| sbezverk_ | kfox1111: agree, but again it is not available now for all kube object we decided to use. | 00:00 |
|---|---|---|
| sbezverk_ | kfox1111: we will have to provide reconfigure feature, otherwise nobody will use it | 00:02 |
| kfox1111 | sbezverk_: still works with daemonsets. you just delete and recreate the daemonset. | 00:02 |
| kfox1111 | not sure how that is really any differnet then delting all the pods? | 00:03 |
| sbezverk_ | kfox1111: deleting daemonset and deleting pod initiated by daemon set is abs different things | 00:04 |
| kfox1111 | how so? | 00:04 |
| sbezverk_ | in latter you really relay on kube to re-initiate killed pod | 00:04 |
| sbezverk_ | daemonset stays the same, what you suggest will create a completely new daemonset object | 00:05 |
| sbezverk_ | I do not know kube well enough to see all possible impacts | 00:05 |
| kfox1111 | yes, but why does it matter? | 00:05 |
| kfox1111 | they acutally recommend it in several workflow documents. | 00:05 |
| kfox1111 | like, if you want to update things, delete and recreate it. or, | 00:06 |
| kfox1111 | if you want to update only new pods, | 00:06 |
| kfox1111 | you delete/recreate the daemonset, but tell it not to delete the pods. | 00:06 |
| kfox1111 | then only terminated/restarted or freshly launched pods get the new config. | 00:06 |
| sbezverk_ | ok strange because with this approach they always need to rely on something else to drive it | 00:06 |
| kfox1111 | yeah. thats why thye claim they aren't an orchestration engine. | 00:07 |
| kfox1111 | they just ensure that the current declarative requests are kept in that state as much as possible. | 00:07 |
| kfox1111 | there is a litlte bit of orchestration in that, but its really not their intention to do orchestration outside of self healing pods. | 00:07 |
| kfox1111 | they kind of expect something outside of k8s will be driving k8s I think. | 00:08 |
| kfox1111 | helm for example. | 00:08 |
| sbezverk_ | I see a bit of conflict here | 00:09 |
| kfox1111 | k. whatcha thinking? | 00:09 |
| sbezverk_ | example they devlop daemonset to be able to deploy | 00:09 |
| sbezverk_ | everywhere without knwoledge and need for external orchestration | 00:10 |
| kfox1111 | right. and I think thats not so much orchestration, but just replication. | 00:10 |
| kfox1111 | k8s is about having a pool of resources and trying to not have to care so much about what lands where specifically. k8s provides a schedulaer and an api to deal with that. | 00:11 |
| kfox1111 | k8s provides object abstractions such as rc or daemonset that helps you express a request in a way that is more abstract that way. | 00:11 |
| kfox1111 | so yo ujust say, I want 3 of these pods, and k8s deals with scheduling and watchdogging them. | 00:12 |
| kfox1111 | the lifecycle of the k8s objects though is left up to the user. | 00:12 |
| sbezverk_ | yeah but all hooks they offer, it is invitation of more granular management/orchstration than just kill everything and recreate object from scratch | 00:13 |
| kfox1111 | It could be used that way, but I don't think its intended for it. more, intended to allow containers to be assembled into pods well. | 00:14 |
| kfox1111 | like, they don't suport having an rc fire off a job before it starts. | 00:14 |
| kfox1111 | that would be good for higher level orchestration. | 00:14 |
| sbezverk_ | anyway it is not phylosophical discussion.. I guess when we need to implement, we will undersatnd all pros/cons | 00:14 |
| kfox1111 | but they exlude that, and helm provides it. | 00:14 |
| kfox1111 | as helm does orchestration things. | 00:15 |
| sbezverk_ | I meant it is more phylosophical | 00:15 |
| kfox1111 | yeah. | 00:16 |
| sbezverk_ | what bugs me is that they have all info needed | 00:16 |
| kfox1111 | right. but I think they havent because it goes against one of their phylosophines. | 00:16 |
| sbezverk_ | but they force people then to implement logic, querying for ojects used this map | 00:16 |
| sbezverk_ | when for them it is already there | 00:17 |
| * kfox1111 shrugs. | 00:18 | |
| kfox1111 | this is just what I've gathered after reading through the docs and talking with some devs. | 00:18 |
| kfox1111 | I totally coudl be wrong. | 00:18 |
| sbezverk_ | my theory is this :-) they did not have time to implement it in time so they invented idea that it conflicts with their idiology ;-) | 00:19 |
| kfox1111 | possible. | 00:20 |
| kfox1111 | I don't think its logically incosistant with their stated phylosophy though. | 00:20 |
| kfox1111 | that of atomic things. | 00:20 |
| kfox1111 | you add a new one and delete an old one, rather then try and patch something up. | 00:21 |
| kfox1111 | thats at the root of everything they do. | 00:21 |
| sbezverk_ | then why failing container get resolved by a container restart and not pod re-create? | 00:21 |
| sbezverk_ | according their logic you just need to kill everything and start from scratch ;-) | 00:22 |
| kfox1111 | I guess I'd agree with killing the container and restarting it to be fresh. | 00:23 |
| kfox1111 | I'm a little surprised it tries to reuse it. | 00:23 |
| kfox1111 | that does seem inconsistent behavior. | 00:23 |
| *** Pavo has quit IRC | 00:29 | |
| *** Pavo has joined #openstack-kolla | 00:30 | |
| *** yingjun has joined #openstack-kolla | 00:37 | |
| *** yingjun has joined #openstack-kolla | 00:37 | |
| *** tovin07 has joined #openstack-kolla | 00:48 | |
| *** tovin07_ has joined #openstack-kolla | 00:51 | |
| *** tovin07_ has quit IRC | 00:52 | |
| *** tovin07 has quit IRC | 00:56 | |
| *** tovin07 has joined #openstack-kolla | 00:57 | |
| *** tovin07_ has joined #openstack-kolla | 00:58 | |
| *** yuanying has joined #openstack-kolla | 01:03 | |
| *** Jeffrey4l has joined #openstack-kolla | 01:03 | |
| sdake_ | sbezverk_ we don't use trivialfix flag anymore | 01:05 |
| openstackgerrit | Serguei Bezverkhi proposed openstack/kolla-kubernetes: DO_NOT_MERGE PoC Configmap update monitor https://review.openstack.org/409493 | 01:07 |
| sbezverk_ | sdake_ What do we use then? | 01:08 |
| openstackgerrit | Pete Birley proposed openstack/kolla-kubernetes: DO_NOT_MERGE PoC Configmap update monitor (alt) https://review.openstack.org/409565 | 01:08 |
| sdake_ | i just said in the review | 01:08 |
| openstackgerrit | Serguei Bezverkhi proposed openstack/kolla-kubernetes: DO_NOT_MERGE PoC Configmap update monitor https://review.openstack.org/409493 | 01:09 |
| sbezverk_ | sdake_: Must have missed it, I mean mailing list ann. | 01:09 |
| sdake_ | sbezverk_ its no sweat - thats why I left in the review ;) | 01:10 |
| sdake_ | mailing list easy to miss unless you filter | 01:10 |
| *** srwilkers_away is now known as srwilkers | 01:10 | |
| srwilkers | o/ | 01:10 |
| openstackgerrit | Li Yingjun proposed openstack/kolla-kubernetes: Fix different repo issue for documentation https://review.openstack.org/405054 | 01:10 |
| sdake_ | sup srwilkers | 01:10 |
| srwilkers | sdake_, not much. just got back from a dinner celebration | 01:11 |
| sdake_ | nice i just had dinner too - but no celebration :) | 01:11 |
| sbezverk_ | portdirect: I guess I prefer 'suicide' approach in compare to your 'mass-killing' ;-) | 01:13 |
| openstackgerrit | Steve Wilkerson proposed openstack/kolla-kubernetes: Helm charts for adding Cinder https://review.openstack.org/409349 | 01:14 |
| portdirect | sbezverk_: your a dark dude :) apt anaolgy though | 01:14 |
| srwilkers | portdirect, his analogies are always good ;) | 01:15 |
| sdake_ | i didn't find it all that good | 01:16 |
| sbezverk_ | russian roots surfaces from time to time ;-) | 01:17 |
| sdake_ | just like all that STONITH crap in HA | 01:17 |
| sdake_ | "shoot the other node in the head" | 01:17 |
| kfox1111 | sdake_: yeah. | 01:17 |
| sdake_ | You wouold have thought someone could have come up with something a ltitle more appropriate | 01:17 |
| kfox1111 | I've seen stonith go wrong too many times. :/ | 01:17 |
| kfox1111 | ha is good if you really need it, but not iwhtout its own risks. | 01:17 |
| kfox1111 | unless well meaintained, it actually becomes anti-ha. | 01:17 |
| *** zhurong has joined #openstack-kolla | 01:18 | |
| kfox1111 | the only way to rally get good at it is with stuff like chaos monkey. | 01:18 |
| kfox1111 | do it super often. | 01:18 |
| *** zhubingbing has joined #openstack-kolla | 01:18 | |
| portdirect | kfox1111: just talking about that with someone :) https://github.com/asobti/kube-monkey | 01:18 |
| srwilkers | yeah, saw the kube-monkey stuff on twitter earlier | 01:19 |
| portdirect | yeah - we should give it a shot once a bit further down the line - good way to show off kolla | 01:21 |
| openstackgerrit | Kevin Fox proposed openstack/kolla-kubernetes: WIP: prototype multiversion/release support https://review.openstack.org/409569 | 01:28 |
| kfox1111 | haha. kube-monkey. awesome! :) | 01:28 |
| kfox1111 | I"m going to find a use for this, somewhere. :) | 01:28 |
| kfox1111 | portdirect: thanks for the link. :) | 01:28 |
| portdirect | kfox1111: np - all part of my ongoing quest to contuniually consume 31 hours of media every 24 hours. | 01:29 |
| kfox1111 | portdirect: sbezverk_ I took a stab at taht multiversion stuff we were talking about. | 01:29 |
| openstackgerrit | Kevin Fox proposed openstack/kolla-kubernetes: WIP: prototype multiversion/release support https://review.openstack.org/409569 | 01:29 |
| openstackgerrit | jianyi proposed openstack/kolla: Install solum dashboard into horizon image https://review.openstack.org/408981 | 01:31 |
| openstackgerrit | jianyi proposed openstack/kolla: Install solum dashboard into horizon image https://review.openstack.org/408981 | 01:32 |
| portdirect | kfox1111: will review properly later - but 1st impression=nice | 01:33 |
| kfox1111 | portdirect: cool. thanks. | 01:34 |
| srwilkers | im going to take a break for a bit. ill look at your versioning stuff kfox1111 and sbezverk_ / portdirect patch sets for the config update stuff in awhile | 01:41 |
| srwilkers | starting to feel crosseyed | 01:41 |
| kfox1111 | srwilkers: no rush. I don't think we can really use it until we get all the things into helm. | 01:43 |
| *** elynn has joined #openstack-kolla | 01:45 | |
| *** elynn_ has joined #openstack-kolla | 01:49 | |
| *** elynn has quit IRC | 01:50 | |
| *** sdake has joined #openstack-kolla | 01:51 | |
| *** lrensing has joined #openstack-kolla | 01:53 | |
| *** sdake has quit IRC | 01:53 | |
| *** sdake has joined #openstack-kolla | 01:54 | |
| *** sdake_ has quit IRC | 01:54 | |
| *** sdake has quit IRC | 01:54 | |
| *** sdake_ has joined #openstack-kolla | 01:55 | |
| *** srwilkers is now known as srwilkers_away | 01:56 | |
| openstackgerrit | dupengfei proposed openstack/kolla: Not to pull the base image when building. https://review.openstack.org/408940 | 02:01 |
| *** sacharya has joined #openstack-kolla | 02:12 | |
| *** yatin_on_leave has quit IRC | 02:16 | |
| *** sacharya has quit IRC | 02:24 | |
| openstackgerrit | jianyi proposed openstack/kolla: Install solum dashboard into horizon image https://review.openstack.org/408981 | 02:27 |
| *** yatin_on_leave has joined #openstack-kolla | 02:27 | |
| *** Pavo has quit IRC | 02:29 | |
| *** Pavo has joined #openstack-kolla | 02:29 | |
| *** fragatina has joined #openstack-kolla | 02:41 | |
| *** yuanying has quit IRC | 02:49 | |
| *** sdake_ has quit IRC | 02:53 | |
| *** fragatina has quit IRC | 02:55 | |
| *** fragatina has joined #openstack-kolla | 02:56 | |
| *** fragatina has quit IRC | 03:11 | |
| *** sdake_ has joined #openstack-kolla | 03:13 | |
| *** mgiles has quit IRC | 03:14 | |
| *** yatin_on_leave has quit IRC | 03:18 | |
| *** fragatina has joined #openstack-kolla | 03:22 | |
| *** yatin_on_leave has joined #openstack-kolla | 03:31 | |
| *** duonghq has joined #openstack-kolla | 03:34 | |
| openstackgerrit | zhubingbing proposed openstack/kolla-ansible: add panko role https://review.openstack.org/400122 | 03:48 |
| *** sdake_ has quit IRC | 03:48 | |
| duonghq | portdirect, around? | 03:51 |
| portdirect | duonghq sup? | 03:52 |
| duonghq | portdirect, if I use halcyon behind proxy, how can I determine no_proxy before the cluster is up? | 03:52 |
| portdirect | hmm - I must confess I dont know offhand | 03:56 |
| duonghq | thought you write this bit in guide | 03:57 |
| portdirect | it has allways been "172.16.35.11" for me, and the other people I know who have had to use a proxy (via vagrant) | 03:57 |
| duonghq | how can I change the no_proxy after vagrant up? | 03:57 |
| portdirect | but I'm not sure where that comes from :( | 03:57 |
| duonghq | portdirect, I got same IPs as you, no idea | 03:58 |
| portdirect | yeah - though I'm not sure how vagrant provisions that interface to be honest | 03:59 |
| duonghq | how can I change the no_proxy after vagrant up? (and other config in general) | 03:59 |
| portdirect | duonghq: https://github.com/att-comdev/halcyon-vagrant-kubernetes/blob/b255019a8e4f33be61ca4691499556cf7abb5b42/Vagrantfile#L62 | 04:00 |
| portdirect | you cannot change the config after the cluster has started for the vagrant parts, you could disable ansible in the vagrantfile and then run that manually if you wanted | 04:01 |
| duonghq | portdirect, but DHCP always assigned same IPs time-to-time? | 04:02 |
| duonghq | can we have kind of upgrade config? | 04:02 |
| duonghq | the cluster can be changed anytime | 04:02 |
| duonghq | if dev need? | 04:02 |
| portdirect | duonghq: I'm not sure I follow | 04:03 |
| portdirect | lets deat with those 1 at a time | 04:03 |
| portdirect | *deal | 04:03 |
| portdirect | I just checked and the ip address range is defined in the link i posted | 04:03 |
| duonghq | portdirect, sorry but I have some business for about 5min, brb, | 04:04 |
| portdirect | regarding cluster changed - what params would you be seeking to change in the cluster? | 04:04 |
| v1k0d3n | portdirect: you're still up! dedication my friend. | 04:13 |
| duonghq | portdirect, at least I can think is change number of minions node? | 04:14 |
| portdirect | v1k0d3n: I'm totally on a PST schedule right now :/ | 04:14 |
| duonghq | or I must teardown and up another cluster? | 04:15 |
| openstackgerrit | Duong Ha-Quang proposed openstack/kolla: Spec - Zero-downtime upgrade in Kolla https://review.openstack.org/409598 | 04:15 |
| *** JRobinson__ is now known as JRobinson__afk | 04:16 | |
| portdirect | 2 sec - just want to try | 04:16 |
| v1k0d3n | portdirect: i've been there...some long nights | 04:17 |
| portdirect | duonghq: you can edit the config.rb and increment the number here: https://github.com/att-comdev/halcyon-vagrant-kubernetes/blob/master/config.rb#L5 | 04:18 |
| portdirect | duonghq: and then adjust https://github.com/att-comdev/halcyon-vagrant-kubernetes/blob/master/config.rb#L16 | 04:18 |
| portdirect | duonghq: and then run 'vagrant up kube5' | 04:19 |
| duonghq | ah, just vagrant up <node_name>? | 04:19 |
| duonghq | tyvm | 04:19 |
| duonghq | I saw 2 lines but don't know how to bring it up | 04:20 |
| duonghq | new to vagrant | 04:20 |
| duonghq | sorry | 04:20 |
| portdirect | duonghq: but I'm doing it for the 1st time now so dont know how the ansible will cope -it should be fine but... | 04:20 |
| portdirect | duonghq: I'm only 30 days ahead of you :) dont be sorry -it how we learn | 04:20 |
| duonghq | :) | 04:20 |
| duonghq | hmm, my kube-init : initialize the kubernetes master task stuck :) | 04:21 |
| duonghq | guess I should update proxy | 04:22 |
| portdirect | if you discover anything, lets get it in the docs - either here or at the att-comdev repo if its going to be usefull to people outside of kolla as well :) | 04:22 |
| *** elynn_ has quit IRC | 04:22 | |
| portdirect | oh yeah - whoops | 04:22 |
| *** saneax-_-|AFK is now known as saneax | 04:23 | |
| duonghq | portdirect, sure | 04:23 |
| *** sacharya has joined #openstack-kolla | 04:24 | |
| duonghq | portdirect, thinking about we can replicate what kolla doing right now for /etc/hosts | 04:25 |
| duonghq | autofill /etc/hosts with appropriate hostname... | 04:26 |
| portdirect | could you give example - paste? | 04:27 |
| duonghq | currently, kolla-ansible w/ bootstrap action can deploy /etc/hosts w/ all hostname and theirs IPs to target nodes | 04:28 |
| duonghq | it's much more convenient than using bare IP? | 04:28 |
| *** Pavo has quit IRC | 04:29 | |
| *** sacharya has quit IRC | 04:29 | |
| openstackgerrit | Duong Ha-Quang proposed openstack/kolla: Spec - Zero-downtime upgrade in Kolla https://review.openstack.org/409598 | 04:31 |
| *** Pavo has joined #openstack-kolla | 04:34 | |
| portdirect | duonghq: that would be possible - but I'm not sure we should do that here - could you provide an example? | 04:34 |
| duonghq | portdirect, "make no_proxy more consistent between deployment" is a good reason? | 04:35 |
| *** elynn_ has joined #openstack-kolla | 04:45 | |
| *** elynn_ has quit IRC | 04:50 | |
| *** elynn_ has joined #openstack-kolla | 04:50 | |
| *** lrensing has quit IRC | 04:52 | |
| *** rstarmer has joined #openstack-kolla | 04:52 | |
| *** JRobinson__afk is now known as JRobinson__ | 04:54 | |
| *** yuanying has joined #openstack-kolla | 04:58 | |
| portdirect | duonghq: sorry got distracted - but could you explain how (I genuinly dont know) | 04:58 |
| *** jascott1_ has quit IRC | 05:00 | |
| duonghq | portdirect, how long you still at here? | 05:01 |
| portdirect | hopfull not long :/ (its 5am here) | 05:02 |
| duonghq | not sleep yet? | 05:02 |
| duonghq | so for short: from user PoV: they can specify no_proxy like "localhost,127.0.0.1,kube1,kube2,kube3,kube4" | 05:03 |
| duonghq | etc | 05:03 |
| portdirect | I only woke up at 2pm... | 05:03 |
| duonghq | strange timetable :) | 05:03 |
| duonghq | portdirect, I'll away for about 30mins in 5mins, see you later | 05:08 |
| portdirect | ah - just saw your comment re poxy - I'll look into that - thanks :) | 05:09 |
| *** mdnadeem has joined #openstack-kolla | 05:13 | |
| *** portdirect is now known as portdirect_away | 05:15 | |
| openstackgerrit | Kien Nguyen proposed openstack/kolla: Remove `$` character for document consistency. https://review.openstack.org/409608 | 05:24 |
| *** sacharya has joined #openstack-kolla | 05:26 | |
| *** sacharya has quit IRC | 05:30 | |
| *** sayantan_ has joined #openstack-kolla | 05:30 | |
| *** sayantan_ has quit IRC | 05:31 | |
| *** fragatina has quit IRC | 05:32 | |
| *** fragatina has joined #openstack-kolla | 05:42 | |
| *** yingjun has quit IRC | 05:45 | |
| openstackgerrit | Surya Prakash Singh proposed openstack/kolla-ansible: Implement Keystone zero-downtime upgrade https://review.openstack.org/405215 | 05:46 |
| *** yingjun has joined #openstack-kolla | 05:47 | |
| *** skramaja has joined #openstack-kolla | 05:59 | |
| *** jascott1_ has joined #openstack-kolla | 06:00 | |
| *** fragatina has quit IRC | 06:02 | |
| *** jascott1_ has quit IRC | 06:05 | |
| *** mdnadeem has quit IRC | 06:18 | |
| *** rsjethani has joined #openstack-kolla | 06:18 | |
| *** sp_ has quit IRC | 06:19 | |
| *** mdnadeem has joined #openstack-kolla | 06:22 | |
| *** openstackgerrit_ has joined #openstack-kolla | 06:23 | |
| *** openstackgerrit_ has quit IRC | 06:25 | |
| *** openstackgerrit_ has joined #openstack-kolla | 06:26 | |
| *** openstackgerrit_ has quit IRC | 06:27 | |
| *** Pavo has quit IRC | 06:29 | |
| *** Pavo has joined #openstack-kolla | 06:33 | |
| *** openstackgerrit_ has joined #openstack-kolla | 06:34 | |
| *** openstackgerrit_ has quit IRC | 06:35 | |
| *** mdnadeem has quit IRC | 06:39 | |
| *** bjolo_ has quit IRC | 06:40 | |
| *** JRobinson__ has quit IRC | 06:44 | |
| *** Satya_ has joined #openstack-kolla | 06:46 | |
| *** mdnadeem has joined #openstack-kolla | 06:51 | |
| *** sp_ has joined #openstack-kolla | 06:55 | |
| *** openstackgerrit_ has joined #openstack-kolla | 06:59 | |
| duonghq | openstackgerrit_, ping | 06:59 |
| *** openstackgerrit_ has quit IRC | 07:00 | |
| *** jascott1_ has joined #openstack-kolla | 07:02 | |
| *** jascott1_ has quit IRC | 07:05 | |
| *** jascott1_ has joined #openstack-kolla | 07:12 | |
| *** sacharya has joined #openstack-kolla | 07:27 | |
| *** sacharya has quit IRC | 07:31 | |
| Satya_ | Hi sdake | 07:42 |
| Satya_ | hi jeffrey4l | 07:42 |
| Jeffrey4l | sup Satya_ | 07:42 |
| Satya_ | regarding the issue where after neuton upgrade | 07:42 |
