Tuesday, 2017-08-08

*** pmannidi has joined #openstack-kuryr00:15
*** limao has joined #openstack-kuryr00:41
*** kiennt has joined #openstack-kuryr00:49
*** gouthamr has quit IRC01:02
*** caowei has joined #openstack-kuryr01:27
*** yedongcan has joined #openstack-kuryr01:48
*** mestery has quit IRC02:17
*** mestery has joined #openstack-kuryr02:22
*** yamamoto_ has joined #openstack-kuryr02:46
*** yamamoto has quit IRC02:46
*** hongbin_ has joined #openstack-kuryr02:49
*** hongbin has quit IRC02:49
*** hongbin_ has quit IRC02:49
*** hongbin has joined #openstack-kuryr02:49
*** caowei has quit IRC02:52
*** caowei has joined #openstack-kuryr02:52
*** hongbin has quit IRC03:56
*** limao has quit IRC04:03
*** reedip_afk is now known as reedip04:17
*** caowei has quit IRC04:26
*** caowei has joined #openstack-kuryr04:26
*** limao has joined #openstack-kuryr04:38
*** caowei has quit IRC05:01
*** janki has joined #openstack-kuryr05:03
openstackgerritJaivish Kothari(janonymous) proposed openstack/kuryr-kubernetes master: [Trivial] Doc fix K8S -> K8s  https://review.openstack.org/49166605:13
*** limao has quit IRC05:18
*** limao has joined #openstack-kuryr05:19
openstackgerritJaivish Kothari(janonymous) proposed openstack/kuryr-kubernetes master: [Trivial] Using Similar Format of path like other places in same Doc  https://review.openstack.org/49166805:32
openstackgerritJaivish Kothari(janonymous) proposed openstack/kuryr-kubernetes master: Fix Incorrect Module Reference  https://review.openstack.org/49166905:39
janonymousapuimedo: irenab : Hi, i was thinking if it would be possible to add a section about loadbalancer for users in docs: somewhere in https://docs.openstack.org/kuryr-kubernetes/latest/installation/index.html ..05:50
*** kiennt has quit IRC05:50
*** caowei has joined #openstack-kuryr05:54
*** kiennt has joined #openstack-kuryr06:07
*** kiennt has quit IRC06:21
*** kiennt_ has joined #openstack-kuryr06:22
*** kiennt has joined #openstack-kuryr06:22
*** kiennt_ has quit IRC06:23
*** livelace2 has joined #openstack-kuryr06:23
*** livelace2 has quit IRC06:24
*** livelace has joined #openstack-kuryr06:27
*** pcaruana has joined #openstack-kuryr07:00
apuimedojanonymous: that's a good point07:25
apuimedolivelace: not yet. There's people testing scale and deployment07:26
apuimedobut not yet07:26
apuimedojanonymous: I'll send a doc patch to explain the loadbalancer situation07:26
apuimedoI'll be offline a while working on this07:26
apuimedootherwise I'll get distracted :P07:26
livelaceapuimedo, Could you point to the main problems by now ?07:29
apuimedoI'm on a meeting now. Will write to you later07:29
livelaceapuimedo, Thanks!07:29
apuimedoyou're welcome07:29
janonymousapuimedo: sure :), let me know if i can add anything to it :)07:31
*** kzaitsev_ws has quit IRC07:33
*** pmannidi has quit IRC07:37
*** yamamoto_ has quit IRC07:44
*** yamamoto has joined #openstack-kuryr07:50
apuimedo;-)07:55
apuimedojanonymous: at the very least reviews will be welcome07:55
apuimedofollow-up patches as well07:55
janonymousapuimedo sure! thanks :)08:00
*** Guest14 has joined #openstack-kuryr08:21
*** kzaitsev_ws has joined #openstack-kuryr08:26
janonymouszengchen: hi, you there08:28
zengchenjanonymous:yes08:29
janonymouszengchen: In watch api of k8s-client there is a resource version returned with each object, just passing the resource version of latest object to the watch and the stream will continue from that point.08:29
*** garyloug has joined #openstack-kuryr08:30
zengchenjanonymous:correct08:30
zengchenjanonymous:from my test, that is a truth.08:30
