| *** saneax-_-|AFK is now known as saneax | 06:05 | |
| *** yedongcan has joined #openstack-kuryr | 06:15 | |
| *** pcaruana has joined #openstack-kuryr | 07:19 | |
| *** yamamoto has quit IRC | 07:25 | |
| openstackgerrit | Luis Tomas Bolivar proposed openstack/kuryr: Make segmentation driver testable https://review.openstack.org/427190 | 07:37 |
| *** tonanhngo has quit IRC | 07:40 | |
| *** pmannidi has quit IRC | 07:43 | |
| irenab | ivc_: I am not sure I got the context of your previous commet. Was the patch to make sure the expected exception log causing trouble? | 07:49 |
| apuimedo | thanks janonymous for the patch! Merged | 08:08 |
| janonymous | apuimedo:thanks! | 08:08 |
| apuimedo | irenab: please review https://review.openstack.org/#/c/427190/4 | 08:09 |
| irenab | ok | 08:09 |
| apuimedo | janonymous: thanks for the check on client-python | 08:10 |
| apuimedo | did you check if they support http2 or websockets for streaming events? | 08:11 |
| openstackgerrit | Merged openstack/kuryr-libnetwork: Tls support configurations https://review.openstack.org/410609 | 08:12 |
| janonymous | apuimedo:checking in a sex | 08:16 |
| janonymous | *sec | 08:16 |
| janonymous | ops, sry | 08:17 |
| openstackgerrit | Merged openstack/kuryr-kubernetes: testing: drop zero hashseed https://review.openstack.org/427495 | 08:18 |
| janonymous | apuimedo: https://github.com/kubernetes-incubator/client-python/pull/2/commits/750c9eaba4dd7e692edb4d2d7d842e05452d92a7 | 08:18 |
| irenab | janonymous: Do you feel the k8s client is table eanough? | 08:25 |
| janonymous | irenab:According to code seems like they have major things that needs for kuryr, but personally i feel i requires some time to get production ready | 08:27 |
| janonymous | *it | 08:27 |
| apuimedo | xD | 08:27 |
| * janonymous I donno where to hide | 08:29 | |
| irenab | janonymous: so you would suggest towait before switching to use the k8s clinet? | 08:29 |
| irenab | client | 08:29 |
| *** yuanying has joined #openstack-kuryr | 08:30 | |
| janonymous | irenab: tricky question, but i am in favor of moving to officially supported client but would wait for some time more | 08:31 |
| janonymous | to check if it wont break our existing enviornment | 08:32 |
| irenab | janonymous: is there pipi for the client? | 08:32 |
| irenab | pypi | 08:32 |
| janonymous | yes | 08:32 |
| janonymous | https://pypi.python.org/pypi/kubernetes/1.0.0b1#downloads | 08:33 |
| janonymous | released after i asked on sig-api machinery few time back | 08:33 |
| irenab | janonymous: cool | 08:35 |
| janonymous | maybe in vtg, we can write down everything we need from client in our kuryr-kuberentes code and check those one-by-one | 08:36 |
| irenab | apuimedo: janonymous : sounds like a plan | 08:36 |
| janonymous | yeah, i would say not to jump directly on using that, instead we can make changes on k8s-client in external repo first according to our needs, when that is done we can safely replace out our internal implementation | 08:38 |
| janonymous | more suggestions are welcomed. | 08:39 |
| openstackgerrit | Merged openstack/kuryr: Make segmentation driver testable https://review.openstack.org/427190 | 08:43 |
| *** jchhatbar has joined #openstack-kuryr | 08:46 | |
| *** janki has quit IRC | 08:48 | |
| irenab | janonymous: longer term, using k8s sdk is the proper way to go. I do not think kuryr should have any notexpected usage for the k8s api. | 08:54 |
| irenab | janonymous: but we should make sure that not only functionality but also performance wirse it fits kuryr needs | 08:54 |
| apuimedo | irenab: agreed | 08:54 |
| janonymous | +1 | 08:55 |
| openstackgerrit | Luis Tomas Bolivar proposed openstack/kuryr: Add randomness to the returned vlan_ids https://review.openstack.org/422641 | 08:58 |
| apuimedo | ltomasbo: the testable patch got much better! Thanks! | 08:58 |
| ltomasbo | thank you for the comments! | 08:59 |