| Satya_ | neutron | 07:42 |
| Jeffrey4l | Satya_, any progress? | 07:43 |
| Satya_ | yes | 07:43 |
| Jeffrey4l | cool. | 07:43 |
| Jeffrey4l | could u explain? | 07:43 |
| Jeffrey4l | is it a bug in kolla or neutron | 07:43 |
| Satya_ | after rebooting the host i am able to connect to old VMs | 07:43 |
| Satya_ | it seems like | 07:43 |
| Jeffrey4l | hrm :) | 07:43 |
| Jeffrey4l | Satya_, which version are trying from and to ? | 07:44 |
| Satya_ | it seems its somewhat related to namespace | 07:44 |
| Satya_ | minor to minor | 07:44 |
| Satya_ | basically i took 2week old code and upgraded to latest trunk | 07:44 |
| SamYaple | hmmm if something has changed and the compute contaienrs now have namespaces... then that would cause it | 07:44 |
| Satya_ | of octa | 07:44 |
| Jeffrey4l | Satya_, can u confirm this? | 07:44 |
| *** rsjethani has quit IRC | 07:45 | |
| Jeffrey4l | how about restart docker engine? Satya_ | 07:45 |
| Satya_ | tried restarting the docker container but same issue | 07:45 |
| Jeffrey4l | Satya_, two weeks code. should no such big change recently. | 07:45 |
| Satya_ | but have not tried with restarting the docker enginer | 07:45 |
| Satya_ | engine | 07:45 |
| Jeffrey4l | Satya_, did u get any error message in logs? | 07:45 |
| Satya_ | nope | 07:45 |
| Satya_ | no errors anywhere | 07:46 |
| Jeffrey4l | weird :( | 07:46 |
| Jeffrey4l | if you have any chance to do this test, please try: 1. restart docker-engine 2. stop docker-engine, remove all namespace, and start again. | 07:47 |
| Satya_ | after upgrading to trunk the old VMs are not reachable but created new VMs which are rechable | 07:47 |
| Satya_ | ok sure | 07:47 |
| Satya_ | trying that | 07:47 |
| Jeffrey4l | Satya_ the issue is: vm can not get ip , right? how about specify a fixed ip in vm direct. then will vm work? | 07:48 |
| openstackgerrit | Duong Ha-Quang proposed openstack/kolla: Spec - Zero-downtime upgrade in Kolla https://review.openstack.org/409598 | 07:50 |
| *** huzhengchuan has joined #openstack-kolla | 07:52 | |
| *** msimonin has joined #openstack-kolla | 07:54 | |
| *** msimonin has quit IRC | 07:57 | |
| *** matrohon has joined #openstack-kolla | 07:57 | |
| *** flaper87 has quit IRC | 08:11 | |
| *** flaper87 has joined #openstack-kolla | 08:11 | |
| *** jascott1_ has quit IRC | 08:23 | |
| *** Pavo has quit IRC | 08:29 | |
| openstackgerrit | Jeffrey Zhang proposed openstack/kolla-ansible: Add enable_cinder_backend_nfs option example https://review.openstack.org/409659 | 08:30 |
| *** pcaruana has joined #openstack-kolla | 08:30 | |
| *** Pavo has joined #openstack-kolla | 08:34 | |
| *** pcaruana has quit IRC | 08:37 | |
| *** pcaruana has joined #openstack-kolla | 08:38 | |
| *** egonzalez90 has joined #openstack-kolla | 08:46 | |
| Satya_ | jeffrey4l after restarting the docker still same issue so experiment 1 failed | 08:51 |
| Satya_ | jeffrey4l deleet the namespace from server or compute container | 08:56 |
| openstackgerrit | chen.xing proposed openstack/kolla: Fix the format issue https://review.openstack.org/407343 | 08:57 |
| openstackgerrit | Sam Yaple proposed openstack/kolla: Add kafka image https://review.openstack.org/406874 | 09:00 |
| openstackgerrit | Sam Yaple proposed openstack/kolla: Add zookeeper image https://review.openstack.org/406878 | 09:00 |
| openstackgerrit | Sam Yaple proposed openstack/kolla: Add monasca images https://review.openstack.org/406422 | 09:00 |
| openstackgerrit | Eduardo Gonzalez proposed openstack/kolla-ansible: Zaqar ansible role https://review.openstack.org/407760 | 09:00 |
| openstackgerrit | chen.xing proposed openstack/kolla: Fix the format issue https://review.openstack.org/407343 | 09:01 |
| *** shardy has joined #openstack-kolla | 09:02 | |
| *** huzhengchuan has quit IRC | 09:10 | |
| *** huzhengchuan has joined #openstack-kolla | 09:10 | |
| *** gfidente has joined #openstack-kolla | 09:10 | |
| zhubingbing | hello guys | 09:11 |
| *** shardy has quit IRC | 09:12 | |
| *** shardy has joined #openstack-kolla | 09:12 | |
| SamYaple | hello | 09:13 |
| SamYaple | Satya_: sounds like you didnt setup docker properly | 09:13 |
| SamYaple | Satya_: http://docs.openstack.org/developer/kolla/quickstart.html | 09:13 |
| SamYaple | Satya_: you need to set MountFlags=shared | 09:14 |
| egonzalez90 | morning SamYaple zhubingbing | 09:15 |
| Satya_ | SamYaple i hope you have not got the full conversation | 09:15 |
| zhubingbing | hi | 09:16 |
| Satya_ | it not related to the docker registrey | 09:16 |
| SamYaple | Satya_: niether is what i posted | 09:16 |
| Satya_ | its about the neutron upgrade issue | 09:16 |
| SamYaple | Satya_: trust me, check what i posted | 09:16 |
| Satya_ | SamYaple can we go for a priv chat? | 09:17 |
| SamYaple | i suppose? | 09:18 |
| *** jascott1_ has joined #openstack-kolla | 09:23 | |
| *** jascott1_ has quit IRC | 09:28 | |
| *** elynn_ has quit IRC | 09:30 | |
| *** yingjun has quit IRC | 09:33 | |
| *** yingjun has joined #openstack-kolla | 09:33 | |
| openstackgerrit | Sam Yaple proposed openstack/kolla: Add kafka image https://review.openstack.org/406874 | 09:37 |
| openstackgerrit | Sam Yaple proposed openstack/kolla: Add monasca images https://review.openstack.org/406422 | 09:37 |
| *** yingjun has quit IRC | 09:37 | |
| *** yuanying has quit IRC | 09:43 | |
| *** Serlex has joined #openstack-kolla | 09:49 | |
| *** strigazi_AFK is now known as strigazi | 09:49 | |
| *** zhurong has quit IRC | 10:02 | |
| *** tovin07_ has quit IRC | 10:09 | |
| *** athomas has joined #openstack-kolla | 10:11 | |
| *** duonghq has quit IRC | 10:11 | |
| *** jascott1_ has joined #openstack-kolla | 10:25 | |
| openstackgerrit | jianyi proposed openstack/kolla: Install solum dashboard into horizon image https://review.openstack.org/408981 | 10:27 |
| *** Pavo has quit IRC | 10:29 | |
| *** jascott1_ has quit IRC | 10:29 | |
| *** Satya_ has quit IRC | 10:33 | |
| *** Pavo has joined #openstack-kolla | 10:34 | |
| *** rafael_sisweb has joined #openstack-kolla | 10:34 | |
| *** igordcard has quit IRC | 10:36 | |
| *** igordcard has joined #openstack-kolla | 10:36 | |
| *** rafael_sisweb has quit IRC | 10:39 | |
| *** rafael_sisweb has joined #openstack-kolla | 10:39 | |
| openstackgerrit | Eduardo Gonzalez proposed openstack/kolla-ansible: Ansible-ize OpenStack Designate https://review.openstack.org/408465 | 10:46 |
| openstackgerrit | Eduardo Gonzalez proposed openstack/kolla-ansible: Ansible-ize OpenStack Designate https://review.openstack.org/408465 | 10:48 |
| egonzalez90 | pbourke Jeffrey4l zhubingbing could your review ^^? | 10:50 |
| zhubingbing | ok | 10:51 |
| zhubingbing | doning | 10:52 |
| egonzalez90 | enable_designate var is not present, and horizon is failing to deploy | 10:52 |
| *** mkoderer has joined #openstack-kolla | 10:52 | |
| pbourke | egonzalez90: have you succesfully deployed designate with this patch/ | 10:53 |
| pbourke | ? | 10:53 |
| egonzalez90 | pbourke: not tested yet, only applied differences between the merge into kolla and changes made after repo split here. | 10:54 |
| egonzalez90 | give me a sec to test | 10:55 |
| zhubingbing | OK | 10:55 |
| pbourke | thanks, it looks good but seems reckless to merge without at least something having verified it stands up | 10:55 |
| pbourke | there's a lot of things to get right in this playbook patches | 10:55 |
| *** n0isyn0ise has joined #openstack-kolla | 10:56 | |
| *** papacz has joined #openstack-kolla | 10:59 | |
| SamYaple | pbourke: im not sure how those package macros work | 10:59 |
| pbourke | SamYaple: when a Dockerfile needs to install packages they're definied in a jinja2 variable | 11:01 |
| pbourke | e.g. https://github.com/openstack/kolla/blob/master/docker/glance/glance-api/Dockerfile.j2#L9 | 11:01 |
| pbourke | then the macro installs them in a consistent way. Users can then append/remove packages to this list | 11:02 |
| pbourke | but if no packages are to be installed the macro shouldn't be called | 11:02 |
| SamYaple | hmmm. shouldnt it still exist for user customizable containres where the actual serice doesn't need aynthing? | 11:03 |
| pbourke | I guess so, we just haven't done that so far | 11:05 |
| SamYaple | if the whole point is to be user extendable, that would only make sense | 11:05 |
| pbourke | well it allows the user to tweak packages we install. I think most of the time if we dont install packages in an image then users wont need to either. If they do they can use the header/footer blocks | 11:06 |
| SamYaple | does the user defined macro thing _override_ or extend? | 11:08 |
| pbourke | http://docs.openstack.org/developer/kolla/image-building.html#package-customisation | 11:09 |
| pbourke | override/append/remove | 11:10 |
| SamYaple | cool. thanks for the link pbourke | 11:10 |
| *** sdake has joined #openstack-kolla | 11:12 | |
| pbourke | documented by yours truly :p | 11:12 |
| pbourke | anywho, I wouldnt be against adding macro.install entries to every file regardless as you say, but its a lot of extra lines for potential little gain | 11:13 |
| SamYaple | i dont have a prefence eithre way to be honest | 11:14 |
| SamYaple | i was just asking | 11:14 |
| egonzalez90 | pbourke: i'm deploying designate to test it | 11:19 |
| pbourke | egonzalez90: kk | 11:19 |
| sdake | morning peeps | 11:20 |
| *** sean-k-m1 is now known as sean-k-mooney | 11:29 | |
| SamYaple | i just wrote designate support of openstack-ansible | 11:30 |
| SamYaple | its a pain egonzalez90 | 11:30 |
| egonzalez90 | SamYaple: can you check if my PS looks good to you? You played with it before https://review.openstack.org/#/c/408465/ | 11:34 |
| Jeffrey4l | designate patch is just a backport. if it won't work, we should fix it in another patch . egonzalez90 pbourke | 11:42 |
| *** kollabot has joined #openstack-kolla | 11:45 | |
| *** mliima has joined #openstack-kolla | 11:45 | |
| *** kollabot5 has quit IRC | 11:45 | |
| SamYaple | egonzalez90: i reviewed it, ahve you tested it? as it stands, what you have in that patchset doesn't do anything. nor does it work | 11:45 |
| egonzalez90 | SamYaple: not, i'm didn't tested ( doing it ATM), is a backport from kolla repo that was lost during repo split | 11:47 |
| SamYaple | esh. it should have never merged into the kolla repo | 11:48 |
| *** zhubingbing has quit IRC | 11:48 | |
| SamYaple | its got a lot of work before its even usable | 11:48 |
| SamYaple | Jeffrey4l: it should be blocked. its not a finished patchset at all | 11:48 |
| SamYaple | Jeffrey4l: it doesnt do anything, its half finished code that never worked | 11:48 |
| egonzalez90 | SamYaple +1 | 11:49 |
| egonzalez90 | even the first api command failed "Unable to determine appropriate api-paste-config file" | 11:49 |
| SamYaple | this should have never merged | 11:50 |
| SamYaple | pbourke: ^^ | 11:50 |
| SamYaple | you know we used to have designate in the repo | 11:50 |
| *** mliima has quit IRC | 11:50 | |
| SamYaple | we removed it because it wasnt' maintained | 11:50 |
| SamYaple | and didnt work | 11:50 |
| pbourke | SamYaple: thanks | 11:52 |
| pbourke | that's basically what I needed to know | 11:52 |
| egonzalez90 | i'll work on it to make it work | 11:52 |
| SamYaple | id say its 70% of a patchset | 11:53 |
| Jeffrey4l | OK thanks for the explanation. no idea about this. but it is merged in kolla repo. but loss during repo split. technecially, it is a backport. | 11:53 |
| SamYaple | needs, at the very least, the ability to update the pool-manager config and some other config info for rndc keys | 11:53 |
| SamYaple | Jeffrey4l: maybe. but isn't a "backport" of something that shouldn't have merged | 11:53 |
| egonzalez90 | thinks we should revert approve this PS https://review.openstack.org/#/c/408714/ | 11:54 |
| egonzalez90 | horizon is not workign at all | 11:54 |
| *** zhurong has joined #openstack-kolla | 11:54 | |
| SamYaple | yea | 11:55 |
| *** mliima has joined #openstack-kolla | 11:55 | |
| Jeffrey4l | hrm. i prefer merge it. and file a new bug report that desinate do not work. because fix it in the *backport* patch is weird. | 11:57 |
| SamYaple | i think the patch should be reverted | 11:57 |
| SamYaple | meaning it shouldnt exist in the repo | 11:58 |
| egonzalez90 | SamYaple: is not needed to be reverted because was lost during repo split (is not present nor in kolla or kolla-ansible) | 11:59 |
| SamYaple | egonzalez90: i know, but thats my point | 11:59 |
| SamYaple | it shouldnt have been merged in the first place | 12:00 |
| SamYaple | so lets just not merge it now | 12:00 |
| SamYaple | and fix it | 12:00 |
| Jeffrey4l | ok i am OK with anyway. since this patch is fixing a bug, not implement a bp. | 12:00 |
| Jeffrey4l | hrm, simple things messed up :( | 12:00 |
| mliima | morning guys | 12:00 |
| SamYaple | yea i think the BP should be reopened and this patchset submitted again | 12:00 |
| Jeffrey4l | s/simple/seems | 12:01 |
| Jeffrey4l | bp is here https://blueprints.launchpad.net/kolla-ansible/+spec/ansible-designate | 12:02 |
| Jeffrey4l | egonzalez90, could u close the bp and re-submit the patch with `implement blueprint`? | 12:03 |
| Jeffrey4l | we have to re-enable the bp if we fix the issue in that patch. | 12:03 |
| Jeffrey4l | re-open* | 12:04 |
| *** dave-mccowan has joined #openstack-kolla | 12:04 | |
| Jeffrey4l | ( re-open bp is weird ) | 12:04 |
| egonzalez90 | Jeffrey4l: thats what i was to ask you ;) | 12:05 |
| Jeffrey4l | ( so i prefer to just backport it and file a new bug to report it do not work ) | 12:05 |
| egonzalez90 | how to do it, status is completed, even if I change Implemented to any other thing, status does not change | 12:05 |
| *** zhurong has quit IRC | 12:06 | |
| *** zhurong has joined #openstack-kolla | 12:07 | |
| *** tovin07_ has joined #openstack-kolla | 12:07 | |
| SamYaple | Jeffrey4l: if we merge it, and work doesnt finish to actually make it _work_, then its dead code just like desginate was last time we removed it | 12:08 |
| Jeffrey4l | since we haven't ci for each role and fewer people test the patch. i have to say this situation is common in kolla. | 12:08 |
| Jeffrey4l | SamYaple, we did not removed it. it is loss by mistake. the kolla-ansible's code is cut around Nov 11 from kolla. | 12:09 |
| Jeffrey4l | and it is push to github at around Nov 15. | 12:09 |
| Jeffrey4l | the patch between these two day is loss by mistake. | 12:09 |
| Jeffrey4l | not removed by purpose. | 12:10 |
| SamYaple | Jeffrey4l: no. designate used to exist in kolla. before Liberty | 12:10 |
| SamYaple | we removed it in riht before mitaka i think | 12:10 |
| SamYaple | because the code didnt work | 12:10 |
| SamYaple | if this code doesnt work, why merge it | 12:10 |
| Jeffrey4l | OK. i have no idea on that. :( . this patch is push days ago. | 12:10 |
| SamYaple | just because at one point the reviwers made the mistake to merge it, why merge it again | 12:10 |
| Jeffrey4l | normally, we do not require the core member to test every patch his is reviewing. it is too overload. | 12:11 |
| SamYaple | of course not, but people shouldnt be merging known broken code | 12:12 |
| SamYaple | thats what would be happening here | 12:12 |
| Jeffrey4l | if every bp is implement very well. there will no bug at all. i prefer to use ci/gate to ensure the code is OK. | 12:13 |
| Jeffrey4l | this is really normally when there is no ci ;( . | 12:14 |
| Jeffrey4l | i do not expect this. it is hard to avoid unless we have ci. | 12:14 |
| Jeffrey4l | because you are working on this and this patch is loss during repo split. so desinate patch issue is exposed. ;( | 12:17 |
| *** rhallisey has joined #openstack-kolla | 12:19 | |
| *** zhurong has quit IRC | 12:19 | |
| openstackgerrit | Merged openstack/kolla-ansible: Change Dockerfile to ansible role in the release note. https://review.openstack.org/409517 | 12:25 |
| openstackgerrit | Merged openstack/kolla-ansible: Add enable_cinder_backend_nfs option example https://review.openstack.org/409659 | 12:25 |
| *** shardy is now known as shardy_lunch | 12:26 | |
| *** williamcaban has quit IRC | 12:26 | |
| *** williamcaban has joined #openstack-kolla | 12:26 | |
| *** Pavo has quit IRC | 12:29 | |
| *** rstarmer has quit IRC | 12:30 | |
| *** msimonin has joined #openstack-kolla | 12:31 | |
| *** Satya_ has joined #openstack-kolla | 12:32 | |
| Satya_ | Hi SamYaple | 12:33 |
| *** saneax is now known as saneax-_-|AFK | 12:33 | |
| *** Pavo has joined #openstack-kolla | 12:34 | |
| SamYaple | hello | 12:34 |
| openstackgerrit | Merged openstack/kolla: Fix some typo https://review.openstack.org/407884 | 12:34 |
| Satya_ | tried with the kolla.conf file | 12:34 |
| Satya_ | still having the issue | 12:34 |
| Satya_ | seems like TASK [neutron : Ensuring OVS bridge is properly setup] ************************* changed: [server1] => (item=[u'br809', u'eno1.809']) | 12:35 |
| Satya_ | not sure though | 12:35 |
| Satya_ | cat /etc/systemd/system/docker.service.d/kolla.conf [Service] MountFlags=shared | 12:36 |
| *** zhubingbing has joined #openstack-kolla | 12:43 | |
| Satya_ | SamYaple please pass me the same command to check the name space | 12:44 |
| *** huzhengchuan has quit IRC | 12:46 | |
| SamYaple | Satya_: docker run --privileged -v /run/netns:/run/netns:shared kolla/oraclelinux-neutron-l3-agent:4.0.0 ip netns add test | 12:47 |
| SamYaple | when that container exits, you should be able to access the 'test' net namespace from the host | 12:47 |
| Satya_ | mount --make-shared /var/run/netns failed: Operation not permitted | 12:49 |
| Satya_ | http://paste.openstack.org/show/592104/ | 12:50 |
| openstackgerrit | Merged openstack/kolla: Install qemu-kvm rather then qemu-system-x86 https://review.openstack.org/409513 | 12:50 |
| openstackgerrit | Merged openstack/kolla: bifrost guide improved https://review.openstack.org/406177 | 12:50 |
| openstackgerrit | Merged openstack/kolla: Remove `$` character for document consistency. https://review.openstack.org/409608 | 12:51 |
| SamYaple | Satya_: well thats your problem | 12:51 |
| openstackgerrit | Merged openstack/kolla: Fix the format issue https://review.openstack.org/407343 | 12:51 |
| SamYaple | Satya_: until you figure that out, its not going to work | 12:51 |
| Satya_ | how can i solve that? | 12:51 |
| SamYaple | Satya_: what kernel version do you have? | 12:51 |
| Satya_ | 4.1.12-61.1.22.el7uek.x86_64 | 12:51 |
| Satya_ | thats the latest i guess | 12:52 |
| SamYaple | Satya_: so you have apparmor/selinunx running? | 12:53 |
| *** zhurong has joined #openstack-kolla | 12:53 | |
| Satya_ | no | 12:53 |
| Satya_ | i disabled selinux | 12:53 |
| Satya_ | getenforce Disabled | 12:53 |
| SamYaple | try running with docker run --privileged -v /run/:/run/:shared | 12:54 |
| Satya_ | docker run --privileged -v /run/:/run/:shared centralizedreg.us.oracle.com:5000/cas/oraclelinux-source-neutron-l3-agent:next ip netns add test mount --make-shared /var/run/netns failed: Operation not permitted | 12:56 |
| SamYaple | cant you post to pastebin the output of `mount` | 12:57 |
| Satya_ | http://paste.openstack.org/show/592105/ | 12:58 |
| SamYaple | im not sure this will work in your running system, but youll want to setup those mount points as shared `mount --make-shared /run/netns` | 13:00 |
| SamYaple | play with it until the command i gave you works. then youll be fine | 13:01 |
| SamYaple | its not a neutron/kolla issue that yo uare having on upgrade | 13:01 |
| Satya_ | is not it done by kolla? | 13:02 |
| Satya_ | i mean kolla-ansible as a part of task? | 13:02 |
| SamYaple | no. this is part of setting up docker | 13:02 |
| SamYaple | which is outside the scope of kolla-ansible | 13:02 |
| Satya_ | ok | 13:04 |
| Satya_ | i did | 13:04 |
| Satya_ | [root@server1 ~]# mount --make-shared /run/netns [root@server1 ~]# | 13:04 |
| Satya_ | but in the mount command i cannt see it | 13:04 |
| Satya_ | http://paste.openstack.org/show/592106/ | 13:06 |
| Satya_ | SamYaple i am bit confused why my old VMs are reachable after the host reboot | 13:07 |
| sdake | yo peeps | 13:11 |
| SamYaple | it has to do with mount namespaces in the kernel | 13:11 |
| *** zhurong has quit IRC | 13:11 | |
| Satya_ | ok | 13:11 |
| Satya_ | reboot solves that problem? | 13:11 |
| Satya_ | and mountflag too? | 13:11 |
| SamYaple | reboot creates new mount namespaces | 13:11 |