janonymouszengchen: so, why are producer consumer model needed..08:31
*** aojea has joined #openstack-kuryr08:32
zengchenjanonymous:as the design doc described, producer generates events and consumer processes events.08:34
zengchenjanonymous: i think you want to say why not call watch api of k8s-client directly?08:34
janonymouszengchen: yup, i guess you got me08:35
*** limao has quit IRC08:35
*** limao has joined #openstack-kuryr08:36
zengchenjanonymous:you see i also called the list api. when calling watch api failed, then it will call list api with resouce version of 0.08:40
*** limao has quit IRC08:41
zengchenjanonymous:it will not miss any events when some exceptions happened.08:41
apuimedolivelace: the main problems by now is to increase testing of failure cases08:42
apuimedolike reconnection to k8s08:42
janonymouszengchen: oh, i havn't gone through code yet..was going through the diagrams and a question came to my mind..i will  go through code once..08:42
apuimedoalso improve the resource pooling08:42
apuimedoLBs and such08:42
apuimedozengchen: janonymous: In my PoC I was always restarting the watch with resourceversion=008:43
apuimedowhich gives you the current state snapshot and then continues with newer events08:43
apuimedogiving another resourceversion number does not guarantee, IIRC, that you'll get it, if the reconnection took too long time and too many events happened in between08:44
apuimedozengchen: janonymous: I even wrote the HTTP parser in the PoC, so I did the whole stack08:44
*** aojea has quit IRC08:46
livelaceapuimedo, How do think, Openshift on Openstack + Kuryr - is it stable solution ? (I saw your github repo with the Heat template)08:46
livelaceapuimedo, Does it support service network completely ?08:47
livelace*openshift/kubernetes service network08:47
zengchenjanonymous: ok. from the point of user, he does not need to know the producer and consumer, it will work by passing the list and watch method of resource and methods of handling different events to controller.08:47
*** aojea has joined #openstack-kuryr08:47
janonymousapuimedo:  zengchen : i see, i am not very familiar with the issues but seems relevant..and how about instead of implementing in project repo add a code upstream to maintain watch when it breaks ..any thoughts?08:48
apuimedolivelace: We're still working to integrate it in openshift-ansible playbooks08:48
apuimedoI have some fork for it. But a bit more work needs to happen08:48
apuimedowe also need to add support for loadbalancer service type08:49
apuimedoyboaron started working on that patch08:49
*** yamamoto has quit IRC08:49
zengchenapuimedo: great. could send me a link or something else about that?08:49
janonymous+108:49
livelaceapuimedo, It's not a big problem to do it by hand. lbaas for services ?08:49
apuimedolivelace: we do lbaas for services already, but a FIP is missing to be associated with the LB vip for loadbalancer type08:50
apuimedozengchen: you mean about how we were handling disconnects in the PoC?08:50
livelaceapuimedo, What about Ironic ? Is there any initial supports ?08:50
zengchenapuimedo:sorry, actually, i don't know the meaning of 'PoC'.08:51
apuimedolivelace: no. We haven't worked nor planned that08:52
apuimedozengchen: before kuryr-kubernetes started upstream. We did exploratory implementation in python308:52
livelaceapuimedo, After your completion with your version of openshift-ansible, will it pushed back to official openshift-ansible ?08:52
zengchenapuimedo:got it.08:53
apuimedolivelace: first we'll merge in openshift/openshift-ansible-contrib, then to openshift-ansible08:53
