Thursday, 2017-11-23

*** aojea has joined #openstack-kuryr00:00
*** salv-orlando has quit IRC00:02
*** aojea has quit IRC00:05
*** hongbin has quit IRC00:17
*** salv-orlando has joined #openstack-kuryr00:59
*** aojea has joined #openstack-kuryr01:01
*** aojea has quit IRC01:06
*** salv-orlando has quit IRC01:10
*** yamamoto has quit IRC01:52
*** yamamoto has joined #openstack-kuryr01:57
*** yamamoto has quit IRC01:57
*** aojea has joined #openstack-kuryr02:02
*** salv-orlando has joined #openstack-kuryr02:07
*** aojea has quit IRC02:07
*** caowei has joined #openstack-kuryr02:09
*** wangbo has joined #openstack-kuryr02:09
*** salv-orlando has quit IRC02:13
*** pmannidi has quit IRC02:19
*** pmannidi has joined #openstack-kuryr02:20
*** yamamoto has joined #openstack-kuryr02:25
*** yamamoto has quit IRC02:28
*** c00281451 has quit IRC02:32
*** c00281451 has joined #openstack-kuryr02:32
*** wangbo has quit IRC03:02
*** aojea has joined #openstack-kuryr03:04
*** salv-orlando has joined #openstack-kuryr03:09
*** aojea has quit IRC03:09
*** wangbo has joined #openstack-kuryr03:13
*** salv-orlando has quit IRC03:14
*** hongbin has joined #openstack-kuryr03:21
*** janonymous has joined #openstack-kuryr03:23
*** yamamoto has joined #openstack-kuryr03:29
*** yamamoto has quit IRC03:32
*** yamamoto has joined #openstack-kuryr03:32
*** yamamoto has quit IRC03:47
*** yamamoto has joined #openstack-kuryr03:50
*** aojea has joined #openstack-kuryr04:05
*** aojea has quit IRC04:09
*** salv-orlando has joined #openstack-kuryr04:10
*** salv-orlando has quit IRC04:16
*** wangbo has quit IRC04:41
*** caowei has quit IRC04:49
*** hongbin has quit IRC04:56
*** aojea has joined #openstack-kuryr05:05
*** aojea has quit IRC05:10
*** salv-orlando has joined #openstack-kuryr05:12
*** salv-orlando has quit IRC05:18
*** wangbo has joined #openstack-kuryr05:19
*** mspreitz has joined #openstack-kuryr05:23
*** mspreitz has left #openstack-kuryr05:24
*** caowei has joined #openstack-kuryr05:38
*** salv-orlando has joined #openstack-kuryr05:57
*** salv-orlando has quit IRC05:57
*** salv-orlando has joined #openstack-kuryr05:57
*** yamamoto has quit IRC05:59
*** yamamoto has joined #openstack-kuryr06:05
*** aojea has joined #openstack-kuryr06:06
*** yamamoto has quit IRC06:09
*** threestrands has quit IRC06:10
*** aojea has quit IRC06:11
*** gcheresh has joined #openstack-kuryr06:12
*** yboaron_ has joined #openstack-kuryr06:13
*** gcheresh has quit IRC06:14
*** kiennt26 has joined #openstack-kuryr06:55
irenabdulek: hi06:57
irenabcan you please check and abandon the https://review.openstack.org/#/c/480028/ ?06:59
janonymousirenab:  i'll do it if that's okay :)07:06
*** caowei_ has joined #openstack-kuryr07:07
*** aojea has joined #openstack-kuryr07:07
irenabjanonymous: sure. thanks a lot!07:09
*** yamamoto has joined #openstack-kuryr07:09
janonymousirenab:  :)07:10
irenabthe alternative patch was merged yesterday, so I guess this one can be abandon07:10
*** caowei has quit IRC07:11
*** caowei_ is now known as caowei07:11
*** aojea has quit IRC07:11
*** yamamoto has quit IRC07:18
*** aojea has joined #openstack-kuryr07:25
*** vikasc has quit IRC07:29
*** aojea has quit IRC07:32
*** yamamoto has joined #openstack-kuryr07:32
*** wangbo has quit IRC07:41
openstackgerritMerged openstack/kuryr-kubernetes master: Add CNI daemon Tempest job  https://review.openstack.org/51961207:42
*** yamamoto has quit IRC07:43
*** wangbo has joined #openstack-kuryr07:43
*** pcaruana has joined #openstack-kuryr07:45
*** wangbo has quit IRC07:54
*** juriarte has joined #openstack-kuryr07:55
openstackgerritMichał Dulko proposed openstack/kuryr-kubernetes master: Fix kubelet retries issues  https://review.openstack.org/51840408:27