| Satya_ | mountflag uses the same one? | 13:12 |
| *** zhurong has joined #openstack-kolla | 13:12 | |
| Satya_ | got a chance to look at the paste SamYaple | 13:12 |
| Satya_ | Hi sdake | 13:13 |
| Satya_ | how are you | 13:13 |
| sdake | just woke up | 13:15 |
| sdake | so tired | 13:15 |
| sdake | other then that good :) | 13:15 |
| Satya_ | :P | 13:15 |
| Satya_ | jetlag over | 13:15 |
| sdake | long ago | 13:15 |
| sdake | takes 1 week per hour of lag | 13:15 |
| Satya_ | cool | 13:15 |
| *** msimonin has quit IRC | 13:15 | |
| Satya_ | lol | 13:15 |
| openstackgerrit | Sam Yaple proposed openstack/kolla: Add kafka image https://review.openstack.org/406874 | 13:18 |
| openstackgerrit | Sam Yaple proposed openstack/kolla: Add zookeeper image https://review.openstack.org/406878 | 13:18 |
| openstackgerrit | Sam Yaple proposed openstack/kolla: Add monasca images https://review.openstack.org/406422 | 13:18 |
| openstackgerrit | Sam Yaple proposed openstack/kolla: Add monasca images https://review.openstack.org/406422 | 13:22 |
| *** papacz has quit IRC | 13:23 | |
| Satya_ | Hi sdake | 13:25 |
| *** zhurong has quit IRC | 13:26 | |
| *** shardy_lunch is now known as shardy | 13:28 | |
| *** zhurong has joined #openstack-kolla | 13:28 | |
| *** tovin07_ has quit IRC | 13:28 | |
| *** papacz has joined #openstack-kolla | 13:32 | |
| *** tovin07_ has joined #openstack-kolla | 13:32 | |
| *** athomas has quit IRC | 13:36 | |
| *** athomas has joined #openstack-kolla | 13:42 | |
| sdake | sup Satya_ | 13:42 |
| *** msimonin has joined #openstack-kolla | 13:42 | |
| sdake | hey SamYaple - zookeeper was in the repo at one point and I removed it on the mesos deprecation | 13:42 |
| sdake | SamYaple did you happen to look at it in the git history? (I recognize the macro stuff is all new and won't necessarily work) | 13:43 |
| SamYaple | sdake: pbourke asked me to remove it since i wasn't using it in a consistent way with the other containers | 13:43 |
| SamYaple | if you mean the package macro stuff | 13:44 |
| sdake | SamYaple i'm pretty certain I removed the old zookeeper | 13:44 |
| SamYaple | oh. yea. that was discussed in the patchset | 13:45 |
| sdake | ok - just wanted to know if you hd a look at the old code | 13:45 |
| SamYaple | the old zookeeper came from mesos repo | 13:45 |
| sdake | haven't looked at the patchset yet | 13:45 |
| sdake | looking shortly | 13:45 |
| openstackgerrit | caoyuan proposed openstack/kolla: Remove the unnecessary "openstack-zaqar-ui" https://review.openstack.org/407576 | 13:45 |
| SamYaple | zookeeper isnt packaged for centos yet | 13:45 |
| sdake | i also think monasca would be a good thing to have | 13:45 |
| SamYaple | im implementing for work | 13:45 |
| sdake | cool | 13:45 |
| sdake | i looked at kafka | 13:45 |
| sdake | partially-implements: blueprint helps show the link in gerrit | 13:46 |
| sdake | otheriwse havve to hunt for the blueprint - to make sure its been approved | 13:46 |
| SamYaple | yea it was just missed on that patchset | 13:46 |
| sdake | np | 13:46 |
| SamYaple | its in the other for that series | 13:46 |
| sdake | i'll get to em shortly :) | 13:46 |
| *** srwilkers_away has quit IRC | 13:46 | |
| sdake | SamYaple you want in the drivers team so you can set your own state on kolla ? | 13:48 |
| SamYaple | nah. ive only go a few toes back in the water at the moment | 13:48 |
| sdake | k | 13:48 |
| SamYaple | if this kolla-salt stuff im working on pans out, well see then | 13:48 |
| sdake | is good progress or started more accurate | 13:48 |
| SamYaple | good progress | 13:49 |
| sdake | thanks | 13:49 |
| SamYaple | we do seperate blueprints for kolla-ansible right? | 13:49 |
| sdake | ya it has an entirely separate tracker | 13:49 |
| sdake | don't blame me | 13:49 |
| sdake | :) | 13:49 |
| SamYaple | i think thats good | 13:49 |
| sdake | inc0's idea | 13:49 |
| SamYaple | the images are almost done, so good progress is accurate | 13:49 |
| sdake | and separate core teams as well | 13:50 |
| SamYaple | i think thats the right idea for that | 13:50 |
| SamYaple | nice clean seperation | 13:50 |
| *** zhenguo has quit IRC | 13:52 | |
| *** zhenguo has joined #openstack-kolla | 13:52 | |
| *** lrensing has joined #openstack-kolla | 13:53 | |
| *** zhurong has quit IRC | 13:53 | |
| *** prameswar has joined #openstack-kolla | 13:56 | |
| openstackgerrit | Merged openstack/kolla-ansible: Enable mistral dashboard when enable_mistral is yes https://review.openstack.org/406728 | 13:57 |
| *** yingjun has joined #openstack-kolla | 13:57 | |
| *** rhallisey has quit IRC | 13:57 | |
| *** alanmeadows has quit IRC | 14:00 | |
| *** alanmeadows has joined #openstack-kolla | 14:01 | |
| *** rhallisey has joined #openstack-kolla | 14:01 | |
| *** msimonin has quit IRC | 14:01 | |
| *** sdake has quit IRC | 14:02 | |
| *** DanyC has joined #openstack-kolla | 14:03 | |
| *** DanyC has quit IRC | 14:03 | |
| *** sdake has joined #openstack-kolla | 14:03 | |
| sdake | what is green monday | 14:03 |
| sdake | is that where the grinch stole christmas? :) | 14:03 |
| sdake | any excuse for a sale | 14:04 |
| *** goldyfruit has joined #openstack-kolla | 14:07 | |
| sdake | well i could buidl that zookeeper package in epel7 | 14:07 |
| sdake | I have the permissions in fedora to do so | 14:07 |
| sdake | but its typically considered rude - I'll ask the maintainer to do an epel7 build | 14:07 |
| *** msimonin has joined #openstack-kolla | 14:09 | |
| *** msimonin has quit IRC | 14:11 | |
| *** saneax-_-|AFK is now known as saneax | 14:13 | |
| Satya_ | Hi sdake | 14:16 |
| *** jheroux has joined #openstack-kolla | 14:16 | |
| Satya_ | i have a issue where i upgrade the neutron containers to trunk | 14:16 |
| Satya_ | after that i am not able to connect to old VMs where as able to connect new VM | 14:17 |
| Satya_ | tried restarting the old VM and seems its not able to get the IP | 14:17 |
| *** mdnadeem has quit IRC | 14:18 | |
| *** lukl has quit IRC | 14:18 | |
| *** zhugx has joined #openstack-kolla | 14:18 | |
| *** msimonin has joined #openstack-kolla | 14:19 | |
| *** mgiles has joined #openstack-kolla | 14:21 | |
| Satya_ | sdake | 14:22 |
| openstackgerrit | Sam Yaple proposed openstack/kolla: Add kafka image https://review.openstack.org/406874 | 14:22 |
| openstackgerrit | Sam Yaple proposed openstack/kolla: Add zookeeper image https://review.openstack.org/406878 | 14:22 |
| openstackgerrit | Sam Yaple proposed openstack/kolla: Add monasca images https://review.openstack.org/406422 | 14:22 |
| *** lukl has joined #openstack-kolla | 14:23 | |
| *** msimonin has quit IRC | 14:23 | |
| *** zhugx has quit IRC | 14:24 | |
| *** mdnadeem has joined #openstack-kolla | 14:26 | |
| *** Pavo has quit IRC | 14:29 | |
| *** ayoung has joined #openstack-kolla | 14:32 | |
| *** fguillot has joined #openstack-kolla | 14:32 | |
| *** srwilkers has joined #openstack-kolla | 14:33 | |
| srwilkers | morning p/ | 14:33 |
| sdake | morning srwilkers | 14:34 |
| srwilkers | hows it going sdake ? | 14:34 |
| *** Pavo has joined #openstack-kolla | 14:34 | |
| sdake | Satya_ not sure what yoru issue is | 14:34 |
| sdake | Satya_ my recommendation is never to run trunk unless doing developemnt | 14:34 |
| sdake | srwilkers good how about yourself | 14:34 |
| Satya_ | yes its my dev env | 14:34 |
| Satya_ | My Old VMs are not getting IP address | 14:35 |
| sdake | Satya_ i see - well if your a developer - debug ;) | 14:35 |
| srwilkers | sdake, not bad man. just waking up | 14:35 |
| sdake | Satya_ i'm not really sure what hte exact problem is | 14:35 |
| Satya_ | :P not getting the error | 14:35 |
| sdake | srwilkers my daughter had an all nighter project that 2 other kids in her school dumped on her | 14:35 |
| sdake | just like real life | 14:35 |
| srwilkers | haha. valuable life lesson right there | 14:36 |
| Satya_ | after i tried upgrading from 2 weeks old code to latest trunk code | 14:36 |
| Satya_ | i am seeing the old VMs are not getting the IP | 14:36 |
| sdake | Satya_ if your a dev, file a bug and debug ;) | 14:36 |
| sdake | Satya_ clealry something needs to be fixed there | 14:36 |
| Satya_ | it should be a bug? | 14:36 |
| srwilkers | Satya_, you can file a bug for it on launchpad to raise visibility so others can try to reproduce it | 14:37 |
| sdake | we can always close it invalid if after ou debug it you find it was user error | 14:37 |
| Satya_ | ok sure | 14:37 |
| sdake | Satya_ i'd not just file a bug tho, work on debugging it too :) | 14:37 |
| sdake | a bug is the first place to start - to record various parts of information | 14:37 |
| sdake | like logs/etc | 14:37 |
| Satya_ | ok sure raising | 14:38 |
| *** huzhengchuan has joined #openstack-kolla | 14:38 | |
| openstackgerrit | Mark Giles proposed openstack/kolla-kubernetes: Helm for installing other nova services https://review.openstack.org/409486 | 14:40 |
| *** jtriley has joined #openstack-kolla | 14:43 | |
| Satya_ | sdake yes trying to debug... from 3 days | 14:43 |
| Satya_ | still not getting any luck | 14:44 |
| Satya_ | it seems to me a namespace issue as restarting the host actually solve the problem | 14:44 |
| Satya_ | but i dont want to restart the host | 14:44 |
| Satya_ | raised the bug "https://bugs.launchpad.net/kolla/+bug/1649290" | 14:45 |
| openstack | Launchpad bug 1649290 in kolla "after kolla-ansible upgrade of neutron containers OLD vms fails to gett IP" [Undecided,New] | 14:45 |
| SamYaple | sdake: Satya_ this isn't a bug | 14:46 |
| *** msimonin has joined #openstack-kolla | 14:46 | |
| SamYaple | this is due to not having shared namespaces with docker setup properly | 14:46 |
| Jeffrey4l | SamYaple, any workaround? | 14:47 |
| SamYaple | Jeffrey4l: set it up correctly | 14:47 |
| *** jheroux has quit IRC | 14:47 | |
| Jeffrey4l | hrm didn't read the whole irc logs. the correct means? did Satya_ make some wrong? | 14:48 |
| Satya_ | SamYaple i tried the mountflag=shared for /run/netns | 14:48 |
| Satya_ | still that didnt solved my issue | 14:48 |
| SamYaple | Jeffrey4l: the shared mount namespace thing is saying "mount --make-shared /var/run/netns failed: Operation not permitted" | 14:49 |
| Jeffrey4l | fyi, i will try test this now. | 14:49 |
| SamYaple | Jeffrey4l: without fixing that,I can't help. I dont know anythign about oraclelinux, maybe pbourke can help | 14:49 |
| SamYaple | but regardless, this is not a kolla bug. perhaps additional documentation, but its outside of kolla/neutron | 14:49 |
| Jeffrey4l | I heard the same issue on oracle linux before. But never tried this. will try reproduce this now. | 14:50 |
| Satya_ | SamYaple it must not be a oraclelinux issue | 14:50 |
| Jeffrey4l | s/oracle linux/centos* | 14:50 |
| SamYaple | Satya_: configuration differs from distro to distro. What works for ubuntu and centos may not work with all the same things in oraclelinux | 14:50 |
| SamYaple | due to the kernel, and different version of different packages | 14:50 |
| Satya_ | hmm ok... | 14:51 |
| SamYaple | it very well could be an oraclelinux issue, but it is certainly not related to neutron or kolla contaienrs themselves since you reproduced it outside of that | 14:51 |
| *** jascott1_ has joined #openstack-kolla | 14:51 | |
| SamYaple | it could be the version of docker too | 14:51 |
| Satya_ | ok | 14:51 |
| Satya_ | i am using docker 1.10 | 14:52 |
| Satya_ | 1.10.3 | 14:52 |
| Jeffrey4l | could u post you environment out? | 14:52 |
| Jeffrey4l | os version, kernel version. | 14:52 |
| Satya_ | 4.1.12-61.1.22.el7uek.x86_64 #2 SMP Fri Dec 2 09:28:44 PST 2016 x86_64 x86_64 x86_64 GNU/Linux | 14:52 |
| Satya_ | Oracle Linux Server release 7.3 | 14:53 |
| Jeffrey4l | globals.yml file? | 14:53 |
| SamYaple | Jeffrey4l: this is reproducable outside of kolla configurations | 14:53 |
| SamYaple | Jeffrey4l: with a single docker run command | 14:53 |
| SamYaple | globals.yml shouldnt be needed | 14:53 |
| Jeffrey4l | could u explain more? SamYaple | 14:53 |
| Jeffrey4l | the more the better. want to know the network interface configuration. ;) | 14:54 |
| Jeffrey4l | it it is reproducable. did u try this on centos or ubuntu? | 14:54 |
| Jeffrey4l | it it/if it* | 14:54 |
| SamYaple | for me, I can run `docker run --privileged -v /run/:/run/:shared ubuntu ip netns add test` and then on the host `ip netns exec test ip a` and it works | 14:54 |
| SamYaple | for Satya_, he gets 'operation not permitted' | 14:55 |
| Jeffrey4l | the core issue is: he is trying to run upgrade. | 14:55 |
| SamYaple | no Jeffrey4l. also when the container is restarted this is present | 14:55 |
| Jeffrey4l | we need simulate this case. | 14:55 |
| SamYaple | the core issue is with no shared namespace support | 14:55 |
| *** jascott1_ has quit IRC | 14:56 | |
| Jeffrey4l | hrm. if there is no shared namespace support, he can never ping the vm, imo. | 14:56 |
| SamYaple | nope | 14:56 |
| SamYaple | thats not how it works | 14:56 |
| SamYaple | he can ping jsut fine, he just cant restart teh container | 14:56 |
| SamYaple | unless he also reboots the host | 14:56 |
| Satya_ | SamYaple i restarted the container before upgrade | 14:57 |
| Satya_ | still my VMs were rechable | 14:57 |
| Jeffrey4l | sound like the old issue, which can not remove the namespace ;( | 14:57 |
| Jeffrey4l | anyway, i am trying reproduce this on centos7 | 14:57 |
| SamYaple | Satya_: earlier you said it didnt work. '07:46:57 < Satya_> tried restarting the docker container but same issue' | 14:58 |
| Satya_ | i mean when i restarted the container after upgrade | 14:58 |
| Satya_ | but before upgrade i restarted the container and still able to reach the VMs | 14:59 |
| SamYaple | if deleting the network namespaces and then restarting the container works, this is an issue with mount propogation | 14:59 |
| SamYaple | i dont know how else to say that | 14:59 |
| *** gomarivera has joined #openstack-kolla | 15:00 | |
| SamYaple | its possible that its your version of docker too i suppose. maybe you found a bug there | 15:00 |
| Jeffrey4l | ( i doubt there are some issue in docker during restart. docker handle the shared /run in wrong ) just a suspect ;) | 15:00 |
| Satya_ | SamYaple 2 scenarios 1. restart the container before upgrade try to access VM works 2. upgrade the container to latest code VM is not reachable restart the container still vms not reachable | 15:01 |
| Jeffrey4l | how about restart docker engine? Satya_ | 15:01 |
| Jeffrey4l | did u try that? | 15:01 |
| Satya_ | tried that | 15:01 |
| Satya_ | yes | 15:01 |
| Satya_ | didnt helped | 15:01 |
| *** jheroux has joined #openstack-kolla | 15:01 | |
| Jeffrey4l | ok. so only restart host save the old vms, right? | 15:02 |
| Satya_ | yes | 15:03 |
| Jeffrey4l | how many node are u using? does the vm and neutron-dhcp-agent on the same physical node? | 15:03 |
| Satya_ | 2 | 15:04 |
| *** TxGirlGeek has joined #openstack-kolla | 15:04 | |
| Satya_ | yes | 15:04 |
| Jeffrey4l | ok. i am setting my AIO env to reproduce. | 15:04 |
| Satya_ | sure | 15:04 |
| -openstackstatus- NOTICE: The Gerrit service on review.openstack.org is restarting now to address acute performance issues, and will be back online momentarily. | 15:06 | |
| srwilkers | hey gomarivera | 15:07 |
| gomarivera | howdy srwilkers :D | 15:08 |
| *** saneax is now known as saneax-_-|AFK | 15:08 | |
| *** yingjun has quit IRC | 15:10 | |
| SamYaple | Satya_: ok. sorry that wasnt what i thought you said before | 15:10 |
| SamYaple | Satya_: i thought you said that just restarting the container also did the same thing | 15:10 |
| SamYaple | Satya_: my apologies | 15:10 |
| Satya_ | nope | 15:10 |
| *** duonghq has joined #openstack-kolla | 15:10 | |
| Satya_ | thats my mistake SamYaple i should be more explanatory | 15:10 |
| sdake | pbourke when triaging a bug it is typical to set the importance | 15:11 |
| *** ayoung is now known as ayoung_afk | 15:11 | |
| pbourke | sdake: ok | 15:11 |
| SamYaple | Satya_: yea if you can restart the container it is not the issue i was talking about at all | 15:11 |
| pbourke | will do | 15:11 |
| sdake | pbourke tx;) | 15:11 |
| SamYaple | Jeffrey4l: what code did you use to upgrade? what version ot what version? | 15:12 |
| Jeffrey4l | build from 2 weeks image to the latest. | 15:12 |
| *** msimonin has quit IRC | 15:12 | |
| Jeffrey4l | i am building the master branch images. | 15:12 |
| SamYaple | Jeffrey4l: just the image changed, or did the config files also change? | 15:12 |
| Jeffrey4l | hrm. just the image changed. | 15:12 |
| sdake | sup duonghq | 15:12 |
| Jeffrey4l | i am trying the same kolla-ansible version. | 15:13 |
| Jeffrey4l | ( master code ) | 15:13 |
| duonghq | sup sdake | 15:13 |
| Satya_ | i used the code at the day "07 11 2016" tried to upgrade to trunk | 15:13 |
| sdake | monday funday full of meetings | 15:13 |
| srwilkers | hey duonghq | 15:13 |
| duonghq | hi srwilkers | 15:13 |
| *** jtriley has quit IRC | 15:15 | |
| openstackgerrit | Jeffrey Zhang proposed openstack/kolla-ansible: Implement octavia ansible role https://review.openstack.org/404648 | 15:18 |
| *** prameswar has quit IRC | 15:19 | |
| *** karlamrhein has quit IRC | 15:23 | |
| *** karlamrhein has joined #openstack-kolla | 15:24 | |
| openstackgerrit | Paul Bourke (pbourke) proposed openstack/kolla-ansible: Update init-runonce to use openstack commands https://review.openstack.org/409832 | 15:29 |
| openstackgerrit | Paul Bourke (pbourke) proposed openstack/kolla-ansible: Update init-runonce to use openstack commands https://review.openstack.org/409832 | 15:30 |
| *** dave-mccowan has quit IRC | 15:31 | |
| *** saneax-_-|AFK is now known as saneax | 15:32 | |
| openstackgerrit | Paul Bourke (pbourke) proposed openstack/kolla-ansible: Update init-runonce to use openstack commands https://review.openstack.org/409832 | 15:32 |
| *** msimonin has joined #openstack-kolla | 15:33 | |
| *** jtriley has joined #openstack-kolla | 15:35 | |
| *** Bico_Fino has joined #openstack-kolla | 15:36 | |
| Bico_Fino | How can I test 3.0.2 newton? | 15:37 |
| egonzalez90 | Bico_Fino: pip install kolla==3.0.1 | 15:39 |
| *** gomarivera has quit IRC | 15:39 | |
| egonzalez90 | sorry, 3.0.2 is not yet available from pip | 15:40 |
| Bico_Fino | Yeap | 15:40 |
| Bico_Fino | Should I clone the master repo and build from it? | 15:40 |
| *** gomarivera has joined #openstack-kolla | 15:41 | |
| egonzalez90 | there is no tag 3.0.2 in git repo IIRC | 15:41 |
| Bico_Fino | but here is a branch for it? | 15:42 |
| openstackgerrit | Mark Giles proposed openstack/kolla-kubernetes: Helm creation of Nova services https://review.openstack.org/409473 | 15:42 |
| Bico_Fino | there* | 15:43 |
| egonzalez90 | no, branch are stable/*, tags are version 3.* | 15:44 |
| Bico_Fino | Wait for the tag then? :) | 15:45 |
| *** dave-mccowan has joined #openstack-kolla | 15:47 | |
| *** mdnadeem has quit IRC | 15:48 | |
| duonghq | bye guys | 15:50 |
| *** duonghq has quit IRC | 15:50 | |
| *** harlowja_at_home has joined #openstack-kolla | 15:52 | |
| egonzalez90 | Jeffrey4l: know the release calendar for stable version? only found the next release calendar | 15:53 |
| *** jascott1_ has joined #openstack-kolla | 15:53 | |
| *** prameswar has joined #openstack-kolla | 15:53 | |
| pbourke | egonzalez90: for your horizon custom policies patch - I think they recommend the files used in horizon match the ones used in the services | 15:56 |
| pbourke | egonzalez90: does it make sense to use the same source for both? | 15:56 |
| egonzalez90 | pbourke: in my previous prod, we used different policies per service and horizon | 15:57 |
| egonzalez90 | for example we do not allow creation of images from horizon | 15:57 |
| *** jascott1_ has quit IRC | 15:57 | |
| egonzalez90 | but i'm to use the same file | 15:58 |
| *** fguillot has quit IRC | 15:59 | |
| *** fguillot_ has joined #openstack-kolla | 15:59 | |
| sdake | anyone have that etherpad handy for the scale testing we did? | 15:59 |
| sdake | Jeffrey4l egonzalez90 ^^ | 16:00 |
| sdake | pbourke ^^ | 16:00 |
| sdake | nm found it | 16:00 |
| pbourke | egonzalez90: makes sense, thanks! | 16:00 |
| *** saneax is now known as saneax-_-|AFK | 16:01 | |
| *** zhubingbing has quit IRC | 16:06 | |
| *** zhubingbing has joined #openstack-kolla | 16:08 | |
| openstackgerrit | Mark Giles proposed openstack/kolla-kubernetes: Helm creation of Nova services https://review.openstack.org/409473 | 16:08 |