apuimedozengchen: https://github.com/midonet/kuryr/blob/k8s/kuryr/raven/raven.py08:54
livelaceapuimedo, When will plan to release a stable version ? Several months, half of year, year ?08:54
zengchenapuimedo:thanks.08:54
apuimedozengchen: the watch code is towards th bottom of that file08:55
apuimedobut I don't see the resourceversion stuff08:55
apuimedomaybe it changed after I left the company08:55
apuimedoI don't know08:55
apuimedoah no08:55
apuimedoit is there!08:55
apuimedoself._resource_version08:55
zengchenapuimedo:ok, i will rand it.08:56
zengchenapuimedo:s/rand/read08:56
apuimedolivelace: for openshift?08:56
livelaceapuimedo, Yes08:56
janonymousapuimedo: Please have a look at client patch: https://review.openstack.org/#/c/454555/ i guess it is ready for reviews now, we can add enhancements to it08:56
apuimedoit will probably take more than 6 months to become a product08:56
apuimedoit will work before that08:56
apuimedobut "support" is a stong word that we are very cautious with at RH08:57
apuimedo:-)08:57
livelacekube is a great tool, but shift is a more convenient tool08:57
janonymouszengchen: apuimedo : i got the whole idea about the issues and will revisit your patches again zengchen :)08:57
apuimedolivelace: agreed08:57
zengchenjanonymous: actually, that design is not mine. I just translate it from client-go to python and do some simplification.08:58
apuimedozengchen: janonymous: If we find the upstream k8s client lacking in reconnection and exception scenarios we should do a mix of improving the upstream client and error control08:58
apuimedoI'd suggest we document the places where it lacks and do something about it08:58
* apuimedo misses the python3 version08:58
janonymous+108:59
zengchen+108:59
livelaceapuimedo, Thanks for your answers, I'm going to inspect internals of kuryr and its possibilities for our needs09:03
apuimedolivelace: if you can provide some details on the needs, I'd love to hear them09:04
livelaceour needs are simple - bidirectional connections between VMs and containers, publishing containers services to bare metals and VMs09:06
apuimedolivelace: in the same tenant network?09:06
livelaceyes, I thinks it's already possible today09:06
apuimedoyes. It is possible09:07
apuimedobecause then you don't need the loadbalaancer type, the clusterip type is enough09:07
apuimedowhich we support with both octavia and neutron-lbaasv2 haproxy code09:08
livelaceI saw Contiv and OVN, and I think - kuryr is a more convenient solution09:08
apuimedolivelace: you should be able to use OVN via the neutron plugin with kuryr09:09
livelace+109:10
apuimedolivelace: are you planning on running pods inside VMs or just on baremetal?09:13
livelaceapuimedo, both, I will try to do that09:15
*** yamamoto has joined #openstack-kuryr09:15
livelaceI don't want have a penalty of full virtualization on part of working nodes09:16
livelacebrb09:16
apuimedolivelace: IIRC for pod-in-VM neutron ovn has a bug. dalvarez posted a fix but it is not merged yet09:22
*** yamamoto has quit IRC09:22
*** yedongcan has left #openstack-kuryr09:30
*** pmannidi has joined #openstack-kuryr09:48
*** yamamoto has joined #openstack-kuryr09:54
*** yamamoto has quit IRC10:07
*** yamamoto has joined #openstack-kuryr10:12
*** yamamoto has quit IRC10:17
*** yamamoto has joined #openstack-kuryr10:18
*** yamamoto has quit IRC10:20
*** yamamoto has joined #openstack-kuryr10:20
*** Guest14 is now known as ajo10:54
*** yamamoto has quit IRC10:58
*** yamamoto has joined #openstack-kuryr10:59
*** garyloug has quit IRC11:00
*** caowei has quit IRC11:00
*** garyloug has joined #openstack-kuryr11:03
*** yamamoto has quit IRC11:05
*** yamamoto has joined #openstack-kuryr11:07
*** yamamoto has quit IRC11:08
*** yamamoto has joined #openstack-kuryr11:10
*** yamamoto has quit IRC11:13
*** yamamoto has joined #openstack-kuryr11:15
*** yamamoto has quit IRC11:16
*** yamamoto has joined #openstack-kuryr11:16
*** atoth has joined #openstack-kuryr11:48
openstackgerritvikas choudhary proposed openstack/kuryr-kubernetes master: Add support to install Kuryr as a network addon  https://review.openstack.org/46667511:50