*** wangbo has joined #openstack-kuryr08:27
*** jistr|off|trng is now known as jistr08:28
*** pmannidi has quit IRC08:32
*** yboaron_ has quit IRC08:34
*** yamamoto has joined #openstack-kuryr08:43
*** gcheresh has joined #openstack-kuryr08:50
openstackgerritBerezovsky Irena proposed openstack/kuryr-kubernetes master: Kubernetes Network Policy support Spec  https://review.openstack.org/51923908:51
*** yamamoto has quit IRC08:51
*** garyloug has joined #openstack-kuryr08:55
*** garyloug has quit IRC08:56
*** yamamoto has joined #openstack-kuryr09:17
apuimedodulek: are you sure https://review.openstack.org/#/c/518404/5/kuryr_kubernetes/cni/binding/bridge.py is right?09:22
apuimedoit seems to me like it means09:22
apuimedothat we are creating a second veth pair09:22
apuimedoinside the container netns09:22
apuimedothen when we try to move one half to the host, we see there's already one there09:23
apuimedoif we ignore the exception, doesn't it mean that we'll end up with both the container and the host side in the container netns09:23
apuimedoplus a now useless veth on the host side?09:23
*** yboaron_ has joined #openstack-kuryr09:23
dulekapuimedo: Let me look. It was a while when we discussed that.09:24
apuimedo;-)09:24
*** yamamoto has quit IRC09:25
*** aojea has joined #openstack-kuryr09:29
dulekapuimedo: What you're saying makes sense. So correct behavior would be to delete host_ifname on the host side in case of exception?09:32
*** wangbo has quit IRC09:32
apuimedodulek: will we still have a process/thread advancing the previous attempt or not?09:32
dulekapuimedo: If that occurs? We should not have one.09:33
*** aojea has quit IRC09:33
dulekapuimedo: I mean this can occur when there was an exception *after* this veth pair was created and h_iface moved to host namespace.09:34
apuimedoso yes, in that case you should kill the previous veth with fire09:34
apuimedoand brimstone09:34
dulekapuimedo: Maybe those issues should be fought from the other way - I mean, why disconnect isn't deleting the pair?09:35
dulekIn case of failure kubelet will make sure to send DEL request.09:35
apuimedodulek: maybe they should be fought both ways09:35
apuimedoyou can never be too sure :-)09:36
apuimedoso better delete it from both places09:36
apuimedoI prefer to fail to delete because it already does not exist09:36
apuimedothat end up failing an addition because a delete was not executed09:36
*** yamamoto has joined #openstack-kuryr09:37
dulekapuimedo: Makes sense, we're not really guaranteed that it will be ADD-DEL-ADD.09:38
dulekIt could be ADD1-ADD2-DEL1.09:38
apuimedodulek: precisely09:39
apuimedo:-)09:39
dulekapuimedo: Still, veth should go away on DEL, isn't it?09:40
apuimedodulek: it should09:40
dulekAll right then!09:40
apuimedoIIRC, when kubelet deletes the container the veth side that was there gets deleted09:41
apuimedoin turn deleting the host side09:41
apuimedothat that is why we don't delete it09:41
apuimedobut if for some reason kubelet is taking its time with the DEL...09:41
apuimedowell, then all hell breaks loose09:41
apuimedoand you end up in the error you are trying to fix09:41
*** vikasc has joined #openstack-kuryr09:42
*** yamamoto has quit IRC09:43
*** vikasc has quit IRC10:00
*** vikasc has joined #openstack-kuryr10:13
*** garyloug has joined #openstack-kuryr10:13
*** aojea has joined #openstack-kuryr10:30
*** aojea has quit IRC10:35
*** wangbo has joined #openstack-kuryr10:35
*** yamamoto has joined #openstack-kuryr10:37
*** caowei has quit IRC10:37
*** kiennt26 has quit IRC10:43
*** wangbo has quit IRC11:05
dulekltomasbo: Ping, port pools.11:23