| *** gomarivera has quit IRC | 16:10 | |
| openstackgerrit | Sam Yaple proposed openstack/kolla: Add kafka image https://review.openstack.org/406874 | 16:10 |
| openstackgerrit | Sam Yaple proposed openstack/kolla: Add zookeeper image https://review.openstack.org/406878 | 16:10 |
| openstackgerrit | Sam Yaple proposed openstack/kolla: Add monasca images https://review.openstack.org/406422 | 16:10 |
| *** kristian__ has joined #openstack-kolla | 16:10 | |
| Jeffrey4l | egonzalez90, re stable release. no idea. i guess it depends on the team. | 16:13 |
| openstackgerrit | Sam Yaple proposed openstack/kolla: Add kafka image https://review.openstack.org/406874 | 16:13 |
| openstackgerrit | Sam Yaple proposed openstack/kolla: Add zookeeper image https://review.openstack.org/406878 | 16:13 |
| openstackgerrit | Sam Yaple proposed openstack/kolla: Add monasca images https://review.openstack.org/406422 | 16:13 |
| *** sayantan_ has joined #openstack-kolla | 16:14 | |
| *** saneax-_-|AFK is now known as saneax | 16:16 | |
| wznoinsk | hi, what's the story with ./ansible? any links or specs? | 16:18 |
| *** shardy has quit IRC | 16:19 | |
| Jeffrey4l | sdake, https://etherpad.openstack.org/p/kolla-N-midcycle-osic | 16:19 |
| wznoinsk | ok, ansible2.0 coming I guess? | 16:19 |
| Jeffrey4l | wznoinsk, sorry? ./ansible folder or? | 16:20 |
| wznoinsk | yes, https://github.com/openstack/kolla/tree/master | 16:20 |
| awiddersheim | so after the repo split | 16:20 |
| awiddersheim | how does kolla-ansible reference the rigth stuff in kolla? | 16:20 |
| awiddersheim | are you using git submodules? | 16:21 |
| Jeffrey4l | awiddersheim, we will make the same tag name for kolla and kolla-ansible. | 16:21 |
| Jeffrey4l | we do not using git submodules right now. | 16:21 |
| Jeffrey4l | there is no ansible folder in kolla repo now wznoinsk | 16:22 |
| wznoinsk | Jeffrey4l, yes, trying to find a spec about it, multi-node ansible? | 16:22 |
| Jeffrey4l | here is the origin ansible multi spec https://github.com/openstack/kolla/blob/master/specs/ansible-multi.rst | 16:23 |
| Jeffrey4l | but it have changed a lot i think. | 16:23 |
| Satya_ | SamYaple i ran docker inspect --format='{{.HostConfig.Privileged}}' neutron_dhcp_agent and the result is true | 16:23 |
| sdake | Jeffrey4l thanks | 16:27 |
| *** huzhengchuan has quit IRC | 16:27 | |
| *** Pavo has quit IRC | 16:29 | |
| Jeffrey4l | egonzalez90, still around? | 16:29 |
| egonzalez90 | Jeffrey4l: yes | 16:29 |
| *** Pavo has joined #openstack-kolla | 16:29 | |
| Jeffrey4l | egonzalez90, how will u handle the desinate patch? https://review.openstack.org/#/c/353261/ | 16:29 |
| egonzalez90 | don't know yet, what you recommend? working in a new bp, merging actual code and work on fixing the bug? | 16:31 |
| Jeffrey4l | i am ok with backport it or not merged it until all issue are fixed. if the latter one is the case, we need fix the horizon ( revert the change ) | 16:31 |
| egonzalez90 | i'm not a fan of merging something that do not work | 16:32 |
| Pavo | so others are having horizon issues? | 16:32 |
| Pavo | because I am | 16:32 |
| Jeffrey4l | hrm, my recommendation is backport it and fix it in another patch. ;) | 16:32 |
| Jeffrey4l | either way works for me. | 16:32 |
| egonzalez90 | Pavo: enable_designate undefined variable? thats the error we have ATM | 16:32 |
| *** sayantan_ has quit IRC | 16:32 | |
| *** sayantan_ has joined #openstack-kolla | 16:33 | |
| Pavo | yeah I am just talking errors in general not desinate | 16:33 |
| Pavo | http://paste.openstack.org/show/592127/ | 16:33 |
| Pavo | this is what I keep getting | 16:33 |
| *** sayantan_ has quit IRC | 16:33 | |
| Jeffrey4l | i only concern is: fix it make the blueprint management weird. | 16:33 |
| Jeffrey4l | Pavo, seems another issue. | 16:34 |
| Pavo | yeah | 16:34 |
| *** sayantan_ has joined #openstack-kolla | 16:34 | |
| Pavo | can not figure out what it is | 16:34 |
| Pavo | has been working fine but now for some reason its not | 16:34 |
| Pavo | this happens when trying to upload images using horizon, and deleteing instances | 16:35 |
| *** diogogmt has joined #openstack-kolla | 16:35 | |
| kfox1111 | morning. | 16:35 |
| Pavo | and btw I am using the offical kolla images | 16:36 |
| Jeffrey4l | Pavo, seems like a keystone v2/v3 issue | 16:37 |
| egonzalez90 | Jeffrey4l: you know how to manage an implemented blueprint so I can work on it? I'm in favour of fixing it all in a PS(push funtional code) | 16:37 |
| Jeffrey4l | we have to re-open the bp. | 16:38 |
| Pavo | Jeffrey4l anyway to narrow that down? | 16:38 |
| sdake | sup kfox1111 | 16:39 |
| Jeffrey4l | Pavo, could u post the keystone logs? | 16:39 |
| Jeffrey4l | sdake, could u give some advise fomr desinate patch egonzalez90 is working on? the desinate ansible part is merged during repo split, it is lost by mistake. and we also found the merge desinate is bust. | 16:40 |
| *** huzhengchuan has joined #openstack-kolla | 16:40 | |
| Jeffrey4l | we have two solution: 1. backport the code from kolla to kolla-ansible, then fix this in another ps. | 16:41 |
| Jeffrey4l | 2. fix the code first, then push to merge. | 16:41 |
| kfox1111 | sdake: hey. :) | 16:41 |
| Jeffrey4l | since the bp is marked as implement. personally 2) is weird to me. | 16:42 |
| Jeffrey4l | but 2) is more meaningless. we should not push bad code to the repo. | 16:42 |
| Jeffrey4l | any advice on this? sdake ^^ | 16:42 |
| Pavo | Jeffrey4l not seeing any errors in keystone logs | 16:43 |
| Jeffrey4l | Pavo, u will get some access log, with 404 return code, i think. | 16:43 |
| openstackgerrit | Paul Bourke (pbourke) proposed openstack/kolla-ansible: Update init-runonce to use openstack commands https://review.openstack.org/409832 | 16:46 |
| *** irtermite has quit IRC | 16:46 | |
| *** sayantan_ has quit IRC | 16:47 | |
| *** irtermite has joined #openstack-kolla | 16:47 | |
| *** sayantan_ has joined #openstack-kolla | 16:48 | |
| openstackgerrit | Mark Giles proposed openstack/kolla-kubernetes: Helm creation of Nova services https://review.openstack.org/409473 | 16:49 |
| *** sayantan_ has quit IRC | 16:49 | |
| *** sayantan_ has joined #openstack-kolla | 16:50 | |
| pbourke | Jeffrey4l: if designate is not ready any time soon we will need https://review.openstack.org/#/c/408714/ | 16:50 |
| *** msimonin has quit IRC | 16:51 | |
| Jeffrey4l | pbourke, yes. either way works for me. just want to clear which way we will choose. | 16:51 |
| pbourke | Jeffrey4l: well, I understand what you were saying about the backport thing, but at the end of the day if the role is non-functional I dont think it makes sense to merge | 16:52 |
| *** zhubingbing has quit IRC | 16:52 | |
| egonzalez90 | pbourke: was looking at the actual designate role and need a big rework to make it funtional | 16:53 |
| pbourke | right SamYaple said as much | 16:53 |
| Jeffrey4l | pbourke, OK. anyway, let's revert your patch first. and fix the designate before merge it. | 16:54 |
| Jeffrey4l | egonzalez90, is it possible to fixing designate in this cycle? | 16:54 |
| Jeffrey4l | seems it depends on lots of other service. | 16:55 |
| *** harlowja_at_home has quit IRC | 16:55 | |
| egonzalez90 | guess yes, have all christmas hollidays to do it | 16:55 |
| *** sayantan_ has quit IRC | 16:55 | |
| egonzalez90 | at the fist glance, the missing part is the backend, the bind docker image is already present in kolla(dont know if is fine) | 16:56 |
| *** cliles has joined #openstack-kolla | 16:56 | |
| egonzalez90 | Jeffrey4l: the endline for Ocata is at the end of January right? | 16:57 |
| *** tovin07_ has quit IRC | 16:57 | |
| Jeffrey4l | egonzalez90, yep | 16:58 |
| *** jrich523 has quit IRC | 16:59 | |
| *** pcaruana has quit IRC | 16:59 | |
| sdake | egonzalez90 ideally shouldn't work over christmas ;) | 16:59 |
| egonzalez90 | sdake: is when more spare time i have ;) | 17:00 |
| sdake | egonzalez90 wfm ;) | 17:00 |
| *** sayantan_ has joined #openstack-kolla | 17:01 | |
| awiddersheim | having issues iwth enable_designate not being defined | 17:02 |
| awiddersheim | https://github.com/openstack/kolla-ansible/blob/6d7c24d4f7418503411d8115a553028479d373eb/ansible/roles/horizon/tasks/start.yml#L10 | 17:02 |
| awiddersheim | gets used there but is never defined | 17:02 |
| egonzalez90 | awiddersheim: we are working on it, for the moment, you can add enable_designate no in globals | 17:02 |
| awiddersheim | ok | 17:02 |
| awiddersheim | thanks | 17:03 |
| awiddersheim | egonzalez90: nice work on the custom policy stuff btw | 17:03 |
| awiddersheim | very useful | 17:03 |
| *** rhallisey has quit IRC | 17:04 | |
| Jeffrey4l | fyi: revert deginate +w | 17:06 |
| srwilkers | sdake, working on this over christmas is how i can avoid my inlaws ;) | 17:06 |
| egonzalez90 | awiddersheim: if you wait some minutes, the fix for your issue is gitting merged | 17:06 |
| *** sacharya has joined #openstack-kolla | 17:06 | |
| *** saneax is now known as saneax-_-|AFK | 17:06 | |
| openstackgerrit | Merged openstack/kolla-ansible: Revert "Enable designate dashboard when enable_designate is yes" https://review.openstack.org/408714 | 17:07 |
| kfox1111 | srwilkers: hehe | 17:07 |
| Jeffrey4l | awiddersheim, egonzalez90 it is merged ^^ | 17:07 |
| *** jrich523 has joined #openstack-kolla | 17:08 | |
| *** egonzalez90 has quit IRC | 17:08 | |
| sdake | srwilkers lol - I like my inlaws ;) | 17:08 |
| *** gagehugo has quit IRC | 17:08 | |
| *** matrohon has quit IRC | 17:11 | |
| openstackgerrit | Merged openstack/kolla: Remove the unnecessary "openstack-zaqar-ui" https://review.openstack.org/407576 | 17:11 |
| *** egonzalez90 has joined #openstack-kolla | 17:17 | |
| *** egonzalez90 has quit IRC | 17:22 | |
| *** egonzalez has joined #openstack-kolla | 17:23 | |
| *** shardy has joined #openstack-kolla | 17:24 | |
| *** jascott1_ has joined #openstack-kolla | 17:25 | |
| Jeffrey4l | cloud u review this pbourke egonzalez https://review.openstack.org/404648 thanks. | 17:26 |
| *** Serlex has quit IRC | 17:36 | |
| *** ntpttr_ has joined #openstack-kolla | 17:37 | |
| *** sayantan_ has quit IRC | 17:41 | |
| *** sayantan_ has joined #openstack-kolla | 17:41 | |
| *** egonzalez has quit IRC | 17:43 | |
| *** athomas has quit IRC | 17:43 | |
| *** kristian__ has quit IRC | 17:46 | |
| *** jascott1_ has quit IRC | 17:51 | |
| v1k0d3n | hello everyone. | 17:51 |
| pbourke | Jeffrey4l: im a bit short on time for such a large patch right now | 17:51 |
| *** matrohon has joined #openstack-kolla | 17:51 | |
| pbourke | Jeffrey4l: I'll try review tomorrow if I get time | 17:51 |
| pbourke | Jeffrey4l: unless someone gets there first :) | 17:51 |
| Jeffrey4l | cool .thanks. | 17:51 |
| v1k0d3n | Jeffrey4l: are you standing in for inc0 while he's moving? | 17:53 |
| Jeffrey4l | v1k0d3n, yep. | 17:54 |
| v1k0d3n | great! | 17:54 |
| v1k0d3n | is there a completed spec for direction of kolla-kuberntes? | 17:54 |
| v1k0d3n | one where everyone can participate. i heard there was a vote. unfortunately we were not able to participate in this vote. | 17:54 |
| Jeffrey4l | check this https://github.com/openstack/kolla-kubernetes/blob/master/specs/kolla-kubernetes-arch.rst | 17:55 |
| v1k0d3n | i believe that changed some of what we expected based on talks in barcelona. | 17:55 |
| *** JoseMello has joined #openstack-kolla | 17:55 | |
| Jeffrey4l | we are implement kolla-k8a by using helm now. | 17:55 |
| v1k0d3n | so the helm implementation... | 17:56 |
| v1k0d3n | i was assuming by this spec that kolla would follow some helm fundumentals. | 17:56 |
| sdake | v1k0d3n only core reviewers participate on the vote of specs since it requires a +2 vote to approve - although the spec was open for review for over a week | 17:56 |
| v1k0d3n | *fundamentals | 17:56 |
| Jeffrey4l | v1k0d3n, sdake is leading the kolla-k8s project now. ;) | 17:57 |
| sdake | i wasn't at the talks in barcelona :) | 17:57 |
| sdake | I am not leading it, I am spearheading it :) | 17:57 |
| *** openstack has joined #openstack-kolla | 18:01 | |
| v1k0d3n | the charts are formatted in an opinionated way. | 18:01 |
| sdake | cool well i've got a meeting for 1 hr - and will be back to discuss further if you like | 18:01 |
| sdake | if the charts are opinionated please submit gi treviews to change em to what you like :) | 18:01 |
| v1k0d3n | i think i've brought up the concerns and implications | 18:01 |
| *** portdirect_away is now known as portdirect | 18:02 | |
| portdirect | o/ | 18:02 |
| v1k0d3n | i think there is a bit of unwillingness to change them because there is the concept to upload other dependancies | 18:02 |
| *** haplo37 has quit IRC | 18:02 | |
| v1k0d3n | i.e. operator, integrators. | 18:02 |
| sdake | bbi1hr | 18:02 |
| portdirect | v1k0d3n: what your saying does resonate with me | 18:03 |
| portdirect | v1k0d3n: there evedently is a disconnect between the work we are doing her and the helm/k8s community | 18:03 |
| v1k0d3n | do i need to document the differences? i mean...what would help Jeffrey4l | 18:03 |
| v1k0d3n | ? | 18:03 |
| *** pbourke has quit IRC | 18:04 | |
| v1k0d3n | the work being done is with the opinion of what's already existed in kolla, not from a fresh greenfield deployment model. | 18:04 |
| *** pbourke has joined #openstack-kolla | 18:04 | |
| Jeffrey4l | i am not working in the k8s project. so you'd better sync with sdake about this ;) | 18:04 |
| v1k0d3n | i.e. how a helm engineer/architect/developer would do it. | 18:04 |
| *** goldyfruit has quit IRC | 18:04 | |
| portdirect | v1k0d3n: and i think we need to ensure that are aritecure does align with helm more | 18:05 |
| v1k0d3n | when i do, i feel like i'm getting a bit railed. | 18:05 |
| sdake | ok - what exactly was the agreemeent in barcelona? | 18:05 |
| *** haplo37 has joined #openstack-kolla | 18:05 | |
| sdake | lets start there | 18:05 |
| sdake | (multitasking ftl) | 18:05 |
| v1k0d3n | this impacts not just our team (who is 100% open source focused...and wants to help with resources), but this is a fundamental helm architectural concern. | 18:06 |
| *** goldyfruit has joined #openstack-kolla | 18:06 | |
| v1k0d3n | align closely with a helm structure. | 18:06 |
| v1k0d3n | as these microservices get carried over, there are architectural issues with them. | 18:07 |
| v1k0d3n | i'm being polite, by bringing this up first...as opposed to -1 on all of them. | 18:07 |
| sdake | iterate the changes you want | 18:07 |
| sdake | nobody is a mindreader | 18:08 |
| v1k0d3n | not sure where that came from. | 18:08 |
| v1k0d3n | i have brought up the differences with kolla-kubernetes and other chart repos. | 18:08 |
| sdake | portdirect do you understand the deltas? | 18:09 |
| portdirect | I think I have an idea 2 sec | 18:09 |
| *** matrohon has quit IRC | 18:09 | |
| *** unicell has quit IRC | 18:10 | |
| portdirect | at a very basic level, we are not utilising/capitalising on the advantages that we can hget from k8s/helm for config: eg https://github.com/att-comdev/aic-helm/tree/master/keystone/templates/etc | 18:10 |
| * srwilkers reads scrollback | 18:10 | |
| portdirect | which would make it very easy to tune/customise keystone to a operators needs, eg support sml2/deferation, which the current direction we ar moving it would make very difficult | 18:11 |
| portdirect | *saml2/federation | 18:11 |
| *** haplo37 has quit IRC | 18:12 | |
| portdirect | also the management of secrets/config in values is much simpler to work with: https://github.com/att-comdev/aic-helm/blob/master/keystone/values.yaml | 18:12 |
| portdirect | i see an advantage in the way we are splitting up the maifests for microservices, but am now thinking that we may have gone too far with the speration into seperate helm charts for each microservice | 18:13 |
| portdirect | I've been thinkng about a way to capitaise on both, archs, but am currently at a bit of a loss | 18:14 |
| *** haplo37 has joined #openstack-kolla | 18:14 | |
| v1k0d3n | ^^ to that point, the general consumption for someone outside to just come in and use the project is extremely difficult without some serious rework. | 18:15 |
| v1k0d3n | one of the goals that got missed eventually making the charts consumable upstream in kubernetes/charts. | 18:15 |
| portdirect | I think we need to actually go out and get a meeting with the helm guys to see if they can help us on this | 18:15 |
| sdake | portdirect tend to agree | 18:16 |
| v1k0d3n | in this case, it would be extremely easy for ANYONE to pick and choose working components of openstack for their own use. they could deploy openstack, but perhaps they just want swift or cinder. | 18:16 |
| v1k0d3n | we've broadened the usefulness of the project. | 18:16 |
| v1k0d3n | and here...check this out... | 18:16 |
| v1k0d3n | look at these sitting PR's | 18:17 |
| v1k0d3n | https://github.com/kubernetes/charts/pulls | 18:17 |
| v1k0d3n | they are sitting there, many for changes...because many of them are not consumable as they are. | 18:17 |
| v1k0d3n | and kolla-kubernetes is unusable in current form if it would take the current approach. | 18:18 |
| srwilkers | this is one of my primary concerns. i agree with portdirect in that there could be an advantage for splitting things up, but if we're truly looking to incubate these charts upstream in the future, it's going to introduce a significant amount of work down the road for our community collectively | 18:18 |
| sdake | hard to say we would submit a pull erquest for kubernetes for kolla-kuberneetes into the charts dir | 18:18 |
| *** huzhengchuan has quit IRC | 18:18 | |
| srwilkers | wouldnt that be required to get our charts upstream? | 18:19 |
| sdake | seems a bit like jumping the gun | 18:19 |
| openstackgerrit | Mark Giles proposed openstack/kolla-kubernetes: Helm creation of Nova services https://review.openstack.org/409473 | 18:19 |
| sdake | sup mgiles :) | 18:19 |
| *** haplo37 has quit IRC | 18:19 | |
| *** harlowja has joined #openstack-kolla | 18:20 | |
| portdirect | there is some stuff in there that we should be considering though, eg https://github.com/kubernetes/helm/pull/1666, I think would help us a lot - and we could really make a strong case for it I think? | 18:20 |
| mgiles | hi sdake | 18:20 |
| sdake | the charts upstream location for kolla-kubernetes is in the kolla-kubernetes repository | 18:20 |
| sdake | if that changes in the future - that changes and we will adapt to that | 18:20 |
| sdake | but kfox1111 isn't around nor sbezverk_ so i don't know their opinions | 18:20 |
| *** sacharya_ has joined #openstack-kolla | 18:21 | |
| v1k0d3n | well, that stops our conversation. | 18:21 |
| v1k0d3n | :( | 18:21 |
| sdake | maybe they hae different opinions | 18:22 |
| sdake | i can't speak for everyone, just myself | 18:22 |
| *** haplo37 has joined #openstack-kolla | 18:22 | |
| sdake | we have enough problems to solve without working this system into something that can go into kubernetes/charts repo | 18:22 |
| srwilkers | its something we need to consider though. the spec we drafted that we keep referencing says the community will decide on whether to upstream them. to say thats open to discussion later if things change goes against that spec, doesnt it? or are we assuming thats not happening now? | 18:22 |
| *** prameswar has quit IRC | 18:22 | |
| v1k0d3n | sdake: this is what i'm saying...if the arch/implementation is correct...that works itself out. | 18:23 |
| *** sacharya has quit IRC | 18:23 | |
| sdake | i'd ask you to define correct ;) | 18:24 |
| sdake | implementations become correct eventually not from the beginning | 18:24 |
| portdirect | i tened to agree here - it's occured to me that we could use the -dry-run fautrue of helm to spit a service chart, eg aic/stackanetes into microservices potentintially much more easily that we could combine microservice charts into a useable blob? | 18:24 |
| v1k0d3n | ok. let me start with this. | 18:24 |
| v1k0d3n | rushing the current implementation for what was....and pushing a date for something outside of the ocata release cycle. | 18:25 |
| v1k0d3n | rather than doing each chart...as written and documented (very well in fact) in the helm repo docs. | 18:25 |
| v1k0d3n | consideration of values.yaml | 18:25 |
| v1k0d3n | and dependancies.yaml | 18:25 |