vikascirenab, ping11:55
*** aojea has quit IRC12:05
*** kiennt has quit IRC12:05
*** aojea has joined #openstack-kuryr12:06
*** aojea has quit IRC12:11
openstackgerritOpenStack Proposal Bot proposed openstack/kuryr-kubernetes master: Updated from global requirements  https://review.openstack.org/48801012:13
*** yamamoto has quit IRC12:18
*** yamamoto has joined #openstack-kuryr12:20
*** yamamoto has quit IRC12:20
apuimedovikasc: irenab is on pto12:35
vikascapuimedo, aah , ok. just wanted to say sorry to her for the time she struggled with network-addon patch12:36
vikasc:)12:36
vikascearlier while local testing i could not catch an issue. patch is updated with the fix12:37
*** yamamoto has joined #openstack-kuryr12:40
*** aojea has joined #openstack-kuryr12:53
*** limao has joined #openstack-kuryr13:36
*** hongbin has joined #openstack-kuryr13:47
*** gouthamr has joined #openstack-kuryr14:00
*** pmannidi has quit IRC14:14
*** aojea has quit IRC14:38
*** mattmceuen has joined #openstack-kuryr15:07
*** limao has quit IRC15:33
*** limao has joined #openstack-kuryr15:34
*** limao has quit IRC15:38
openstackgerritMerged openstack/kuryr-libnetwork master: Make the name of existing net tag unique  https://review.openstack.org/49030815:45
*** livelace2 has joined #openstack-kuryr15:48
*** janki has quit IRC15:48
*** yamamoto has quit IRC15:54
*** tonanhngo has joined #openstack-kuryr15:56
*** ajo has quit IRC15:57
*** yamamoto has joined #openstack-kuryr15:58
*** yamamoto has quit IRC16:03
*** pcaruana has quit IRC16:27
*** tonanhngo has quit IRC16:58
*** garyloug has quit IRC17:05
*** tonanhngo has joined #openstack-kuryr17:06
*** tonanhngo has quit IRC17:07
*** livelace2 has quit IRC17:55
*** tonanhngo has joined #openstack-kuryr17:55
*** tonanhngo_ has joined #openstack-kuryr17:58
*** tonanhngo has quit IRC17:59
*** tonanhngo_ has quit IRC18:02
*** tonanhngo has joined #openstack-kuryr18:02
*** aojea has joined #openstack-kuryr19:30
*** aojea has quit IRC19:51
*** aojea has joined #openstack-kuryr20:09
*** aojea has quit IRC20:24
*** aojea has joined #openstack-kuryr20:28
*** aojea has quit IRC20:51
*** yamamoto has joined #openstack-kuryr21:09
*** aojea has joined #openstack-kuryr21:10
*** yamamoto has quit IRC21:21
*** tonanhngo has quit IRC21:33
*** tonanhngo has joined #openstack-kuryr21:38
*** tonanhngo has quit IRC21:43
*** tonanhngo has joined #openstack-kuryr21:53
*** tonanhngo has quit IRC21:57
*** yamamoto has joined #openstack-kuryr21:59
*** tonanhngo has joined #openstack-kuryr21:59
*** tonanhngo has quit IRC22:03
*** tonanhngo has joined #openstack-kuryr22:05
*** tonanhngo has quit IRC22:09
*** tonanhngo has joined #openstack-kuryr22:11
*** tonanhngo has quit IRC22:16
*** tonanhngo has joined #openstack-kuryr22:17
*** tonanhngo has quit IRC22:22
*** tonanhngo has joined #openstack-kuryr22:24
*** tonanhngo has quit IRC22:27
*** tonanhngo has joined #openstack-kuryr22:28
*** aojea has quit IRC23:12
*** hongbin has quit IRC23:31
*** mattmceuen has quit IRC23:45

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