ltomasbodulek, tell me11:24
dulekltomasbo: Just a few questions.11:24
dulekltomasbo: So currently adding ports to pool and removing them is triggered manually?11:24
ltomasbono, it is automatically done11:25
ltomasboyou can adjust some variables11:25
ltomasboand it will re-populate a pool when the number of remaining ports is below X11:25
ltomasboand you can set a max number of ports per pool11:25
ltomasboso that when the pod is removed, the ports goes back to the pool or gets deleted11:26
ltomasbowhat is not there (yet) is the initial pre-population11:26
ltomasboi.e., first time you use it, it will generate X ports while allocating the first pod11:26
dulekltomasbo: Where's that code? I've looked through controller.managers.pool and it haven't noticed it.11:26
ltomasbodulek, ^^11:26
ltomasbothat code is in the vif_pool driver11:27
dulekAh, so everything is in VIF handler?11:27
ltomasbocontroller/drivers/vif_pool.py11:27
ltomasbothere you have 2 classes, well 411:27
ltomasbono op class11:27
ltomasbobasic one11:27
ltomasboand then the pool for the baremetal and the one for the nested case11:28
dulekltomasbo: Okay and those react to events from Watcher?11:28
*** wangbo has joined #openstack-kuryr11:28
ltomasbobasically, whenever the vif_handler trigger the release_vif or request_vif action11:28
ltomasbothe vif_pool makes sure you get one port from the pool11:29
dulekltomasbo: Okay, so vif_pool driver is superset of vif driver abilities?11:29
ltomasboand if there are not enough remainin ports (or no ports at all) it triggers a re-population action to create X amount of ports in a bulk request11:29
ltomasboyep, pretty much11:29
dulekltomasbo: Additionaly it manages the pool.11:29
dulekOkay, makes perfect sense! :)11:30
ltomasboin fact, we tell vif_pool to use the required vif_driver11:30
dulekAnd NoopVIFPool driver is the one used with configurations without pools?11:31
dulekltomasbo: ?11:32
ltomasboyes11:32
ltomasboif you set it to noop (which is the default)11:32
ltomasboit will mainly call the request_vif from vif_driver11:32
ltomasboso, it does nothing, just pass the call11:32
dulekltomasbo: Awesome. Thanks for explaining. It's pretty nice design.11:32
ltomasbothanks dulek!11:33
ltomasbodulek, what are you at now?11:33
ltomasbodulek, trying to move them to the CNI side?11:33
dulekltomasbo: Not whole pool management, but only the port choice.11:33
dulekltomasbo: The idea is that controller populates and assigns pools to nodes.11:34
dulekltomasbo: And then it's nodes who do the choice, annotate the VIF, etc.11:34
ltomasbook11:35
dulekltomasbo: It should scale better in situations when we have a single controller.11:35
ltomasbojust note that there is already a pool per node, actually a pool per (node-sg-projectid)11:35
ltomasboyes yes, I agree on that!11:35
ltomasboand it was in my to do list, but had to focus on other stuff11:35
dulekltomasbo: Oh, nice, good to know pools are already separated.11:38
*** yamamoto_ has joined #openstack-kuryr11:45
*** yamamoto has quit IRC11:48
apuimedoand it should survive controller reboots if the pools are full enough11:55
*** yamamoto_ has quit IRC12:02
irenabdulek: please drft your design ideas to dicuss since it will have an impact once Network Policies are applied. Lets try to make sure the alignment between these features12:31