| v1k0d3n | those are totally missing... | 18:26 |
| v1k0d3n | because of being rushed. | 18:26 |
| sdake | not being rushed | 18:26 |
| v1k0d3n | ok. | 18:26 |
| v1k0d3n | well, i am pretty locked with the back and forth. | 18:26 |
| v1k0d3n | i give up. :) | 18:26 |
| v1k0d3n | bringing this issue up has gotten nowhere. | 18:26 |
| sdake | which issue specifically? | 18:27 |
| sdake | that the charts aren't "helm native"? | 18:27 |
| v1k0d3n | gtg. | 18:27 |
| portdirect | working from the initial implementation (pre helm) and shoehorning it into helm, vs starting from a blank sheet of paper | 18:28 |
| sdake | portdirect is that the specific issue? | 18:28 |
| sdake | portdirect software is iterated on, not replaced | 18:28 |
| sdake | atleast in openstack | 18:29 |
| portdirect | from my persepeive yes, but I cannot speak for anyone else. | 18:29 |
| *** Pavo has quit IRC | 18:29 | |
| sdake | its not like someone comes along with a new versoin of nova and says "nova we want you to remove your codebase and use our version" | 18:29 |
| sdake | that invalidates all of the work the core team has done on the original implementation of nova | 18:29 |
| sdake | and causes a whole bunch of chaos in the process | 18:30 |
| *** matrohon has joined #openstack-kolla | 18:30 | |
| SamYaple | thats a bad example sdake. that would be for the best | 18:30 |
| sdake | SamYaple lol | 18:30 |
| portdirect | no, and the original work was by no means invaild, what they did was get a working set of k8s templates whihc is a really impressive feat | 18:31 |
| *** sp_ has quit IRC | 18:31 | |
| kfox1111 | back. | 18:31 |
| sdake | portdirect tend to agree - working k8s setup | 18:31 |
| srwilkers | hey kfox1111 | 18:32 |
| portdirect | o/ | 18:32 |
| *** yatin_on_leave has quit IRC | 18:32 | |
| *** ayoung_afk is now known as ayoung | 18:32 | |
| *** fragatina has quit IRC | 18:33 | |
| kfox1111 | sdake: yeah. the issue is, as I understand it, is v1k0d3n wants a complete restart as of the start of helm, | 18:33 |
| kfox1111 | and do it in kubernetes/charts, not under the openstack kolla-kubernetes repo. | 18:33 |
| v1k0d3n | lol | 18:33 |
| kfox1111 | v1k0d3n: is that correct? | 18:33 |
| portdirect | so the issue from my perescective is noe we have those how we use them - we are changing from a jinja2 rendering to goptl - and trying to take advantage of the tools we can get in the process, and that i thing requies the archtecure of how we template the maifests to change refecting the tooling we are using | 18:33 |
| v1k0d3n | no kfox1111 | 18:33 |
| *** Pavo has joined #openstack-kolla | 18:34 | |
| kfox1111 | v1k0d3n: then I appologize. that is what I gathered from the discussion on greenfield. | 18:34 |
| kfox1111 | can you please clarify then? I do want to understand your request. | 18:34 |
| v1k0d3n | as charts are introduced into kolla-kubernetes, create values.yaml in the chart dir....just like helm suggests | 18:34 |
| *** jascott1_ has quit IRC | 18:34 | |
| v1k0d3n | create requirements.yaml as helm suggests | 18:34 |
| *** jascott1_ has joined #openstack-kolla | 18:34 | |
| v1k0d3n | and as the helm charts take shape... | 18:34 |
| kfox1111 | v1k0d3n: ah. so no code sharing like we're doing. | 18:34 |
| kfox1111 | just flattened out. | 18:34 |
| v1k0d3n | they will be easy to merge into a kubernetes/chart repo...because it's a complete usable framework in a single dir (as documented in helm). | 18:35 |
| kfox1111 | we're been doing it with prebuild to just limit the amount of maintenence we have to do for now for each one. once we get all the bits in, could we then reevaluate flattening? | 18:35 |
| *** JoseMello has quit IRC | 18:35 | |
| kfox1111 | it will be obvious at that point what will ongoing maintencence and what wont. | 18:36 |
| kfox1111 | during the bootstrapping though, its not so obvious. | 18:36 |
| *** JoseMello has joined #openstack-kolla | 18:36 | |
| kfox1111 | or alternately, we can get some features merged into helm to better support variable sharing? | 18:36 |
| v1k0d3n | that's where the timeline hurts the project. | 18:36 |
| portdirect | not entirely flattening kfox1111 - but I think loosing kolla-ansible (ie kolla gencofig) for now would help massively, and then we could potentially put it back in once we had stand alone charts that could be proped up without it. | 18:37 |
| kfox1111 | portdirect: my concern is in the common vars and the other vars that are shared between templates. | 18:37 |
| v1k0d3n | because deadlines are causing architectural decisions that are impacting our ability to use and participate. | 18:37 |
| *** unicell has joined #openstack-kolla | 18:37 | |
| v1k0d3n | oh my.... | 18:37 |
| kfox1111 | flatting makes those have to be touched all over the place if we ever have to change them. | 18:37 |
| v1k0d3n | there's a gap here. | 18:37 |
| kfox1111 | portdirect: I think we can do that, in kolla-kubernetes/helm/services. | 18:38 |
| kfox1111 | portdirect: everyone's been focused on the under layers and no one's yet payed too much attention to the over layer. | 18:38 |
| v1k0d3n | kfox1111: i think it may help to get an extra set of eyes on the project... | 18:39 |
| kfox1111 | I was thinking aic-helm's glance service could be rewritten to use the microservices and placed there? | 18:39 |
| v1k0d3n | someone who wrote or contributes heavily to helm. | 18:39 |
| v1k0d3n | kfox1111: question.... | 18:39 |
| v1k0d3n | what is the main focus....microservices? | 18:39 |
| v1k0d3n | architecture? | 18:39 |
| v1k0d3n | usability/portability? | 18:39 |
| portdirect | v1k0d3n: do you know who would be best? I was thinking about reaching out to lachie83, to see if he could help? | 18:40 |
| v1k0d3n | microservices (or your vision of it) is warping the project into something that is not very portable. i know you think it is, but as portdirect explained...it's not. | 18:40 |
| v1k0d3n | i spoke with gabe yesterday... | 18:40 |
| v1k0d3n | and i think getting matt butcher and lachlan on this would be extremely helpful. | 18:41 |
| v1k0d3n | there would be a lot of pitfalls avoided i think. | 18:41 |
| sdake | microservices first is my vision not kfox1111 's | 18:41 |
| sdake | or maybe its a joint vision | 18:41 |
| sdake | who knows :) | 18:41 |
| sdake | thats how these things work - chaotically | 18:42 |
| sdake | nobody is out to get you v1k0d3n | 18:42 |
| sdake | we want you to participate | 18:42 |
| *** sayantan_ has quit IRC | 18:42 | |
| kfox1111 | sorry. someone showed up in my office. back. | 18:42 |
| v1k0d3n | so how do i use nuetron l3 agent all by itself without neutron? | 18:43 |
| openstackgerrit | Francisco Edigleison Barbosa proposed openstack/kolla-ansible: Allow cinder-volume to be configured to use HNAS iscsi https://review.openstack.org/409883 | 18:43 |
| kfox1111 | v1k0d3n: more eyes always good. "to enough eyes, all bugs are shallow" :) | 18:43 |
| v1k0d3n | this starts getting into religion on what a microservice unit is. | 18:43 |
| kfox1111 | v1k0d3n: focus on microservices is its the bottom of the perimid, as per the spec. | 18:43 |
| *** sp_ has joined #openstack-kolla | 18:43 | |
| kfox1111 | its building blocks the rest of the system is built on top of. | 18:43 |
| v1k0d3n | sdake: i'm not worried about anyone "out to get me", but i want everyone to realize that going against the grain is uncomfortable and time consuming. | 18:44 |
| v1k0d3n | i'm not trying to throw rocks at it because i like controversy. | 18:44 |
| kfox1111 | v1k0d3n: I think you believe its not portable, but I think we disagree there. | 18:44 |
| sdake | it is time consuming for everyone involved | 18:44 |
| SamYaple | lies! i know v1k0d3n! thats what he is doing! | 18:44 |
| v1k0d3n | SamYaple: :) | 18:44 |
| srwilkers | busted | 18:44 |
| sdake | SamYaple you troll you :) | 18:44 |
| kfox1111 | sdake: to v1k0d3n: agreed. we're not out to get you. we're actuaaly workign very hard to cover your use cases. | 18:44 |
| * v1k0d3n evil smirk | 18:44 | |
| v1k0d3n | actually SamYaple is awesome...i want to get the container image sizes down too! :) | 18:45 |
| kfox1111 | v1k0d3n: say you deploy neutron with packstack. | 18:45 |
| v1k0d3n | not thinnking packstack | 18:45 |
| v1k0d3n | that's the point... | 18:45 |
| portdirect | everyone wants that... | 18:45 |
| kfox1111 | v1k0d3n: you can then rpm -e neutron-l3-agent and helm install kolla/neutron-l3-agent. | 18:45 |
| SamYaple | idk portdirect. i heard someone saying they wanted to make them larger | 18:45 |
| *** yatin_on_leave has joined #openstack-kolla | 18:45 | |
| kfox1111 | v1k0d3n: I know your not thinkging about it. but I am. | 18:45 |
| *** gagehugo has joined #openstack-kolla | 18:45 | |
| sdake | SamYaple bigger is better doncha know :) | 18:45 |
| sdake | supersize ftw | 18:46 |
| v1k0d3n | tbh kfox1111 i'm thinking about how helm would do it :) | 18:46 |
| kfox1111 | v1k0d3n: I'm contributing code to make it easier for folks to migrate from some existing system, to kolla-kubernetes. | 18:46 |
| v1k0d3n | and what that all looks like only. | 18:46 |
| SamYaple | one massive container it is! | 18:46 |
| kfox1111 | v1k0d3n: yeah. your goal is a subset of mine. | 18:46 |
| v1k0d3n | how would a helm developer implement this. | 18:46 |
| v1k0d3n | lol | 18:46 |
| v1k0d3n | ok | 18:46 |
| srwilkers | +1 SamYaple | 18:46 |
| portdirect | SamYaple: https://microbadger.com/images/port/openstack-neutron-manager <-- this would be nice (though I'm obviously biased) | 18:46 |
| *** jascott__ has joined #openstack-kolla | 18:46 | |
| v1k0d3n | so i think we've uncovered the issue... | 18:46 |
| *** jascott1_ has quit IRC | 18:46 | |
| SamYaple | portdirect: what am i looking at | 18:47 |
| kfox1111 | v1k0d3n: please summerize your undertanding of the issue. | 18:47 |
| portdirect | SamYaple: 65mb for neutron api | 18:47 |
| SamYaple | portdirect: oh. have we not talked? I built 40MB keystone, and I bet I could beat your 65MB neutron-api :) | 18:48 |
| v1k0d3n | asking for changes is a disruptive process, and kolla-kubernetes first priority is microservices... | 18:48 |
| SamYaple | portdirect: with ubuntu! | 18:48 |
| sdake | v1k0d3n you don't ask for changes, you submit them | 18:48 |
| * portdirect feels very small | 18:48 | |
| v1k0d3n | at the cost of a helm deployment that is consumable to the masses. | 18:48 |
| *** vhosakot has joined #openstack-kolla | 18:48 | |
| v1k0d3n | i'm discussing first. | 18:48 |
| portdirect | SamYaple: wow - thats awesome :) | 18:49 |
| sdake | we do that in the review process | 18:49 |
| v1k0d3n | we did it in the spec | 18:49 |
| kfox1111 | v1k0d3n: everyone's free to contribute in the way that furthers their own goals. | 18:49 |
| v1k0d3n | but i think there was a vote after the spec | 18:49 |
| v1k0d3n | am...i...wrong? | 18:49 |
| sdake | we also do so in the review process | 18:49 |
| SamYaple | portdirect: i can bring it to kolla, but the images need major reworks for that. its doable for sure | 18:49 |
| kfox1111 | if you want the focus on helm deployment consumable to the masses, thats where your devs should be contributing? | 18:49 |
| sdake | indeed there was a vote | 18:49 |
| *** shardy is now known as shardy_afk | 18:50 | |
| kfox1111 | by your definition of "helm deployment consumable to the masses" | 18:50 |
| *** yatin_on_leave has quit IRC | 18:50 | |
| kfox1111 | we set out a framework where all parties could coixist. | 18:50 |
| SamYaple | portdirect: it does take build time longer. so its build images (time it takes now) then recompute layers (additional time). The final images are tiiiiiny | 18:50 |
| sdake | kfox1111 +2 | 18:50 |
| portdirect | SamYaple: yeah, actuallky I rememeber us chatting now :) | 18:50 |
| kfox1111 | the framework doesn't state who puts in dev time at what point though. | 18:50 |
| v1k0d3n | ok. | 18:50 |
| v1k0d3n | i brought it up. | 18:51 |
| v1k0d3n | it's over | 18:51 |
| portdirect | SamYaple: I tried a rough pass at that https://review.openstack.org/#/c/392115/ <<-worked locally | 18:51 |
| openstackgerrit | Mark Giles proposed openstack/kolla-kubernetes: Helm creation of Nova services https://review.openstack.org/409473 | 18:51 |
| kfox1111 | I'd be happy to review contributions that work on the service level helm packages. | 18:51 |
| kfox1111 | which i belive is where your interests are? | 18:52 |
| sdake | i'd be happy to make a blueprint for such work as well if people want to workon that | 18:52 |
| kfox1111 | sdake: +1 | 18:52 |
| v1k0d3n | i think you guys are doing it how you need to. | 18:52 |
| v1k0d3n | we just have architectural differences, and different goals. | 18:52 |
| v1k0d3n | i won't cause any more issues. | 18:53 |
| portdirect | v1k0d3n: would you have time to write a bp? | 18:53 |
| kfox1111 | I think the architectural differences are not great. as for goals, my goal is to support your use cases. | 18:53 |
| v1k0d3n | hate having to bring up. | 18:53 |
| v1k0d3n | i really don't feel like it fits portdirect | 18:53 |
| kfox1111 | but my personal contributions have focus on the bottom most layers at the moment. | 18:53 |
| kfox1111 | that would allow aic-helm developers though to quickly match up their contributations with kolla-kubernetes contributions though, so we all can work together. | 18:54 |
| sdake | kfox1111 i dont see how you cuold possibly write the whole codebase yourself ;) | 18:54 |
| kfox1111 | sdake: right. | 18:54 |
| sdake | that is why we are a team | 18:54 |
| kfox1111 | +1 | 18:54 |
| v1k0d3n | kfox1111: hi | 18:54 |
| kfox1111 | we work as a team, allowing each dev to work on the bits they, or their org, cares about most. | 18:54 |
| v1k0d3n | aic-helm here. | 18:54 |
| v1k0d3n | :) | 18:55 |
| srwilkers | so before this conversation wraps up, i just want to go back to the spec for a second | 18:55 |
| v1k0d3n | trying. | 18:55 |
| v1k0d3n | but it won't work | 18:55 |
| SamYaple | portdirect: yea. i actually did the exact same thing with docker_squash about a year ago | 18:55 |
| SamYaple | portdirect: abandonded PS of course | 18:55 |
| v1k0d3n | i think we need to publicly document why there are differences. | 18:55 |
| sdake | SamYaple figured out how to merge his code base into kolla-ansible | 18:55 |
| sdake | SamYaple it was hard, wasnt it? | 18:55 |
| kfox1111 | v1k0d3n: I'd be happy to discuss the issues on why you don't believe architecturally why aic-helm couldn't reusie kolla-kubernetes microservice packages. | 18:56 |
| SamYaple | sdake: it required me to pretty radically change the way i was doing things to conform to the community | 18:56 |
| sdake | but in the end, everyone won | 18:56 |
| kfox1111 | v1k0d3n: if there is a problem with that, then we really need to fix it. | 18:56 |
| v1k0d3n | kfox1111: please read portdirect's statements above. | 18:56 |
| SamYaple | but luckily, I can get back 90% of what I want in kolla-salt | 18:56 |
| kfox1111 | srwilkers: sure. | 18:56 |
| v1k0d3n | kfox1111: you said that...but i don't feel a willingness for that. | 18:56 |
| SamYaple | for example, I always wanted config files to live on the host at /etc/keystone | 18:56 |
| v1k0d3n | otherwise i would start merging in. | 18:56 |
| kfox1111 | v1k0d3n: that kolla-kubernetes doesn't ship config stuff yet? | 18:56 |
| SamYaple | but i put aside that to conform to community | 18:56 |
| kfox1111 | v1k0d3n: why can't aic-helm ship them exactly as they are today for now? | 18:57 |
| SamYaple | now I can do that with kolla-salt | 18:57 |
| srwilkers | i feel this is a moot point because we keep circling around back to it, but the spec should be the single source of truth here. i feel a large part of the disconnect here and why there's concern is that the spec specifically mentions we'll have helm charts for each openstack service. in terms of our project, defined by the terms set forth in review, a "service" by definition is what's been done in other openstack helm | 18:57 |
| srwilkers | implementations, not just our own. in fact, we link to SAP's repo specifically in the spec for this | 18:57 |
| v1k0d3n | so SamYaple and sdake does community include SAP and Stackanetes, and other projects using helm? | 18:58 |
| v1k0d3n | look...i'm just saying that this project is the ONLY project i've seen doing this. | 18:58 |
| srwilkers | it doesnt make mention of microservices for helm, and i feel thats why this conversation has been and will likely continue to happen | 18:58 |
| v1k0d3n | i spent all weekend, looking over countless chart repos, and they all look similar | 18:58 |
| srwilkers | if we're to use the spec as the single source of truth, which we should | 18:58 |
| sdake | srwilkers please have a look at this: https://github.com/openstack/kolla-kubernetes/blob/master/specs/README.rst | 18:59 |
| sdake | srwilkers inc0 asked me to put that in the specs dir | 18:59 |
| kfox1111 | v1k0d3n: just because they do it differently doesn't mean we're wrong in doing something different? | 18:59 |
| v1k0d3n | unless they are helmc charts...in which they somewhat resemble this format, but that is not what's in kubernetes/helm or kubernetes/charts. | 18:59 |
| sdake | srwilkers and he was first to review it and +2 it | 18:59 |
| SamYaple | v1k0d3n: i have no clue what helm is or how it relates to Kolla, so im of no help here | 18:59 |
| kfox1111 | srwilkers: the microservices are around layer2. the templates for k8s objects. | 18:59 |
| v1k0d3n | kfox1111 please just listening to me for a second... | 18:59 |
| v1k0d3n | i'm just saying does community count these others? | 18:59 |
| srwilkers | sdake, i agree. however, we keep referencing the spec when we're discussing differing opinions and why things are done. we cant both say "its the path set forward because its in the spec" and "its in the spec, but the spec isn't set in stone" | 18:59 |
| srwilkers | contradictory in nature | 18:59 |
| sdake | srwilkers this is why specs are evil :) | 19:00 |
| v1k0d3n | SamYaple: i know...just point out about the merge in this case. | 19:00 |
| jascott__ | SamYaple helm its the official "pkg manager" for k8s | 19:00 |
| sdake | srwilkers i never wanted a spec, everyone else did | 19:00 |
| *** sayantan_ has joined #openstack-kolla | 19:00 | |
| v1k0d3n | very difficult as portdirect alluded to. | 19:00 |
| SamYaple | jascott__: that does not help :P, i don't know kubernetes at all either | 19:00 |
| v1k0d3n | sdake: in other projects there are specs. we work on many projects. | 19:01 |
| sdake | v1k0d3n yes - kolla operates without specs because its a young project | 19:01 |
| jascott__ | SamYaple: oh ok ;) its scheduler for docker basically | 19:01 |
| kfox1111 | v1k0d3n: helm is brand new. 2.0 was just relaeased a complete rewrite, so much closer really to a 1.0. what makes you believe everyone knows the best way to use such a new and innovative tool right out of the gate? | 19:01 |
| sdake | specs result in these types of arguments | 19:01 |
| sdake | that is why inc0 asked me to put the following in the repo: | 19:01 |
| SamYaple | sdake: kolla has been around for 1/3 the life of openstack | 19:01 |
| *** tonanhngo has joined #openstack-kolla | 19:01 | |
| sdake | The single source of truth for development work is recorded in blueprints as specifications are immutable once committed to the repository. Blueprints on the other hand are dynamic and may be changed at any time along with the codebase. | 19:02 |
| portdirect | SamYaple: i think you've just hit on the findimaental issue, none of us know k8s as well as we should, we need to work with peeps that do, eg CoreOS and get their resoning for whay they did things the way they did | 19:02 |
| v1k0d3n | kfox1111: this goes to my point of reaching out to that community; which your response is "i don't have time". | 19:02 |
| v1k0d3n | look...this is going nowhere. | 19:02 |
| kfox1111 | the same happend with docker. containers became "hot" and then everyone used them for things that they really weren't suited for. | 19:02 |
| v1k0d3n | religions are set | 19:02 |
| kfox1111 | but after a while, it settled out into some really good practices. | 19:02 |
| SamYaple | kfox1111: i have only ever used containers as an imaging tool to wrap up my chroots | 19:02 |
| kfox1111 | v1k0d3n: right.k I do believe we shoudl reach out to the community. who is going to do that? | 19:02 |
| *** tonanhngo_ has joined #openstack-kolla | 19:03 | |