| ltomasbo | I just rebased the other patch to see if know unit tests pass | 08:59 |
| ltomasbo | apuimedo, ^^ | 08:59 |
| apuimedo | I saw | 09:00 |
| apuimedo | I'm looking at it now | 09:00 |
| ltomasbo | yep, now they pass unit tests! | 09:04 |
| *** garyloug has joined #openstack-kuryr | 09:08 | |
| irenab | ltomasbo: hi | 09:12 |
| ltomasbo | irenab, hi! | 09:13 |
| irenab | ltomasbo: checking the randomness patch now. | 09:13 |
| ltomasbo | great! | 09:13 |
| irenab | not related tot he change, but general question about the local_vlans | 09:13 |
| irenab | is it sort of the local cache? | 09:14 |
| ltomasbo | yes | 09:14 |
| ltomasbo | these vlans are local to each trunk port | 09:15 |
| ltomasbo | and independent of the vlans used for the subnets | 09:15 |
| ltomasbo | so, each kuryr running inside the trunk VM will have their own local range | 09:16 |
| irenab | it starts with full vlan range. How is it synced with the host? | 09:16 |
| ltomasbo | that is handled by kuryr-libnetwork (or kuryr-kubernetes) | 09:16 |
| ltomasbo | where they check (by calling neutron) what vlans are already in used by that trunk port | 09:17 |
| irenab | and this is passed into allocate_xxx method | 09:17 |
| ltomasbo | and it will be passed in allocated_ids, so that those are removed from the full range in the allocate_segmentation_id call | 09:17 |
| ltomasbo | yep | 09:18 |
| irenab | I am not sureI see the value of the local cache | 09:18 |
| irenab | can you please expain how is it used | 09:18 |
| ltomasbo | not sure I get what is the problem with that (and that was in the patch before anyway) | 09:19 |
| ltomasbo | the idea is that you get a vlan_id from the list of available vlans | 09:19 |
| ltomasbo | and we used the available_local_vlans just to keep which vlans are already taken | 09:20 |
| irenab | ltomasbo: debug purposes? | 09:20 |
| ltomasbo | we don't want to call neutron all the time to check what vlan_ids are available | 09:20 |
| ltomasbo | so we keep track of that to avoid neutron calls | 09:21 |
| ltomasbo | at starting kuryr, we sync with neutron (for the specific VM trunk port) | 09:21 |
| ltomasbo | and then we handle the vlan_ids | 09:21 |
| ltomasbo | is it not for debuging, we need to figure out the vlan_ids as neutron don't give you one by default | 09:22 |
| ltomasbo | https://review.openstack.org/#/c/421880 | 09:22 |
| ltomasbo | I'm trying to get that supported, but didn't get too far | 09:23 |
| irenab | ok. thereis something bothering me here :-), but it is not related to your change anyway | 09:24 |
| ltomasbo | irenab: not sure this answer your question/concern | 09:24 |
| ltomasbo | what do you have in mind to remove the cache? | 09:24 |
| ltomasbo | calling neutron? or simply using allocated_ids? | 09:25 |
| irenab | ltomasbo: we keep neutron as a single source of truth, regardless what is the status in the VM | 09:25 |
| ltomasbo | we still do that | 09:25 |
| ltomasbo | and even if we are not completely in sync at some point | 09:26 |
| ltomasbo | neutron will reject to use an already allocated vlan | 09:26 |
| ltomasbo | for the specific purpose | 09:26 |
| irenab | I mean for the case is ther is some restart of the kuryr service, we loose the cache | 09:26 |
| ltomasbo | the problem here is neutron should give you a default vlan_id by definition, but it is not | 09:26 |
| ltomasbo | yes, but if we restart the kuryr service, then we call neutron at restart time to get the vlans | 09:27 |
| apuimedo | ltomasbo: that's the problem :-) "neutron should give you a default vlan_id by definition, but it is not" | 09:27 |
| apuimedo | any advance in getting the neutron patch in? | 09:27 |
| ltomasbo | https://github.com/openstack/kuryr-libnetwork/blob/master/kuryr_libnetwork/port_driver/drivers/vlan.py#L39:L42 | 09:28 |
| ltomasbo | no, armax said it will not work with ironic | 09:28 |
| ltomasbo | I tried to reach him but didn't get any more info | 09:28 |