*** aojea has joined #openstack-kuryr12:31
*** aojea has quit IRC12:36
dulekirenab: Hm… I'll definitely draft it, as it's too much for my imagination to handle. :)12:38
dulekirenab: But I'm not sure how it'll affect policies.12:38
dulekirenab: I mean - policies are basically SGs applied to ports/VIFs. It shouldn't matter who is writing the VIF annotation.12:39
irenabdulek: policies are applied as security groups on pod-ports. Probably most of the work will have to be done at the controller level, but I am concerned how it will work with the pools, when pools are arranged accordng to SGs12:40
dulekirenab: Pools need to be arranged according to SGs?12:41
dulekirenab: I thought it's more of a port-SG mapping.12:41
irenabltomasbo: Am I wrong about it? ^^12:42
* dulek wouldn't be surprised if that's his misunderstanding.12:42
ltomasbopools right now are differentiated by a key made out of: node/vm + sgs + projectId12:44
ltomasboso, not sure how the pools will be handled12:44
openstackgerritBerezovsky Irena proposed openstack/kuryr-kubernetes master: Kubernetes Network Policy support Spec  https://review.openstack.org/51923912:44
ltomasboif there is a new rule being added to the SG, it should be independent12:44
ltomasboas the pool will be unmodified12:44
irenabltomasbo: it will be not only rules, but new groups too12:45
ltomasboif it is adding a new SGs, then we will have a new pool matching that SG12:45
irenabI think leyal is back from PTO next week, so maybe we can discuss in more details during the weekly12:45
ltomasboanyway, new SGs will be added to running pods12:45
dulekltomasbo: Why there's a pool per node-sg-projectId?12:46
ltomasboso, what I expect to happen is that, a new pool will be created for a new pod after the SGs had been modified12:46
ltomasbodulek, we want to skip calls to neutron, and if we need to take a port from the pool and apply the proper SG ids, then we need to call neutron again12:46
dulekltomasbo: Ouch. That makes stuff much harder.12:47
irenabltomasbo: right to optimize more pods to be created that match the policy12:47
ltomasbobesides, it was just a configuration sg until now, it was not changing over time12:47
dulekapuimedo: Interesting read. ^12:48
apuimedowhat is an interesting read?12:48
dulekltomasbo: So whole deployment had a single SG, or was is configured on per-node basis?12:48
* apuimedo lunch12:48
ltomasboI'm going to have lunch now, but I agree with irenab. It is worth to have a discussion about this12:48
ltomasbodulek, until now, you define the sg in the kuryr.conf12:48
ltomasboso it is not changing between pods12:48
yboaron_I'll happy to join this discussion12:49
dulekltomasbo: Okay, nice.12:49
ltomasbobut I did include the SG into the key, to ensure the right SG are still on the ports12:49
dulekapuimedo: Discussion when ltomasbo chimed in. Basically daemon-side-vif-choice + policies support.12:49
dulekltomasbo: In case someone reconfigured it, right?12:49
irenabltomasbo:  apuimedo reno support patch https://review.openstack.org/#/c/522393/12:59
*** wangbo has quit IRC13:00
*** yamamoto has joined #openstack-kuryr13:02
*** yamamoto has quit IRC13:09
dulekirenab: I've just commented with an issue on the patch.13:19
*** salv-orlando has quit IRC13:21
*** salv-orlando has joined #openstack-kuryr13:22
apuimedothanks13:25
*** salv-orlando has quit IRC13:27
openstackgerritBerezovsky Irena proposed openstack/kuryr-kubernetes master: Add Reno release notes support to Kuryr-Kubernetes  https://review.openstack.org/52239313:28
*** yboaron has joined #openstack-kuryr13:29
*** yboaron_ has quit IRC13:31
irenabapuimedo: dmellado : Seems we will need to add reno job to CI13:31