| *** fragatina has joined #openstack-kolla | 19:03 | |
| *** matrohon has quit IRC | 19:03 | |
| kfox1111 | SamYaple: some used the containers as vm's and instantiate, then install crap into them. a counter best practice now. | 19:03 |
| sdake | SamYaple have a look at how old that file is in git | 19:03 |
| SamYaple | sdake: ? | 19:03 |
| *** yatin_on_leave has joined #openstack-kolla | 19:03 | |
| rafael_sisweb | Hey guys. I'm with some http 500 with horizon when using brazilian language.. on horizon logs got "AttributeError: DjangoTranslation instance has no attribute 'plural". I hapens only with some endpoints. | 19:03 |
| rafael_sisweb | Is there any way to disable multi language support on horizon? or has any fix to this "DjangoTranslation" issue? | 19:03 |
| kfox1111 | v1k0d3n: sorry, but you don't get to dictate where my time goes. | 19:03 |
| kfox1111 | v1k0d3n: thats against the openstack rules. | 19:04 |
| SamYaple | kfox1111: haha! thats openstack-ansible. and its why i hate that | 19:04 |
| sdake | SamYaple that readme.rst that inc0 +2'ed first to review, and was committed nov 9th | 19:04 |
| kfox1111 | v1k0d3n: I would love to have more time to do everything emaginable that would helm kolla | 19:04 |
| portdirect | kfox1111: I could do that (and have begin to, for various things I'minvovled in) - but I'm not core in this, so dont know if its really appropriate | 19:04 |
| v1k0d3n | kfox1111: thanks for that | 19:04 |
| sdake | portdirect CORE REVIEWER | 19:04 |
| kfox1111 | v1k0d3n: but thats not real. I have to make my own priorities, and my current priority is getting something in the kolla-kubernetes space to actually work. | 19:04 |
| SamYaple | sdake: i may have missed a message somewhere | 19:04 |
| sdake | core reviewers review code | 19:04 |
| sdake | anyone can contibute | 19:04 |
| v1k0d3n | funny you bring up dictate after fridays convo. | 19:05 |
| sdake | SamYaple https://github.com/openstack/kolla-kubernetes/blob/master/specs/README.rst | 19:05 |
| kfox1111 | v1k0d3n: tehn priority number 2 is to get it in line with convention wherever possible. | 19:05 |
| kfox1111 | and contribute back to upstreams | 19:05 |
| v1k0d3n | i take my priorities outside of os projects. | 19:05 |
| kfox1111 | we're bootstrapping still. | 19:05 |
| v1k0d3n | look. i know where we stand. | 19:05 |
| v1k0d3n | there's no path that i see until helm 2.0 is considered | 19:05 |
| *** tonanhngo has quit IRC | 19:06 | |
| kfox1111 | v1k0d3n: we just seem to be going around and around on what you believe I shoudl be focused on. :/ | 19:06 |
| SamYaple | sdake: i may have missed something. i agree with that message though. I dont know why you are linking it to me | 19:06 |
| v1k0d3n | (which was brought up in BARCELONA AND IS IN THE SPEC!) | 19:06 |
| kfox1111 | sorry, but getting pretty frustrated about that. | 19:06 |
| v1k0d3n | so kfox1111 if you're going with "like helm 1.0" you should review the spec | 19:06 |
| *** JoseMello has quit IRC | 19:06 | |
| sdake | SamYaple you said kolla has been around for 1/3rd the lifetime of openstack, however the kolla-kubernetes deliverable has not | 19:06 |
| kfox1111 | v1k0d3n: how is helm 2 not being considered? | 19:06 |
| v1k0d3n | i send a bunch of links...on friday | 19:07 |
| sbezverk_ | v1k0d3n: who said anything about using helm 1.0?? we use helm 2.0.2 | 19:07 |
| SamYaple | sdake: ohhhh right. but we still dont do specs in kolla often was my only point | 19:07 |
| v1k0d3n | values.yaml, depends.yamls. concepts of base64, concepts of bring your own secrets, there are lots of gaps honestly. | 19:07 |
| sdake | SamYaple right - and this conversation is the precise reason why:) | 19:07 |
| v1k0d3n | sbezverk_: kfox1111 said that the layout is more like helm 1.0 (or helmc...which is is). | 19:08 |
| v1k0d3n | not helm 2.0 layout/architecture. | 19:08 |
| Satya_ | jeffrey4l got any luck with centos | 19:08 |
| portdirect | kfox1111, we work we are producing is very similar to the helm charts from 1.0, even though we are using 2.0 | 19:08 |
| v1k0d3n | ^^^ yup | 19:08 |
| v1k0d3n | that architecture is the missing link here gentlemen | 19:09 |
| v1k0d3n | just saying | 19:09 |
| jascott__ | so whats the difference in helm 1 vs 2 as far as architecure? | 19:09 |
| *** rafael_sisweb has quit IRC | 19:09 | |
| jascott__ | i have seen nothing | 19:09 |
| kfox1111 | v1k0d3n: how do we allow launching just neutron-l3-agent without the rest of neutron if we don't split it up into microservices? | 19:09 |
| v1k0d3n | one sec. jascott__ links incoming. | 19:09 |
| v1k0d3n | there's a lot | 19:09 |
| kfox1111 | v1k0d3n: your saying we | 19:09 |
| *** rafael_sisweb has joined #openstack-kolla | 19:09 | |
| v1k0d3n | ok | 19:09 |
| kfox1111 | are designing around helm 1.0. I'm saying we're not . | 19:09 |
| jascott__ | ive read the docs | 19:09 |
| kfox1111 | were designing around functionality, | 19:09 |
| kfox1111 | and it happens to look kind of like helm 1.0. | 19:09 |
| jascott__ | oh you mean Helm-classic? | 19:10 |
| kfox1111 | but the goal is the functionality, not the look of the repo. | 19:10 |
| kfox1111 | I really could care less what the repo looks like so long as the use cases are met. | 19:11 |
| v1k0d3n | jascott__: yes helm classiv | 19:11 |
| jascott__ | oh yeah they are diff | 19:11 |
| jascott__ | thats not the same as 1.0 and 2.0 | 19:11 |
| jascott__ | thats why they changed it to helm-classic | 19:11 |
| v1k0d3n | right..true...i was repeating kfox1111 1.0 | 19:11 |
| jascott__ | we have been using the new helm the whole time | 19:12 |
| jascott__ | oh ok | 19:12 |
| kfox1111 | yup. helm 2.0. | 19:12 |
| v1k0d3n | helmc is dir/ with chart.yaml and template folder | 19:12 |
| sdake | we adapt to our upstreams | 19:12 |
| kfox1111 | was just saying, as helm-classic != helm, helm 2.0 is closer to a 1.0 release in principal. | 19:12 |
| v1k0d3n | helm 2.0 has all sorts of added features... | 19:12 |
| kfox1111 | as its brand new code. | 19:12 |
| kfox1111 | brand new phylosophy in some cases. | 19:12 |
| kfox1111 | so very little experience with folks trying to base things on the new phylosophy. | 19:13 |
| sdake | kfox1111 ure spell checker is broken -> principal -> princple | 19:13 |
| v1k0d3n | ingesting values.yaml, requirements.yaml, (still templates) | 19:13 |
| kfox1111 | sdake: yeah. I have a bad spell checker. :/ | 19:13 |
| sdake | ;=) | 19:13 |
| portdirect | sdake, least he has one :) (my smelling is awlsys perfect) | 19:13 |
| jascott__ | my spell checking neural net is "adversarial" | 19:13 |
| kfox1111 | jascott__: hehe. | 19:14 |
| v1k0d3n | the values is where your defaults reside...but those are extremely easy for someone to modify or better yet...bring their own. | 19:14 |
| sdake | jascott__ mine is cohesive ;) | 19:14 |
| kfox1111 | v1k0d3n: yup. we're using values as defaults. | 19:14 |
| kfox1111 | and its easy to helm install foo --values <users values here> | 19:14 |
| kfox1111 | I really like that about helm. | 19:14 |
| portdirect | yeah - this is wehre my concer is kfox1111, i know you say you dont care how the repo looks, but i do: | 19:14 |
| v1k0d3n | kfox1111: to be fair...there's a lot of hardcoded in "your" charts as well. | 19:15 |
| v1k0d3n | (since you got me on "we" earlier....) | 19:15 |
| v1k0d3n | look...this is uncomfotable. | 19:15 |
| kfox1111 | portdirect: I'm ok changing up the repo so long as someone can chart a clear path forward that doesn't break functionality. | 19:15 |
| v1k0d3n | dangit...spell check! | 19:15 |
| portdirect | i want to be able to take a chart and warp it to my needs easily, and the current ayout means a new user could not do that | 19:15 |
| v1k0d3n | anyway...i don't want to bring this up anymore. it's too disruptive to the team, and too disruptive to the community. | 19:15 |
| v1k0d3n | i said my peace. it's not going to change. i'm not going to -1 reviews. | 19:16 |
| sdake | v1k0d3n we want you to be part of that team | 19:16 |
| kfox1111 | portdirect: can you propose an example of that, so it is really clear what the use case is? | 19:16 |
| v1k0d3n | i brought up, and it's too far for me personally. | 19:16 |
| jascott__ | v1k0d3n how bout a list of perceived gaps? (pardon if you have already shared one) | 19:16 |
| v1k0d3n | sdake: :) me too. but i have to back it. i'm telling you what it would take...i need a technical path. no excuses why it's good as is for "us to come sit at the table". | 19:17 |
| *** ccesario has quit IRC | 19:17 | |
| v1k0d3n | well, jascott__one thing would be super easy... | 19:17 |
| v1k0d3n | using helm...do it helm's way. | 19:17 |
| v1k0d3n | period. | 19:17 |
| portdirect | np, the keystone one I had earlier was good for that: witht the work I'm doing: if I wanted to use federated auth (and I do) I've been having diffucilty seing how I would incoprorate that into what we are doing | 19:17 |
| jascott__ | right thats pretty vague | 19:17 |
| v1k0d3n | actually it's not though. | 19:18 |
| kfox1111 | v1k0d3n: to me, it still feels like your saying, "your not using helm the way I expect. do it my way" :/ | 19:18 |
| jascott__ | im asking for a list of grievances nailed to the kolla church door that people can reflect on and discuss | 19:18 |
| portdirect | it would require work to kolla-ansible, kolla, and kolla-k8s - wehreas one of the other projects would allow me to get stuck in and make those changes with minimal effore quickly i think | 19:18 |
| v1k0d3n | kfox1111: let's start listening to each other and stop putting words into each others mouths here.... | 19:18 |
| kfox1111 | you could be right. we're using it wrong. but I could be right, and helm just doens't have the features we need yet and that will become obvious to others too in time. | 19:19 |
| jascott__ | one here and there in irc is not getting us far | 19:19 |
| v1k0d3n | what i said exactly is, this is the only project using charts in this manner. | 19:19 |
| sdake | v1k0d3n +1 listen to each other | 19:19 |
| kfox1111 | v1k0d3n: k. | 19:19 |
| kfox1111 | trying to listen. | 19:19 |
| portdirect | jascott__: yeah :( | 19:19 |
| v1k0d3n | over time i think you would find it would be hard for others to use if they have other helm things they deploy. | 19:19 |
| v1k0d3n | and that's my point | 19:19 |
| v1k0d3n | if you offer me a helm project...the likelyhood is that i'm a helm user and look forward to this. | 19:20 |
| kfox1111 | helm's built so that deps can be external. | 19:20 |
| sdake | jascott__ sounds like a fine approach to me - on the mailing list | 19:20 |
| kfox1111 | the build packages can be depended on in their own packages. | 19:20 |
| sdake | jascott__ care to start the thread? | 19:20 |
| kfox1111 | is that hard to deal with? | 19:20 |
| v1k0d3n | it will be very frustrating for me to find that i can't just add this to my current helm repo and consume. | 19:20 |
| kfox1111 | or are you very concerned their own build tooling cant build kolla-kubernetes packages? | 19:20 |
| kfox1111 | v1k0d3n: that is the exact use case I'm tryign to cover. | 19:21 |
| kfox1111 | being able to build the kolla-kubernetes packages, and add them to a repo for consumption. | 19:21 |
| jascott__ | sdake I can but dont know what to put in, assumed one of these very passionate and intelligent main debaters could ;) | 19:21 |
| v1k0d3n | i think you're defending the current strategy, rather than listening to the concerns of a helm shop. | 19:21 |
| sdake | v1k0d3n can you take up jascott__ 's suggestoin to the ml? | 19:21 |
| sdake | we don't have a strategy, we have tactics... | 19:21 |
| v1k0d3n | sorry...lot of messages being passed. hard to keep up with. | 19:22 |
| kfox1111 | v1k0d3n: possibly. I guess the issue is, you want to always use helm chart installs, and I think helm packages can provide the solution? | 19:22 |
| kfox1111 | both are helm solutions to the smae problem, but maybe my helm based solution is different then yours? | 19:22 |
| sdake | v1k0d3n jascott__ 's suggestion was as follows: [12:18:38] <jascott__>im asking for a list of grievances nailed to the kolla church door that people can reflect on and discuss | 19:22 |
| sdake | v1k0d3n and my followup was as follows: [12:20:21] <sdake>jascott__ sounds like a fine approach to me - on the mailing list | 19:23 |
| kfox1111 | sdake: +1 from me. | 19:23 |
| kfox1111 | mailing list woudl be much easer to keep track of the threads. | 19:23 |
| portdirect | kfox1111: I always want to use helm chart installs | 19:23 |
| v1k0d3n | i would like to see a BP or a spec | 19:23 |
| *** fragatina has quit IRC | 19:23 | |
| *** fragatina has joined #openstack-kolla | 19:24 | |
| portdirect | kfox1111: otherwise i'd use ansible | 19:24 |
| kfox1111 | portdirect: and is a git checkout/helm_prebuild; then do stuf not sufficent to your needs? | 19:24 |
| sdake | v1k0d3n i think the right approach is what jascott__ offered up in a spot of brilliance | 19:24 |
| openstackgerrit | Mark Giles proposed openstack/kolla-kubernetes: Helm creation of Nova services https://review.openstack.org/409473 | 19:24 |
| sbezverk_ | kfox1111: I think the issue here is a target. with packages and microservices we try to make it more usable for the end user, but from v1k0d3n perspective it seems he cares more of contributing and easy re-use.. | 19:24 |
| v1k0d3n | a mailing list? | 19:24 |
| kfox1111 | sbezverk_: and there's a place for that. kolla-kubernetes/helm/services/ | 19:25 |
| v1k0d3n | i'm concerned that's not expedient? my only concern. | 19:25 |
| v1k0d3n | and everyone has work to do. | 19:25 |
| kfox1111 | sbezverk_: just wrap up the microservices in a service package via deps, and it gains the easy to deploy use case. | 19:25 |
| portdirect | kfox1111: for me *possibly* but for someone who has not been invovled in this project as much i really doubt it | 19:25 |
| jascott__ | well we could create a doc a someone link that in ML | 19:25 |
| v1k0d3n | i recommend letting kfox1111 continue his vision? | 19:25 |
| v1k0d3n | as he stated, maybe in time they will be similar. | 19:25 |
| *** rhallisey has joined #openstack-kolla | 19:25 | |
| v1k0d3n | but this brings up a whole other issue about how concerns are addressed. | 19:26 |
| sbezverk_ | so my question is what is our goal, to build something can be used by users or build something which can be endlessly being developed?? | 19:26 |
| v1k0d3n | which we can cover later, elsewhere. | 19:26 |
| kfox1111 | sbezverk_: users. but some users don't want orchestration. such as me. Or need to slowly work up to it. | 19:26 |
| v1k0d3n | sbezverk_: this is a loaded question. isn't that just "development"? | 19:26 |
| portdirect | kfox1111: thats wher we have fundimental disconnect, i want to use k8s to orchestrate openstack for me | 19:27 |
| v1k0d3n | what i'm saying is that helm was first class citizen | 19:27 |
| v1k0d3n | that changed | 19:27 |
| sbezverk_ | v1k0d3n: but what you push is "easy" re-use easy contribute etc.. there is nothing about how end user will use what you propose easily | 19:27 |
| kfox1111 | v1k0d3n: I would love to see the glance package in aic-helm try and target kolal-kubernetes glance microservice packages and see if the ease of launching aic-helm can be built on top of the microservcies. I believe it can easily be done. | 19:27 |
| kfox1111 | portdirect: and I want to support htat use case too. | 19:27 |
| v1k0d3n | how big of a scope is kolla solving?! | 19:28 |
| portdirect | sbezverk_: have you tried to use it? I have :( | 19:28 |
| kfox1111 | portdirect: thje issue is, do we architect it in such a way that its impossible to not orchestrate with helm? | 19:28 |
| portdirect | sbezverk_: (the frown face is beause it was) | 19:28 |
| v1k0d3n | OSAD is laser focused...ubuntu and ansible. kolla is the only project that makes so many exceptions and morphs into something where nothing is first class. | 19:28 |
| v1k0d3n | that in fact in the end...hurts the user. | 19:29 |
| v1k0d3n | hey, sorry i brought all this up guys. | 19:29 |
| kfox1111 | v1k0d3n: your assuming again. :/ | 19:29 |
| v1k0d3n | i've taken my last bow. | 19:29 |
| v1k0d3n | whatever kfox1111 | 19:29 |
| kfox1111 | we have a goal. that covers your exact use case. | 19:29 |
| sbezverk_ | portdirect: what looks easy in the lab, does nor always the easiest way for real life.. | 19:29 |
| v1k0d3n | everyone knows this... | 19:29 |
| kfox1111 | we have developers working towards it. | 19:29 |
| v1k0d3n | :-/ | 19:29 |
| rhallisey | yo | 19:29 |
| portdirect | kfox1111: looking at it a different way (and playing devils advocate) whoul it not make more sense for you to take a helm chat (eg aic/stackanettes type) and then render it using --dry-run and then do what you needed? | 19:30 |
| kfox1111 | we don't have developers focused on exactly your use case ths very moment. | 19:30 |
| *** rstarmer has joined #openstack-kolla | 19:30 | |
| kfox1111 | portdirect: I dont' belive you can individually select templates. | 19:30 |
| openstackgerrit | Qin Wang (qwang) proposed openstack/kolla-kubernetes: WIP - Hel-ming horizon components https://review.openstack.org/408975 | 19:30 |
| kfox1111 | portdirect: most of the work is in the templates themselves anway. the microservice package wrapping is very thin. | 19:30 |
| kfox1111 | in fact, | 19:30 |
| kfox1111 | if we wanted to generate, | 19:30 |
| kfox1111 | we could put the templates into some common dir, | 19:30 |
| portdirect | no - but you could very simpley build a tool to take the templates you wanted | 19:30 |
| kfox1111 | copy them into service packages or microservice packages. | 19:31 |
| kfox1111 | but I felt doing it the "helm native " way using deps was more inline with the community. | 19:31 |
| * srwilkers reads scrollback | 19:31 | |
| portdirect | rhallisey: you dude :) | 19:31 |
| portdirect | (yo!) | 19:31 |
| rhallisey | portdirect, :) | 19:31 |
| jascott__ | of all these projects, who has demonstrated day 2 helm based deployment and upgrades in a gate/CI? seems everyone has very strong opinions with no proof | 19:32 |
| jascott__ | (of which was is best) | 19:32 |
| kfox1111 | jascott__: that is one of my big conserns. | 19:32 |
| rhallisey | jascott1, it's a good point. It's a concern I have with helm orchestration. Everything is too low level and hidden | 19:33 |
| sbezverk_ | jascott__: oh yeah, the gate :-) | 19:33 |
| kfox1111 | I'm trying to be pragmatic and say, I don't think anyone knows yet how to build it. not even us. so lets give our operators flexability until its discovered. | 19:33 |
| jascott__ | sbezverk_: lol ;) | 19:33 |
| kfox1111 | sorry. to clearify that statement, I mean "human operators" | 19:34 |
| kfox1111 | our users. | 19:34 |
| rhallisey | kfox1111, I call them vessels now :) | 19:34 |
| rhallisey | will make life easier :) | 19:34 |
| * portdirect facepalms | 19:34 | |
| rhallisey | not humans. The software operators | 19:34 |
| kfox1111 | rhallisey: there's been a lot of miscommunication on this thread. just trying to be very very clear to avoid misunderstanign. | 19:35 |
| rhallisey | ya I'm kinda jumping into the middle of it. Carry on :0 | 19:35 |
| rhallisey | :) | 19:35 |
| srwilkers | oh hey rhallisey | 19:35 |
| srwilkers | fancy seeing you here | 19:35 |
| rhallisey | hey oh! | 19:35 |
| kfox1111 | layer 2 should allow operators to self orchestrate until a rock solid orchestration layer emerges. | 19:36 |
| sbezverk_ | take lego approach, when you buy for your kid lego, there is no magic button to put everything together in one shot.. doing piece by piece triggers learning. I think the same approach is needed to kolla-kubernetes since we do not have yet proven way of operate it | 19:36 |
| portdirect | kfox1111: is orchestratin not what k8s is about? | 19:36 |
| kfox1111 | once it emerges, may it makes sense to drop layer 2 and merge it with the orchestration stuff. | 19:36 |
| kfox1111 | portdirect: no. not at all. | 19:36 |
| rhallisey | sbezverk_, well legos aren't cloud native. That's why everyone plays minecraft now | 19:37 |
| portdirect | kfox1111: ? | 19:37 |
| rhallisey | :) | 19:37 |
| jascott__ | haha | 19:37 |
| srwilkers | rhallisey, sweeping in for the kill | 19:37 |
| sbezverk_ | rhallisey: I am old for new fancy staff ;-) | 19:37 |
| rhallisey | I like legos too :) | 19:37 |
| *** jmccarthy has quit IRC | 19:38 | |