| ltomasbo | and I did check ironic a bit, and it seems trunk-ports are not supported at all | 09:28 |
| irenab | ltomasbo: maybe worth to check with Ironic team | 09:29 |
| ltomasbo | so not sure what the problem may be with my patch and ironic | 09:29 |
| ltomasbo | I saw there is a spec to include trunks support | 09:29 |
| ltomasbo | but it is not yet there, so I'm not sure how the patch can break it! | 09:29 |
| apuimedo | ltomasbo: so you are saying that trunk ports are not supported in Ironic yet it is blocked on those grounds? | 09:29 |
| apuimedo | I must be missing something | 09:30 |
| ltomasbo | armax did not get back to me with more info, but that is what I found by looking at trunk ports at ironic | 09:30 |
| apuimedo | ltomasbo: does ironic always use vlans? | 09:30 |
| irenab | apuimedo: As far as I know its eiter flat or VLANs | 09:31 |
| ltomasbo | the think I'm forecasting, is that they plan to support trunk ports by sending the traffic encapsulated as vlan ids | 09:31 |
| ltomasbo | on the wire | 09:31 |
| ltomasbo | so, there is a need of those vlans_ids to be accepted/configured in the physical network switches (ToR) | 09:32 |
| apuimedo | irenab: oh, I had thought maybe with some switches it could do vxlan | 09:32 |
| ltomasbo | but neutron is not configuring physical switches directly | 09:32 |
| apuimedo | but even if it vlan | 09:32 |
| apuimedo | if you get a "physical trunk" port to the host | 09:32 |
| apuimedo | you could run containers in vlans | 09:33 |
| irenab | ltomasbo: it can with ml2 drivers. It waht it does for Arista switches | 09:33 |
| apuimedo | so neutron trunk ports could be supported in that way | 09:33 |
| apuimedo | or nested tags (though that sucks) | 09:33 |
| apuimedo | ltomasbo: I thought ironic did some switch interaction | 09:33 |
| *** saneax is now known as saneax-_-|AFK | 09:33 | |
| apuimedo | I wonder how they do it | 09:34 |
| ltomasbo | apuimedo, I think they do, but not sure how static/dynamic that may be | 09:34 |
| irenab | https://specs.openstack.org/openstack/ironic-specs/specs/not-implemented/ironic-ml2-integration.html | 09:34 |
| apuimedo | it seems it is implemented (despite the spec path) | 09:35 |
| apuimedo | at least the linked bugs are | 09:35 |
| irenab | apuimedo: so it mentions the case with single interface on multiple networks | 09:35 |
| irenab | as TBD | 09:36 |
| apuimedo | yeah | 09:36 |
| apuimedo | irenab: but in any case, it depends on ml2 being able to configure switches, right? | 09:38 |
| irenab | yes | 09:38 |
| ltomasbo | I think the problems was related to not having ovs agent on ironic nodes | 09:39 |
| ltomasbo | so, the vlan-aware-vm, instead of having the tagging only up to br-int | 09:40 |
| ltomasbo | for ironic it will go out of the server tagged | 09:40 |
| ltomasbo | and therefore it needs to use proper tags | 09:40 |
| ltomasbo | and we cannot get the tags by asking neutron about the range and the already used ones | 09:40 |
| ltomasbo | unless the range is the same as the one configured for the ironic nodes/switches | 09:40 |
| ltomasbo | and for that specific case, the vlan tags needs to be the same as the ones used for the subnets | 09:41 |
| ltomasbo | while for normal vlans-aware-vms, that tag is independent of the subnets tags | 09:41 |
| ltomasbo | as it is only local to the OVS trunk-port created for the VM, which will translate it to the subnet vlan at br-int | 09:42 |
| apuimedo | ltomasbo: but there must be some sort of agent, right? | 09:43 |
| irenab | what I don’t understadn is why not enableAPI to support both scenarios. It seems to be the caller responsibility to decide if provide vlan or use neutron to pick itfor him. Like with ml2 type drivers for provider network | 09:43 |
| apuimedo | irenab: ltomasbo's patch just made it optional to automatically get an id | 09:44 |
| apuimedo | blocking it seems needless | 09:44 |
| ltomasbo | yep, but maybe it should be optional just non-ironic scenario... | 09:44 |