*** aojea has joined #openstack-kuryr13:32
dmelladoirenab: yeah13:32
dmelladosaw your patch!13:32
dmelladowill review later13:32
*** juriarte has quit IRC13:32
*** aojea has quit IRC13:37
apuimedoDoes it happen to any of you that sometimes the screen goes blank after not being in use and can't turn back on without rebooting13:39
apuimedo?13:39
dulekapuimedo: Thinkpad P-90 with a dock? There's a thread on memo-list about that.13:49
*** juriarte has joined #openstack-kuryr13:51
*** janki has joined #openstack-kuryr13:53
apuimedodulek: nope, I have t460s13:58
dulekapuimedo: T470S works fine for me with Fedora 27.14:01
*** dougbtv__ has quit IRC14:02
apuimedoso you can swap it with mine :-)14:02
apuimedoxD14:02
dulekBut I'm already used to it…14:02
*** janki has quit IRC14:05
*** yamamoto has joined #openstack-kuryr14:05
*** janki has joined #openstack-kuryr14:06
*** yboaron_ has joined #openstack-kuryr14:07
*** juriarte has quit IRC14:08
*** yamamoto has quit IRC14:10
*** yboaron has quit IRC14:11
apuimedo:P14:12
apuimedoI wouldn't trade it, it's already too full of sunflower seed parts14:13
apuimedonothing helps review patches like salted sunflower seeds14:14
apuimedook, maybe a bigger screen14:14
apuimedobut apart from that...14:14
*** salv-orlando has joined #openstack-kuryr14:22
*** juriarte has joined #openstack-kuryr14:23
openstackgerritMichał Dulko proposed openstack/kuryr-kubernetes master: Add OpenShift gates  https://review.openstack.org/52107714:25
apuimedoirenab: still not gone for the weekend?14:27
irenabapuimedo: sick today, so on and off working from home14:28
irenabanything specific to check?14:28
apuimedoirenab: It would be great if you have some time to review this openshift gates patch14:28
irenabapuimedo: sure, asap14:29
irenabbut seems zuul posted -114:29
dulekirenab: Uh, oh, my bad.14:30
dulekirenab: Just a second…14:30
openstackgerritMichał Dulko proposed openstack/kuryr-kubernetes master: Add OpenShift gates  https://review.openstack.org/52107714:31
*** salv-orlando has quit IRC14:31
*** aojea has joined #openstack-kuryr14:33
*** aojea has quit IRC14:37
irenabdulek: few comments posted14:44
dmelladoirenab: dulek14:45
dmelladolate to the party14:45
dmelladoit's been a busy day :\14:45
dmelladoirenab: +1 to the renos14:45
dmelladoI don't really mind merging the patch and then adding a gate + a release note14:45
irenabdmellado: fine with me14:46
apuimedoltomasbo: don't be lazy and add the release note to https://review.openstack.org/#/c/519704/14:49
apuimedoI want to merge it!14:49
ltomasbo:D14:49
apuimedo(you'll have to rebase after reno support gets merged in about 5mins)14:49
ltomasboI'll do later today, I thought I had to wait for the reno support...14:49
apuimedoltomasbo: I just clicked w+114:49
apuimedo:-)14:49
openstackgerritMichał Dulko proposed openstack/kuryr-kubernetes master: Add OpenShift gates  https://review.openstack.org/52107714:50
ltomasboapuimedo, and that is being lazy!!! xD it is not even merged yet! :D14:50
ltomasboapuimedo, jokes aside, I'll do after our meeting14:50
apuimedoltomasbo: xD14:51
apuimedoI'm just in a merge spree14:51
irenabapuimedo: if you are there, please check few trivial patches I +2 this morning14:52
apuimedoirenab: I think I just did, but let me know if I missed any14:53
*** janki has quit IRC14:53
irenabapuimedo: https://review.openstack.org/#/c/509950/14:56
apuimedoirenab: thanks!15:04
*** yamamoto has joined #openstack-kuryr15:06
openstackgerritMerged openstack/kuryr-kubernetes master: Add Reno release notes support to Kuryr-Kubernetes  https://review.openstack.org/52239315:07