| kfox1111 | portdirect: kubernetes is about providing a cloud api that allows resources to be requested and containers to be lauched and kept alive. without caring so much about scheduling and keeping alove. | 19:38 |
| kfox1111 | orchestration has been kept out of scope beyond that in kubernetes. | 19:38 |
| portdirect | kfox1111: http://kubernetes.io/ "Production-Grade Container Orchestration" ? | 19:38 |
| *** jmccarthy has joined #openstack-kolla | 19:38 | |
| * jascott__ quits to start cloud native lego startup | 19:38 | |
| kfox1111 | so, "you need this service to talk to this backend, used by these 3 other things" | 19:38 |
| kfox1111 | thats orchestration. | 19:38 |
| * portdirect applies to jascott__'s new multibillion dollar unicon | 19:39 | |
| kfox1111 | portdirect: heh. ok. lets say, there are many different ways to define any word. :/ | 19:39 |
| jascott__ | hired. foosball ensues. | 19:39 |
| srwilkers | foosball jascott__ ? sign me up please | 19:39 |
| kfox1111 | let me try and define things a little more clearly then.... | 19:39 |
| openstackgerrit | Qin Wang (qwang) proposed openstack/kolla-kubernetes: WIP - Hel-ming horizon components https://review.openstack.org/408975 | 19:39 |
| portdirect | sweet :) | 19:39 |
| kfox1111 | "container orchestration" starting containers in sets (pods) and sets of sets. | 19:39 |
| kfox1111 | "deployment orchestration" lauching all the k8s objects in a way that produces a usable end result. | 19:40 |
| kfox1111 | k8s is focused on just the first. letting you request resources and fulfil them. | 19:40 |
| kfox1111 | the later, is explicitly out of scope of kubernetes. | 19:41 |
| kfox1111 | helm is focused on the second. | 19:41 |
| kfox1111 | is that a little more clear? | 19:41 |
| kfox1111 | k8s didn't want to aporach the second, as that can be deeply religious or different tools must be employeed for different reasons. | 19:41 |
| kfox1111 | helm is one implementation, but kpm is another. | 19:42 |
| kfox1111 | its not k8s's stated goal to decide which, or any other tool that talks to k8s. | 19:42 |
| portdirect | I'm not sure kfox1111, my view is that the k8s manifest's, service dicovery etc, are examples of "deployment orchestration", i view helm and kpm as just ways of preparing them for use | 19:42 |
| kfox1111 | portdirect: they are building blocks, that can be used for deployment orchestration. | 19:42 |
| kfox1111 | but they in of themselves don't provide the orchestration. | 19:43 |
| kfox1111 | how does k8s know, you need a service, a deamonset and a deployment to make a usable thing? | 19:43 |
| *** rstarmer has quit IRC | 19:44 | |
| kfox1111 | they have stated that is outside of scope of "container orchestration" | 19:44 |
| *** rstarmer has joined #openstack-kolla | 19:44 | |
| sbezverk_ | or roder containers need to be started?? k8s does not care, it just starts them.. | 19:44 |
| kfox1111 | either a human loads the objects in, or a nother tool outside of k8s does. | 19:44 |
| *** rstarmer has quit IRC | 19:44 | |
| kfox1111 | sbezverk_: right. | 19:44 |
| *** rstarmer has joined #openstack-kolla | 19:45 | |
| kfox1111 | could the deployment need an external maridb or an internal ceph? k8s doesn't care. it only does what you tell it to do. laucn hand keep track of some pods. | 19:45 |
| *** rstarmer has quit IRC | 19:45 | |
| *** rstarmer has joined #openstack-kolla | 19:46 | |
| kfox1111 | esnuring all the objects that are important to the application are loaded in properly into k8s is outside of k8s's stated scope. | 19:46 |
| *** rstarmer has quit IRC | 19:46 | |
| openstackgerrit | Qin Wang (qwang) proposed openstack/kolla-kubernetes: WIP - Hel-ming horizon components https://review.openstack.org/408975 | 19:46 |
| portdirect | I'm lost to be honest now | 19:46 |
| *** rstarmer has joined #openstack-kolla | 19:46 | |
| portdirect | i'm not sure how what you are discribing is different from ansible say? | 19:47 |
| *** rstarmer has quit IRC | 19:47 | |
| kfox1111 | portdirect: ok. maybe this would help. if k8s was to suport deployment orchestration, | 19:47 |
| kfox1111 | there would be a document type that described "i need you to pull these 3 k8s objects from this site and ensure they all exist) | 19:47 |
| *** rstarmer has joined #openstack-kolla | 19:47 | |
| kfox1111 | the the user would kubectl create -f app.yaml | 19:47 |
| openstackgerrit | Qin Wang (qwang) proposed openstack/kolla-kubernetes: WIP - Hel-ming horizon components https://review.openstack.org/408975 | 19:47 |
| *** rstarmer has quit IRC | 19:47 | |
| kfox1111 | and k8s would do the deployment orchestration. | 19:48 |
| *** rstarmer has joined #openstack-kolla | 19:48 | |
| kfox1111 | k8s has stated in docs and in revies they don't want to support that though. | 19:48 |
| kfox1111 | so, think if it this way. | 19:48 |
| *** rstarmer has quit IRC | 19:48 | |
| kfox1111 | k8s is like nova. | 19:48 |
| kfox1111 | helm is like heat. | 19:48 |
| portdirect | https://github.com/portdirect/harbor/blob/latest/kubernetes/templates/keystone/controllers.yaml ? | 19:48 |
| portdirect | is that not waht you were desibing? | 19:49 |
| kfox1111 | k8s has an api, that takes in a request for a resource (daemsonet) and makes sure it runs. | 19:49 |
| *** rstarmer has joined #openstack-kolla | 19:49 | |
| *** rstarmer has quit IRC | 19:49 | |
| kfox1111 | helm takes in some paramaters, and a template, renders it into a useful set of templates, and then asks k8s to create a set of objects needed to run the stack. | 19:49 |
| kfox1111 | portdirect: the templating is really a key part of it. | 19:50 |
| kfox1111 | k8s takes in a very spicific instantiation of something. | 19:51 |
| portdirect | yeah - this is where I'm confused though, and to be honest i think you are presenting a moving target | 19:51 |
| kfox1111 | templating makes it generic enough to ship, and then orchestrates the actual deployment (rendering it into something that it can hand to k8s) | 19:51 |
| awiddersheim | I noticed when installing kolla with ansible I no longer get the python libraries for like genpwd | 19:52 |
| *** msimonin has joined #openstack-kolla | 19:52 | |
| kfox1111 | portdirect: I've been pretty consistent with this I think. we do need to discuss it more though so everyone's on the same page. | 19:52 |
| awiddersheim | using git directly | 19:52 |
| kfox1111 | portdirect: I totally acknolege my understanding of it will evolve as I learn to use it though, just like everyone else though. | 19:52 |
| kfox1111 | portdirect: we'rea all still learning. | 19:53 |
| portdirect | yeah, im just trying to get on the same page :) | 19:53 |
| kfox1111 | helm has two pieces of funcitonality in it though. | 19:54 |
| sdake | i'll bbi45 mins or so | 19:54 |
| kfox1111 | a templating engine, and an orchestration (workflow?) engine. | 19:54 |
| portdirect | so that tempate i shoued you above, is very rigit, but would allow you to do exacktly the kubectl create -f app.yaml, you gave as an example to get a running yestone (ok a bit of sed involved...) | 19:54 |
| portdirect | rigit - rigid | 19:54 |
| kfox1111 | portdirect: yeah. very minimally sed then is your templating engine, and its outside of k8s. | 19:54 |
| kfox1111 | k8s doesn't want to be involved in that piece. they want others to define the templateing engine thats best for them too. | 19:55 |
| portdirect | so we are inagreement that k8s does dempoyment orchestaration? but not template management? | 19:55 |
| portdirect | dempoyment - deployment | 19:55 |
| kfox1111 | so jinja2 for those that like that, or gotl for those. k8s is unopionated about it. | 19:55 |
| kfox1111 | hmm.... | 19:55 |
| kfox1111 | I guess, the way I define deployment orchestration, I'd say it depends on a templating engine. | 19:56 |
| kfox1111 | so no. | 19:56 |
| kfox1111 | but if you exclude that requirement, I could be convinced to say yes. | 19:56 |
| kfox1111 | in some cases. | 19:56 |
| *** gomarivera has joined #openstack-kolla | 19:56 | |
| kfox1111 | that, being there are no deps between objecgts, or the objects have orchestration built in. | 19:56 |
| kfox1111 | though, I'd argue k8s isn't orchestrating then, but the pods themselves. | 19:57 |
| portdirect | ok lets look at the last point | 19:57 |
| kfox1111 | k | 19:57 |
| portdirect | so a pod will only launch when the volume, secrets, config are avaible for it | 19:57 |
| portdirect | it will then use a livlyness propb to signal that its reeady | 19:58 |
| portdirect | and enpoints will be attached to a service (dns) so it can be found | 19:58 |
| kfox1111 | that is "container orchestration". | 19:58 |
| kfox1111 | yeah. | 19:58 |
| portdirect | and then another pod can find it | 19:58 |
| kfox1111 | right. | 19:58 |
| kfox1111 | but how many non trivial apps can be deployed with a static set of objects? | 19:59 |
| kfox1111 | I think the number is very small. | 19:59 |
| kfox1111 | thats why things like helm exist. | 19:59 |
| kfox1111 | or why everyone tries to do some kind of templating. | 20:00 |
| portdirect | lets not go there for a moment | 20:00 |
| kfox1111 | k. | 20:00 |
| portdirect | please desribe a not trilial app that cannot be dolpyed using the methodology i just described | 20:00 |
| kfox1111 | simple 2 tier web app. web server with backend db. often you need some extra pieces of info: | 20:01 |
| kfox1111 | * root pw for db. | 20:01 |
| kfox1111 | * ssl certificates for the https endpoint. | 20:01 |
| kfox1111 | * some mechanism to load the schema into the db before it becomes usable. | 20:02 |
| kfox1111 | so just lauching the k8s objects concatinated into a single object and loaded into k8s don't provide a usable system all by itself. | 20:02 |
| portdirect | ok you are moving back to templating for the password, the ssl cert can either be generated by a job, and the schmea can also be performed via a job | 20:02 |
| portdirect | all from a sinle template | 20:02 |
| kfox1111 | not and be trusted, but ok. | 20:02 |
| kfox1111 | assuming you have already a set of objects that are presetup the way you would expect, | 20:03 |
| portdirect | how not - a job generates cert and puts it in secret - the pod that depends on it will not launch untill it is present | 20:03 |
| kfox1111 | then k8s could potentially deploy it, assuming everything could be mapped into k8s objects. | 20:03 |
| kfox1111 | portdirect: most ssl cert authorizties don't nessisarily have automated processes to allow signing. | 20:04 |
| portdirect | the db bigration could be performed wither by a job that signials its done (eg kube-entrypoint) or run as in init container for the pod of the service it consumes | 20:04 |
| kfox1111 | portdirect: yes, but thats not k8s doing the orchestration. | 20:04 |
| sbezverk_ | portdirect: I do not think this appraoch will work for everything, when we tried to have all components of opensatck service bootstrap in the same job but using multiple containers it was a nightmare | 20:04 |
| kfox1111 | thats the job itself doing it. | 20:04 |
| kfox1111 | the sofware being loaded into k8s is doing the orcestration then. | 20:04 |
| portdirect | *into* k8s | 20:04 |
| kfox1111 | somethign has to keep track of non trivial orcehstration | 20:05 |
| sbezverk_ | portdirect: k8s cannot do this orchestration for us | 20:05 |
| portdirect | ok - so not a three teir app | 20:05 |
| kfox1111 | "Has the db been inited yet. now, run a job to init it." | 20:05 |
| portdirect | ok - thats easy | 20:05 |
| kfox1111 | the problem is, that logic is very application specific. | 20:05 |
| portdirect | that maifest i posed above does exactly that | 20:05 |
| kfox1111 | it can be wrapped in a k8s job for sure. | 20:06 |
| kfox1111 | but its not k8s thats orchestrating the deployement. its the sofware itself running in k8s thats orchestrating the deployment. | 20:06 |
| portdirect | there is a db pod, a manager that gets the certs etc, and then a bootstrapper that runs after keystone ahs come up to populate users and endpoints | 20:06 |
| kfox1111 | yup. | 20:07 |
| portdirect | so this is why i think we should be doing everything in k8s, not trying to perform orcehstration from outside | 20:07 |
| kfox1111 | thats again, not k8s doing the deployment orchestration, but it the "container orchestration" that k8s does, can end up loading sofware that does "deployment orchestration" | 20:07 |
| kfox1111 | portdirect: thats what kubernetes-endpoint was written for. | 20:08 |
| kfox1111 | to allow for easier "internal deployment orchestration" | 20:08 |
| kfox1111 | some folks swear by it. I'm all for supporting the use case. | 20:08 |
| portdirect | nice - lets use it - it's on the spec | 20:08 |
| kfox1111 | I personally don't believe in it yet though. as it can be disasterious if designed wrong. and its hard to get right. | 20:08 |
| kfox1111 | so we made it optional in the spec. | 20:09 |
| kfox1111 | with the design in the spec, | 20:09 |
| kfox1111 | deployment orchestration can be manual, external, or internal. | 20:09 |
| kfox1111 | whoever wants to work on any of those parts has room to work on it. :) | 20:09 |
| kfox1111 | microservices are key in the first part. manual. | 20:10 |
| openstackgerrit | Andrew Widdersheim proposed openstack/kolla-ansible: Fix kolla-genpwd and kolla-mergepwd https://review.openstack.org/409909 | 20:10 |
| kfox1111 | helm based deployment orchestration is a combination of both external and internal orchestration. | 20:10 |
| portdirect | ok - im stating now for the record I'm only inertested in internal | 20:10 |
| kfox1111 | portdirect: great. :) | 20:10 |
| kfox1111 | I support you in that. :) | 20:11 |
| portdirect | so lets work out if its possible to make my goals and your' both 1st class citizens | 20:11 |
| kfox1111 | for the record, I only want manual for now. with a goal longer term, once internal or external has proven itself, I may swithc to that. | 20:11 |
| kfox1111 | +1 | 20:12 |
| kfox1111 | I belive it is. | 20:12 |
| portdirect | I hope so :) | 20:12 |
| portdirect | so lets look at the requirements on the templating side | 20:12 |
| kfox1111 | so, lets talk through the workflow a bit. | 20:12 |
| kfox1111 | lets leave templates out for a second. | 20:13 |
| kfox1111 | starting at k8s. | 20:13 |
| portdirect | can we dend the problems first before gettin into the detail? | 20:13 |
| kfox1111 | k. | 20:13 |
| portdirect | define | 20:13 |
| kfox1111 | go ahead. | 20:13 |
| portdirect | sry my cat is attacking me :) | 20:13 |
| kfox1111 | hehe. I toootaly get that. :) | 20:13 |
| kfox1111 | your not worshiping hard enough ! :) | 20:14 |
| sbezverk_ | portdirec: what kind of cat lion, kind ;-)?? | 20:14 |
| portdirect | grumpy - and overly affectionate - depends on the day | 20:14 |
| portdirect | :) | 20:14 |
| kfox1111 | hehe. | 20:15 |
| jascott__ | dogs rule! http://i.imgur.com/b4ZbnjB.gifv | 20:15 |
| portdirect | so getting on to templates - I want what we produce to be consumable by a user familar with k8s immediatly | 20:15 |
| * portdirect quits jascott__'s startup in discust | 20:15 | |
| kfox1111 | portdirect: at what level? | 20:15 |
| kfox1111 | portdirect: one to one mapping with k8s objects, which is more normal to direct k8s, | 20:16 |
| jascott__ | lol. you're stock is worthless! you spent too much time playing foos! | 20:16 |
| kfox1111 | or at an "application" level? | 20:16 |
| kfox1111 | they are two different use cases I think. | 20:16 |
| kfox1111 | microservices are targeted to the first use case. | 20:16 |
| portdirect | at an application level - again refering to the template i shoed you - auser could take that and fill it in my hand in 10 mins if they knew k8s | 20:17 |
| kfox1111 | lauch helm microservice package template, get a k8s object out the other side. | 20:17 |
| portdirect | and I want that same experience for a helm user | 20:17 |
| kfox1111 | k. the other route is service package. | 20:17 |
| kfox1111 | a collection of templates rendered to prodcue a working app. | 20:17 |
| kfox1111 | or in this case, openstack service | 20:17 |
| portdirect | as otehrwise its like saying we support redhat, but here is our own version of rpm. | 20:17 |
| kfox1111 | nova say. | 20:17 |
| *** msimonin has quit IRC | 20:18 | |
| *** msimonin has joined #openstack-kolla | 20:18 | |
| kfox1111 | the nova service package contains deps to all the microservice packages. | 20:18 |
| kfox1111 | its functionally equiv to if the individual templates were bundled all in the main package. | 20:18 |
| kfox1111 | its just an implementation detail that they are subpackages instead of in templates directly. | 20:18 |
| kfox1111 | the user helm install nova --variables nova_overrides | 20:19 |
| kfox1111 | and it renders the templates basically into one big blob of k8s objects lodaed into k8s. | 20:19 |
| portdirect | but its really hard to consume, thats why i suggested (a long way back in the scrollback) for your use case would it not may more sense instread of building a pebuild.py to have a helm_decompose.py ? | 20:19 |
| portdirect | that way we could meet both use cases? | 20:19 |
| kfox1111 | portdirect: I guess the question here related to that is, | 20:20 |
| kfox1111 | do we believe the primary use case of someone git checking out the repo, to be to launch the service packages with helm? | 20:20 |
| kfox1111 | sorry. let me refrase that slightly. | 20:21 |
| kfox1111 | do we believe the primary use case of someone git checking out the repo, to be to launch the service charts with helm? | 20:21 |
| portdirect | i would hope so - I want this project to be sucessfull and that means ultimatly more users than developers | 20:21 |
| kfox1111 | because I think that use case is better handled differently. | 20:21 |
| portdirect | and we will only be sucessfull if we intergrate with what the wider community is doing | 20:22 |
| kfox1111 | I think the easiest way to support helm deployment for users, is not that the users should have to gknow how to use git, | 20:22 |
| kfox1111 | but just know how to use helm. | 20:22 |
| kfox1111 | so, rather then git checkout kolla-kubernetes | 20:22 |
| jascott__ | hey check this out Chaos monkey for k8s, might be useful https://github.com/asobti/kube-monkey | 20:22 |
| kfox1111 | cd kolla-kubernetes/helm/services/<service-x>; helm install .... | 20:22 |
| kfox1111 | they use pure helm: | 20:22 |
| kfox1111 | helm repo add kolla http://helm.openstack.org/ | 20:23 |
| kfox1111 | helm install kolla/<service-x> .... | 20:23 |
| kfox1111 | the same way kolla provides containers, they can provide packages too. | 20:23 |
| kfox1111 | then its even more native helm. | 20:23 |
| portdirect | no that fundimentally doesnt work i think | 20:24 |
| kfox1111 | ok. why not? | 20:24 |
| kfox1111 | curious. | 20:24 |
| portdirect | as it mean we are too optionated in the packages we provide | 20:24 |
| kfox1111 | why? the vars are all overridable? | 20:24 |
| portdirect | how could a user modify a package to their orgs needs easily youing that workflow | 20:25 |
| kfox1111 | in what way would modifications be needed? and how common is that? | 20:25 |
| portdirect | very | 20:25 |
| sbezverk_ | kfox1111: can we do what you describe, but without built packages? | 20:25 |
| portdirect | eg - add federation support to keystone | 20:25 |
| portdirect | or a different sdn layer for neutron | 20:26 |
| kfox1111 | sbezverk_: without build packages, it places certain requirements on the repo. | 20:26 |
| sbezverk_ | kfox1111: I mean with source templates | 20:26 |
| kfox1111 | portdirect: ah. | 20:26 |
| portdirect | or a manufacturors drivers for cinder | 20:26 |
| kfox1111 | portdirect: those ideally are either one of two things, feature enhancements more then config changes, | 20:27 |
| kfox1111 | or selecting different microservice packages? | 20:27 |
| portdirect | or use images that have gone through their own ci system? | 20:27 |
| *** gomarivera has quit IRC | 20:27 | |
| kfox1111 | use of images not through kollas ci can be done via var overrde, or | 20:27 |
| portdirect | that doesnt work - no one is gonoin to get that inversted, they will just use somthing else | 20:27 |
| kfox1111 | building wht the helm prebuild --version flag in the poc? | 20:28 |
| kfox1111 | they won't use it either if we can't maintain it. :/ | 20:28 |
| kfox1111 | gotta find the right balance. :/ | 20:28 |