| apuimedo | ltomasbo: does that not depend just on ironic usage | 09:44 |
| apuimedo | I don't see what they are being so protective about in this specific case | 09:45 |
| ltomasbo | apuimedo, I'm not sure how ironic works, but I'm guessing configurations happens though pxe | 09:45 |
| ltomasbo | and they just connect to vlans by eth0.101 or something like that | 09:45 |
| apuimedo | ltomasbo: yes, I assume that's the case, but there may be some dynamic aspect | 09:45 |
| ltomasbo | yes, but maybe it is in an ironic controller instead of the ironic nodes itselfs | 09:46 |
| ltomasbo | then again, I do not have enough knowledge about ironic to support any of this, just some reading, never tried myself | 09:46 |
| apuimedo | ltomasbo: I've been wanting to experiment with ironic for some time | 09:48 |
| openstackgerrit | Luis Tomas Bolivar proposed openstack/kuryr: Add randomness to the returned vlan_ids https://review.openstack.org/422641 | 09:51 |
| ltomasbo | me too, I just don't have the hardware... | 09:52 |
| ltomasbo | it would be nice to have ironic support at kuryr when trunk ports are supported | 09:53 |
| apuimedo | exactly | 09:53 |
| ltomasbo | that way we can just have containers veth connected to the nick | 09:53 |
| ltomasbo | s/nick/nic | 09:53 |
| ltomasbo | I think that could be an awesome solution for some telco operators... | 09:54 |
| apuimedo | ltomasbo: you'd probably use either nic vfs or ipvlan/macvlan | 09:58 |
| apuimedo | wouldn't you? | 09:58 |
| apuimedo | (instead of veths) | 09:59 |
| ltomasbo | for ironic? | 10:00 |
| ltomasbo | you want to use the subnet vlan tag | 10:00 |
| ltomasbo | so, why using ipvlan/macvlan? | 10:00 |
| ltomasbo | it is just plug the container on, let's say eth0.101 for subnet on vlan tag 101 | 10:01 |
| ltomasbo | similarly to what we do today for vlans nested containers at VMs with kuryr | 10:02 |
| ltomasbo | but as if the ironic node was the trunk VM | 10:02 |
| ltomasbo | (unless I'm missing something) | 10:03 |
| apuimedo | right | 10:08 |
| apuimedo | ltomasbo: https://github.com/openstack/pyghmi/blob/master/bin/virshbmc | 10:17 |
| apuimedo | there is a fake ipmi server that uses libvirt python module | 10:17 |
| apuimedo | maybe it can be used to try ironic | 10:17 |
| apuimedo | :-) | 10:17 |
| ltomasbo | umm | 10:17 |
| ltomasbo | thanks! | 10:17 |
| ltomasbo | I may try it later! | 10:18 |
| *** saneax-_-|AFK is now known as saneax | 10:38 | |
| *** yedongcan has left #openstack-kuryr | 10:45 | |
| openstackgerrit | Merged openstack/kuryr: Add randomness to the returned vlan_ids https://review.openstack.org/422641 | 11:06 |
| *** v1k0d3n has quit IRC | 12:02 | |
| *** jchhatbar_ has joined #openstack-kuryr | 12:31 | |
| *** jchhatbar has quit IRC | 12:34 | |
| *** dougbtv has joined #openstack-kuryr | 12:35 | |
| *** garyloug has quit IRC | 12:48 | |
| *** garyloug has joined #openstack-kuryr | 12:50 | |
| openstackgerrit | Luis Tomas Bolivar proposed openstack/kuryr-kubernetes: Kuryr Kubernetes Resource Manager design reference document https://review.openstack.org/427681 | 13:00 |
| *** limao has joined #openstack-kuryr | 13:36 | |
| *** limao_ has joined #openstack-kuryr | 13:40 | |
| *** limao has quit IRC | 13:41 | |
| *** v1k0d3n has joined #openstack-kuryr | 13:48 | |
| *** limao_ has quit IRC | 14:08 | |
| *** limao has joined #openstack-kuryr | 14:09 | |
| *** limao has quit IRC | 14:10 | |
| *** limao has joined #openstack-kuryr | 14:10 | |
| apuimedo | ivc_: sometimes I'm tempted to use https://github.com/h2o/picohttpparser for the requests | 14:40 |
| apuimedo | btw, review ltomasbo's https://review.openstack.org/#/c/427681/1/doc/source/devref/resource_manager.rst | 14:45 |
| apuimedo | ;-) | 14:45 |
| apuimedo | ltomasbo: just posted some comments | 14:50 |
| *** limao has quit IRC | 14:50 | |
| apuimedo | on to the lbaas patch review | 14:50 |
| *** limao has joined #openstack-kuryr | 14:50 | |