openstackgerritMerged openstack/kuryr-kubernetes master: Use OVS_BRIDGE variable for bridge configuration  https://review.openstack.org/51610615:07
apuimedoyay!15:07
apuimedorelease notes incoming!15:07
*** yamamoto has quit IRC15:12
*** salv-orlando has joined #openstack-kuryr15:13
openstackgerritMerged openstack/kuryr-kubernetes master: Remove SCREEN_LOGDIR from devstack  https://review.openstack.org/50995015:23
openstackgerritMichał Dulko proposed openstack/kuryr-kubernetes master: Add OpenShift gates  https://review.openstack.org/52107715:28
*** aojea has joined #openstack-kuryr15:34
*** aojea has quit IRC15:41
*** gcheresh has quit IRC15:42
*** yamamoto has joined #openstack-kuryr16:08
*** pcaruana has quit IRC16:09
*** yamamoto has quit IRC16:13
dmelladolol16:45
dmelladoltomasbo: today's meetings day16:45
* dmellado sighs16:46
dmelladowhenever I get to start working on something, I've got to leave it for a meeting16:46
dmelladoxD16:46
*** yamamoto has joined #openstack-kuryr16:52
openstackgerritLuis Tomas Bolivar proposed openstack/kuryr-kubernetes master: Add oslo_cache to default_subnet driver  https://review.openstack.org/51970416:53
ltomasbodmellado, I know the feeling...16:54
ltomasboapuimedo, reno added (first one in my life...)16:54
dmelladoltomasbo: please don't forget the raindeer16:54
ltomasbodmellado, ??16:55
dmelladoreno16:55
* dmellado hides, too many meetings16:55
dmelladoxD16:55
ltomasboxD16:55
dmelladoand it was REINdeer16:55
ltomasboreindeer...16:55
dmelladomy bad16:55
dmelladoyep16:55
dmelladowhen I'm tired I also do usually type openshuft, you know...16:56
*** yboaron_ has quit IRC17:01
apuimedodmellado: I tend to type opensihft17:03
*** yamamoto_ has joined #openstack-kuryr17:09
*** yamamoto has quit IRC17:13
*** salv-orlando has quit IRC17:26
*** salv-orlando has joined #openstack-kuryr17:27
*** juriarte has quit IRC17:31
*** salv-orlando has quit IRC17:31
*** gcheresh has joined #openstack-kuryr17:36
*** janonymous has quit IRC17:37
*** salv-orlando has joined #openstack-kuryr17:37
*** salv-orlando has quit IRC17:37
*** salv-orlando has joined #openstack-kuryr17:37
*** garyloug has quit IRC17:53
*** yamamoto_ has quit IRC18:17
*** gcheresh has quit IRC18:34
openstackgerritOpenStack Proposal Bot proposed openstack/kuryr-kubernetes master: Updated from global requirements  https://review.openstack.org/50978118:46
*** yamamoto has joined #openstack-kuryr19:17
*** yamamoto has quit IRC19:25
*** yboaron_ has joined #openstack-kuryr19:36
*** aojea has joined #openstack-kuryr19:38
*** aojea has quit IRC19:42
*** gcheresh has joined #openstack-kuryr20:17
*** gcheresh has quit IRC20:33
*** yamamoto has joined #openstack-kuryr20:49
*** yamamoto has quit IRC20:49
*** salv-orlando has quit IRC20:56
*** salv-orlando has joined #openstack-kuryr20:57
*** yamamoto has joined #openstack-kuryr20:59
*** yamamoto has quit IRC20:59
*** salv-orlando has quit IRC21:01
*** hongbin has joined #openstack-kuryr21:16
*** aojea has joined #openstack-kuryr21:23
*** gcheresh has joined #openstack-kuryr21:34
*** gcheresh has quit IRC21:38
*** yboaron_ has quit IRC21:47
*** salv-orlando has joined #openstack-kuryr21:58
*** yamamoto has joined #openstack-kuryr22:00
*** salv-orlando has quit IRC22:03
*** yamamoto has quit IRC22:07
*** salv-orlando has joined #openstack-kuryr22:59
*** yamamoto has joined #openstack-kuryr23:03
*** salv-orlando has quit IRC23:05
*** yamamoto has quit IRC23:08
*** aojea has quit IRC23:11

Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!