| apuimedo | irenab: ltomasbo: what about https://review.openstack.org/#/c/425040/ | 14:52 |
| apuimedo | ? | 14:52 |
| ivc_ | apuimedo does that mean we also want to move to C instead of Py? xD but seriously, C extensions do not work well with eventlet | 14:54 |
| ltomasbo | apuimedo, thanks! It was a first draft. I will take a look asap | 14:54 |
| apuimedo | ivc_: I know, I just like to have speedup plans for all the parts | 14:55 |
| apuimedo | :P | 14:55 |
| ivc_ | apuimedo now i want to cite donald knuth :) | 14:56 |
| apuimedo | ivc_: he only spoke of premature optimization, not of premature optimization plans | 14:57 |
| apuimedo | :-) | 14:57 |
| apuimedo | so I'm in the clear | 14:57 |
| ivc_ | thats even more premature xD | 14:57 |
| ltomasbo | apuimedo: ahh, not that many comments! :D | 14:57 |
| apuimedo | ivc_: and less of an optimization | 14:58 |
| ivc_ | apuimedo well, we can also talk about 'go' then :) | 15:00 |
| apuimedo | ivc_: it is less performant, but it gets you free client | 15:01 |
| irenab | apuimedo: totally forgot about it, will update later today | 15:09 |
| apuimedo | thanks | 15:10 |
| *** saneax is now known as saneax-_-|AFK | 15:17 | |
| *** hongbin has joined #openstack-kuryr | 15:19 | |
| *** limao_ has joined #openstack-kuryr | 15:34 | |
| *** limao has quit IRC | 15:36 | |
| *** jchhatbar_ has quit IRC | 15:43 | |
| openstackgerrit | Merged openstack/kuryr-libnetwork: Add nested-containers limitations https://review.openstack.org/425040 | 15:49 |
| openstackgerrit | Merged openstack/kuryr-libnetwork: README: fix nested container rendering https://review.openstack.org/425309 | 15:50 |
| *** limao_ has quit IRC | 16:15 | |
| *** tonanhngo has joined #openstack-kuryr | 16:37 | |
| *** tonanhngo has quit IRC | 16:42 | |
| *** tonanhngo has joined #openstack-kuryr | 17:31 | |
| *** tonanhngo has quit IRC | 17:36 | |
| *** v1k0d3n has quit IRC | 17:56 | |
| *** v1k0d3n has joined #openstack-kuryr | 18:00 | |
| *** tonanhngo has joined #openstack-kuryr | 18:20 | |
| *** garyloug has quit IRC | 18:23 | |
| *** tonanhngo has quit IRC | 18:23 | |
| *** tonanhngo has joined #openstack-kuryr | 18:23 | |
| *** pcaruana has quit IRC | 18:28 | |
| *** yuanying has quit IRC | 19:03 | |
| openstackgerrit | Merged openstack/kuryr-libnetwork: Optimize add subnetpool tag https://review.openstack.org/420610 | 19:08 |
| alraddarla | Does libnetwork follow the same versioning of kuryr | 20:40 |
| alraddarla | Also, is this a bug: https://github.com/openstack/kuryr/blob/master/releasenotes/source/conf.py#L61-L63 | 20:44 |
| alraddarla | Shouldn't version and release be flipped? | 20:44 |
| openstackgerrit | Hongbin Lu proposed openstack/kuryr-libnetwork: [WIP] Optimize the search of subnet https://review.openstack.org/427923 | 21:06 |
| openstackgerrit | Hongbin Lu proposed openstack/kuryr-libnetwork: [WIP] Optimize the search of subnet https://review.openstack.org/427923 | 21:24 |
| hongbin | apuimedo: irenab : hey, when you have a chance, would like to get your early feedback on this one: https://review.openstack.org/#/c/427923/ | 21:25 |
| alraddarla | After looking into it, seems as though these have been updated here: https://github.com/openstack-dev/pbr/blob/master/pbr/version.py#L468-L469 | 21:52 |
| alraddarla | I can volunteer to update Kuryr as well. | 21:52 |
| apuimedo | alraddarla: did you check how neutron does it? | 22:38 |
| apuimedo | hongbin: I'll bring it up with irenab tomorrow | 22:40 |
| apuimedo | thanks | 22:40 |
| hongbin | apuimedo: ok | 22:40 |
| *** pc_m has quit IRC | 22:49 | |
| *** pc_m has joined #openstack-kuryr | 22:50 | |
| *** yuanying has joined #openstack-kuryr | 23:07 | |
| *** v1k0d3n has quit IRC | 23:07 | |
| *** pmannidi has joined #openstack-kuryr | 23:14 | |
| *** v1k0d3n has joined #openstack-kuryr | 23:22 | |
| *** yamamoto has joined #openstack-kuryr | 23:24 | |
| *** v1k0d3n has quit IRC | 23:41 | |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!