| sdake | kfox1111 is right on the money there | 20:28 |
| kfox1111 | what about a release specifically for folks wanting to tweak the charts? | 20:29 |
| sdake | maintaing it comes from building a big big community | 20:29 |
| *** Pavo has quit IRC | 20:29 | |
| portdirect | yeah - but they wont try if they cant get a poc running quickly | 20:29 |
| kfox1111 | getting th epoc running as quickly as possible I think is the prebuilt package route. | 20:29 |
| kfox1111 | then git and stuff isn't an issue. | 20:29 |
| *** Pavo has joined #openstack-kolla | 20:29 | |
| kfox1111 | just helm and go. | 20:29 |
| portdirect | it'll be quicker to build it themselves, or juts use another openstack implementation, or even just not use openstack at all? | 20:30 |
| kfox1111 | getting dev's, dev-env up is a bit different. | 20:30 |
| kfox1111 | portdirect: if we implement service packages, and even the overarching package, | 20:30 |
| kfox1111 | its as easy as helm add repo kolla http:.... | 20:30 |
| kfox1111 | helm install kolla/openstack | 20:30 |
| kfox1111 | I'm not sure how to make it much easier then that. | 20:30 |
| portdirect | I dont want helm install kolla/openstack | 20:32 |
| kfox1111 | portdirect: k. what do you want? | 20:32 |
| portdirect | i want helm install kolla/keystone | 20:32 |
| v1k0d3n | ^^^ portdirect yes!!!!!!!!! | 20:32 |
| kfox1111 | that works too under this model. | 20:32 |
| v1k0d3n | "microservices" :) | 20:32 |
| portdirect | then helm install kolla/cinder | 20:32 |
| sdake | yup thats exactly what i want too portdirect | 20:32 |
| kfox1111 | v1k0d3n: we differ on our definitions of microservice. thats what I keep telling you! | 20:32 |
| sdake | nova is not a microservice, its a service | 20:33 |
| portdirect | i do not want helm install kolla/cinder-api | 20:33 |
| kfox1111 | v1k0d3n: portdirect thats the model I've been laying out. | 20:33 |
| sdake | nova-novncproxy is a microservice | 20:33 |
| portdirect | helm install kolla/cinder-svc | 20:33 |
| kfox1111 | cinder is a service package, cinder-api is a microservice package. | 20:33 |
| portdirect | helm install kolla/cinder-db-job | 20:33 |
| sdake | portdirect i also want helm install kolla/cinder-api | 20:33 |
| portdirect | helm install kolla/cinder-db-migration | 20:33 |
| portdirect | etc | 20:33 |
| v1k0d3n | https://github.com/att-comdev/aic-helm/tree/master/keystone << this is close to what we're after for a single service. | 20:34 |
| v1k0d3n | just saying. | 20:34 |
| *** lbeliveau has quit IRC | 20:34 | |
| portdirect | sdake, and we should work to support that | 20:34 |
| v1k0d3n | or something that can be consumed as a whole of something else. | 20:34 |
| sdake | portdirect 100% agree | 20:34 |
| jascott__ | ya so the keystone package lives in helm/services so you get what you want | 20:34 |
| kfox1111 | portdirect: 100% agree too. | 20:34 |
| sdake | portdirect that i the direction ew are headed | 20:34 |
| kfox1111 | v1k0d3n: +1. it shoudl be consumable as a piece of a greater thing too. | 20:35 |
| sdake | jascott__ that is one apporach - i think we can hash that out in the reviews once we do a blueprint for services | 20:35 |
| sdake | i tihnk where everyone is stuck atm is "how do we do services with the current microservice model" | 20:35 |
| kfox1111 | v1k0d3n: can you do something for me please? | 20:35 |
| kfox1111 | v1k0d3n: have a review of this: | 20:35 |
| kfox1111 | https://review.openstack.org/#/c/407771/ | 20:36 |
| kfox1111 | I think a lot of our misunderstanding is covered in that. | 20:36 |
| sbezverk_ | kfox1111: what prevents us from building helm/services structure and then each service would reference to corresponding microservice? I do not think anything prevents us doig it | 20:36 |
| portdirect | sdake: i think this where helm-decompose.py (or wahtever comes in) - as if you want helm install cinder-api you are most likley inversted enouth to use such a tool | 20:36 |
| sdake | portdirect any chance you can make a review that looks like that? | 20:37 |
| portdirect | i can try | 20:37 |
| kfox1111 | sbezverk_: thats what I've been trying to say for a long time now. :) | 20:37 |
| sdake | portdirect so folks can visualize it | 20:37 |
| sdake | sbezverk_ bingo :) | 20:37 |
| sdake | sbezverk_ maybe you and portdirect can work on that together since this is such a point of contention | 20:38 |
| sbezverk_ | kfox1111: I think we should just build a service chart with misroservices attached so this endless discussion would be closed.. | 20:38 |
| kfox1111 | sbezverk_: +1. | 20:38 |
| kfox1111 | sbezverk_: and ideally from the aic-helm glance example. | 20:38 |
| sbezverk_ | sdake: would be more than happy :-) | 20:38 |
| kfox1111 | showing how to close the loop. | 20:38 |
| v1k0d3n | kfox1111: personally? | 20:39 |
| kfox1111 | v1k0d3n: ? | 20:40 |
| v1k0d3n | i feel like it's reinventing a lot of information already out there, but treating this as a source of truth. | 20:40 |
| kfox1111 | v1k0d3n: the terms doc? | 20:40 |
| v1k0d3n | absolutely | 20:40 |
| v1k0d3n | for instance... | 20:40 |
| v1k0d3n | microservice... | 20:41 |
| kfox1111 | feel free to review and post alternative terms. | 20:41 |
| jascott__ | sbezverk_ +1 | 20:41 |
| v1k0d3n | do you go down so far as a single process, or is a microservice somethig that can be reused in it's own right along with other unlike services? | 20:41 |
| kfox1111 | we've been using "microservice" to in the way used in terms. if there is a better way, we could switch. | 20:41 |
| kfox1111 | v1k0d3n: right. we just picked microservice to = k8s object in our terminology to have something to discuss. | 20:42 |
| sbezverk_ | kfox1111: hopefully not swithcing, otherwise everybody will be out of sync ;-) | 20:42 |
| kfox1111 | it could be called "foo" and still we could then talk about it. | 20:42 |
| *** jascott__ has quit IRC | 20:42 | |
| kfox1111 | we just need a set of terms to prevent so much arguing without understanding that we're talking about different things. | 20:43 |
| *** jascott1_ has joined #openstack-kolla | 20:43 | |
| *** lbeliveau has joined #openstack-kolla | 20:43 | |
| kfox1111 | sbezverk_: if switching can prevent 10 more hours not being consumed on discussing terminology issues, I'd be happy to swithc. | 20:43 |
| v1k0d3n | kfox1111: this is an interesting conversation. | 20:43 |
| kfox1111 | human lanugage is about relaying ideas back and forth. | 20:44 |
| v1k0d3n | got stranger for me. | 20:44 |
| kfox1111 | if we can't agree to common terminology for the discussion, we're not speaking the same langauge, | 20:44 |
| kfox1111 | and cant' talk. :/ | 20:44 |
| v1k0d3n | ok. | 20:44 |
| kfox1111 | does that make sence? | 20:45 |
| v1k0d3n | does what make sense? | 20:45 |
| kfox1111 | how important that we're on the same page term wise for discussion? | 20:45 |
| *** jtriley has quit IRC | 20:45 | |
| v1k0d3n | kfox1111: yes. you have made your points. | 20:46 |
| kfox1111 | k. | 20:46 |
| kfox1111 | I think a lot of the issues have been around the term microservices being used differently by different folks. | 20:46 |
| kfox1111 | and then not understanding what we are all talking about. | 20:46 |
| kfox1111 | its not that we're not listening to each other, but mishearing. | 20:46 |
| v1k0d3n | the difference is in the implementation model from other helm usage and creation to be frank. | 20:47 |
| v1k0d3n | i know what a microservice is. | 20:47 |
| v1k0d3n | and i don't freaking care. | 20:47 |
| v1k0d3n | what i care about is the architecture. | 20:47 |
| v1k0d3n | is it usable compared to other implementations of helm. | 20:47 |
| portdirect | yup | 20:47 |
| srwilkers | sdake, Jeffrey4l: not sure what the process entails, but is it possible to consider getting kolla-k8s its own weekly meeting? I feel one of the short comings right now is that we have to piggyback off kolla/kolla-ansible, and i think having an hour time set up each week would really help us get this discussion along each week | 20:47 |
| v1k0d3n | no it's not from what i've seen. | 20:47 |
| v1k0d3n | move on | 20:47 |
| kfox1111 | helm provides the ability to use subcharts. are you saying that, since few use them yet, we hsould not use them? | 20:48 |
| srwilkers | i just feel that as this goes on, we're going to have a lot of latency induced by never really having time set aside to discuss this project specifically | 20:48 |
| v1k0d3n | i don't need a lesson in microservices though. if THAT's the focus... | 20:48 |
| sdake | srwilkers that is being discussed on the mailing list as we speak | 20:48 |
| srwilkers | where everyones involved and knows what time it happens | 20:48 |
| srwilkers | sdake, good | 20:48 |
| v1k0d3n | then we have bigger gaps than i originally thought | 20:48 |
| portdirect | srwilkers: great suggestion | 20:48 |
| sdake | srwilkers you can participate if you like - it would be helpful :) | 20:48 |
| sdake | in the mailing list discussion i mean srwilkers | 20:49 |
| srwilkers | sdake, yep | 20:49 |
| * portdirect needs to check his email more often | 20:49 | |
| kfox1111 | so the discussion is in, do we make a single service chart, or allow it to be more modularly assembled? | 20:49 |
| *** shardy_afk has quit IRC | 20:49 | |
| *** g3ek has quit IRC | 20:50 | |
| v1k0d3n | kfox1111: this is how i really feel about microservices. | 20:50 |
| v1k0d3n | https://www.youtube.com/watch?v=Nosa5-xcATw&t=4m11s | 20:50 |
| kfox1111 | I need what is reffered to as microservice packages in the terminology doc. if kolla decides to not support those at all, its fine, but I don't think I can contribute. | 20:51 |
| kfox1111 | v1k0d3n: I thihnk we're pretty well in violent disagreement on path. | 20:51 |
| kfox1111 | I can only work on the things I need to work on, as by openstack rules. | 20:52 |
| kfox1111 | I'm spending an awefully lot of time outside of that right now, and I can only do that for so long. :/ | 20:52 |
| kfox1111 | so do we spend it endlessly rehashing the path, or actually working on getting to the end goal where we all can be happy? | 20:53 |
| v1k0d3n | all good. that's why i don't want to disrupt that anymore. | 20:53 |
| v1k0d3n | i wouldn't have used the word violent though. | 20:53 |
| kfox1111 | k. I truely think we can get you waht you want. and if you can put devs on the task of service packages, we can get it sooner. | 20:53 |
| v1k0d3n | i would just say that our focus is different. | 20:53 |
| kfox1111 | yeah. fair enough. | 20:54 |
| *** g3ek has joined #openstack-kolla | 20:54 | |
| v1k0d3n | my focus, and what i brought up...was using a helm native approach as primary. | 20:54 |
| kfox1111 | v1k0d3n: we disagree on that term. | 20:54 |
| kfox1111 | "helm native" | 20:54 |
| * portdirect is sad | 20:54 | |
| kfox1111 | I believe I'm following it too. but just in a different way then you. | 20:54 |
| v1k0d3n | well, helm was called out in the spec | 20:54 |
| v1k0d3n | helm 2.0 | 20:55 |
| kfox1111 | helm 2.0. yes. | 20:55 |
| sbezverk_ | v1k0d3n: if we can show that microservices we built do not prevent, to deploy at service level being glance would it work for you? | 20:55 |
| kfox1111 | sbezverk_: I think the concern is more, the repo isn't layed out like aic-helm so its not "helm native" | 20:56 |
| sbezverk_ | or the only option you accept is getting your repo as is and start using ti? | 20:56 |
| v1k0d3n | i'd like to have a spec for how helm is being delivered in this project. | 20:56 |
| v1k0d3n | that;s what it would take | 20:56 |
| kfox1111 | nothing short of dropping microservice deployment I think will resolve that. :/ | 20:56 |
| portdirect | sbezverk_: the issue is not at the stage of running helm install but everythin leading to that step. | 20:56 |
| kfox1111 | portdirect: right. :/ | 20:57 |
| kfox1111 | and I'm concerned thats not solvable without throwing features under the bus. | 20:57 |
| sbezverk_ | portdirect: so no any type of prepping is accepted? | 20:58 |
| portdirect | I'm concerned that if we build our own coach, everyone else will be on a nice buss driving off into the sunset | 20:58 |
| sbezverk_ | other then what helm offers | 20:58 |
| kfox1111 | portdirect: I believe the infra is short term. others will run into the same issues we are, and helm will grow to cover the needed use cases. | 20:58 |
| portdirect | sbezverk_: no - thats why i sugessted the parsing of charts.yaml rather than tempating it | 20:58 |
| kfox1111 | portdirect: like, a "helm install --version-override xx.xx.xx.xx to overrdie the version number int he chart. | 20:59 |
| kfox1111 | rather then rewriting the file. | 20:59 |
| sbezverk_ | portdirect: sorry I must have missed it, parsing it with what? | 20:59 |
| portdirect | helm prebuild - kfox1111 has a ps in to parse a standard helm chart yaml and then update the version number as required | 21:00 |
| portdirect | rather than templating it | 21:00 |
| kfox1111 | portdirect: yeah. but I don't think the same thing can be done for values at the moment. | 21:00 |
| *** haplo37 has quit IRC | 21:00 | |
| kfox1111 | the only way to really do that is to flatten them, and take the maintence burdon. | 21:01 |
| SamYaple | portdirect: dont be sad! get glad! | 21:01 |
| * portdirect is a drama queen | 21:01 | |
| kfox1111 | at this stage in the game, I don't think we want to take on the maintenence berson of values until at least all the existing services are converted. | 21:02 |
| *** haplo37 has joined #openstack-kolla | 21:02 | |
| portdirect | but - its hardly hard - we just build out values with defaults, check out what keystone in the aic-helm stuff - is that an unnacpetable maintaince burdon? | 21:03 |
| kfox1111 | otherwise, each change to a common var wil have to update all values files individually taking more work, risking foretting one, and creating more merge conflicts slowing down development. | 21:03 |
| *** matrohon has joined #openstack-kolla | 21:04 | |
| kfox1111 | maybe common vars is solid enough now. I don't know. | 21:04 |
| kfox1111 | or maybe we wait 2 weeks and then flatten? | 21:04 |
| kfox1111 | or 1 week? nova/glance/cinder should be in by then most likely. | 21:05 |
| kfox1111 | I really would rather helm graow a feature to allow a subchart's vars to be included drectly in the main charts vars. | 21:06 |
| portdirect | so thats why we have python and the ability to parsw yaml files. | 21:06 |
| portdirect | i did it with sed and bash ok | 21:06 |
| kfox1111 | hmm... | 21:07 |
| kfox1111 | ok... what if we did this... | 21:07 |
| *** goldyfruit has quit IRC | 21:07 | |
| kfox1111 | we split up kolla-common into a few sub packages, that match up with the var sections. | 21:08 |
| kfox1111 | we then put the vars stuff specific to each set in the vars of that subpackage. | 21:08 |
| kfox1111 | and we write a gate test that looks in variables for each microserices chart, | 21:08 |
| kfox1111 | and seeis if its subcharts vars are in it. | 21:08 |
| *** haplo37 has quit IRC | 21:08 | |
| kfox1111 | if not, we reject the merge. | 21:09 |
| *** rstarmer has joined #openstack-kolla | 21:09 | |
| kfox1111 | then at least breaking changes arn't pushed in. | 21:09 |
| kfox1111 | feels a little wierd using vars that way, but might feel better to others? | 21:10 |
| rhallisey | anyone have a galera config handy? | 21:10 |
| *** rstarmer has quit IRC | 21:10 | |
| portdirect | rhallisey: na you should use mangdb, it's web scale | 21:11 |
| *** haplo37 has joined #openstack-kolla | 21:11 | |
| kfox1111 | portdirect: heh | 21:11 |
| portdirect | rhallisey: sorry to me of abosulty no help | 21:11 |
| portdirect | *be | 21:11 |
| rhallisey | darn | 21:11 |
| rhallisey | want to get galera on kube going | 21:12 |
| kfox1111 | rhallisey: sek. let me see what I can dig up | 21:12 |
| rhallisey | I think I keep screwing up the config | 21:12 |
| *** sacharya_ has quit IRC | 21:12 | |
| SamYaple | hey rhallisey. how ya been? | 21:12 |
| *** sacharya has joined #openstack-kolla | 21:12 | |
| rhallisey | SamYaple, hey! | 21:13 |
| rhallisey | SamYaple, ok you can probably help :) | 21:13 |
| SamYaple | oh boy | 21:14 |
| rhallisey | does the placement of wsrep config settings matter in the config file? | 21:14 |
| SamYaple | no! | 21:14 |
| SamYaple | well, you mean order? | 21:14 |
| rhallisey | ya that order | 21:14 |
| SamYaple | no! | 21:14 |
| rhallisey | ok that's good | 21:14 |
| kfox1111 | rhallisey: http://pastebin.com/bt7JT6ZF | 21:14 |
| rhallisey | for my bootstrap do I want to specify all the nodes that will join the cluster | 21:15 |
| kfox1111 | rhallisey: no. | 21:15 |
| rhallisey | or should I add them with --gcom... | 21:15 |
| kfox1111 | but the trick is keeping all the ips in it and only changing to bootstap. | 21:15 |
| SamYaple | rhallisey: so im not sure if this is more or less useful for you, but you can start a new cluster without it all coming up | 21:15 |
| *** gomarivera has joined #openstack-kolla | 21:16 | |
| SamYaple | rhallisey: for example 'gcomm://server01,server02,server03?pc.wait_prim=no' | 21:16 |
| portdirect | rhallisey: might be worth reaching out to, alanmeadows? think he's got galera running in k8s? | 21:16 |
| *** haplo37 has quit IRC | 21:16 | |
| SamYaple | that wil lallow galera to start | 21:16 |
| rhallisey | https://paste.fedoraproject.org/505510/48157742/ | 21:17 |
| SamYaple | but it wont be working until you set "pc.bootstrap=1" on one of the hosts | 21:17 |
| kfox1111 | SamYaple: scary. :) | 21:17 |
| kfox1111 | ah. interesting. | 21:17 |
| rhallisey | ^ that paste has my config file | 21:17 |
| SamYaple | kfox1111: its really cool actually, because then mysql responds to commands and you can check WSREP variables to check the state | 21:17 |
| kfox1111 | right. so would be great for bootstraping. | 21:18 |
| rhallisey | I bootstrap and start both services, but the result is nothing in my cluster and the second service failing to start | 21:18 |
| SamYaple | i decided against it for the ansile implemenation because i found it cleaner (and more understandable) to not use it | 21:18 |
| kfox1111 | but could allow load balancers to talk to members, so shouldn't be used for 2 day ops. | 21:18 |
| SamYaple | but for k8s.... i dont know. it might make sense | 21:18 |
| alanmeadows | We have 3 x mariadb bootstrapping working (not using pets until https://github.com/kubernetes/helm/pull/1657 is solved) | 21:18 |
| *** haplo37 has joined #openstack-kolla | 21:18 | |
| SamYaple | kfox1111: im not sure about that. i dont know if all users can auth when its in that stated | 21:18 |
| kfox1111 | SamYaple: ah. yeah. if it was root only, then it might still work. | 21:19 |
| kfox1111 | unless you were foollish enought to use root for everything. ;) | 21:19 |
| SamYaple | anyway. just an option | 21:19 |
| rhallisey | alanmeadows, awesome | 21:19 |
| *** srwilkers has quit IRC | 21:20 | |
| *** sdake has quit IRC | 21:20 | |
| portdirect | alanmeadows: i've been using the image in that ps - works well :) | 21:20 |
| alanmeadows | mariadb:https://github.com/att-comdev/aic-helm/tree/master/mariadb - it hasnt gone through our latest layout refactor | 21:22 |
| alanmeadows | (like keystone) | 21:22 |
| *** diablo_rojo has joined #openstack-kolla | 21:23 | |
| qwang | Horizon service needs to be exposed to 2 ports(80 and 443). The current _common_svc does not support exposure to multiple ports. Do I modify the _common_svc template, or just define two service objects for horizon? | 21:24 |
| jascott1_ | qwang i think just add the additional one to the yaml directly | 21:30 |
| *** haplo37 has quit IRC | 21:31 | |
| qwang | jascott1_: you mean horizon-svc.yaml? It is using the common-svc template and I'm not sure if I can just add another port in the yaml | 21:31 |
| *** gomarivera has quit IRC | 21:34 | |
| *** fguillot_ has quit IRC | 21:34 | |
| *** Jeffrey4l has quit IRC | 21:34 | |
| *** Jeffrey4l has joined #openstack-kolla | 21:36 | |
| *** sdake has joined #openstack-kolla | 21:36 | |
| portdirect | qwang: i think you can add it into the horizon-svc.yaml | 21:37 |
| *** sdake has quit IRC | 21:38 | |
| *** haplo37 has joined #openstack-kolla | 21:38 | |
| *** sdake has joined #openstack-kolla | 21:39 | |
| *** sdake_ has joined #openstack-kolla | 21:40 | |
| *** gfidente has quit IRC | 21:41 | |
| jascott1_ | qwang it might be tricky to find where it goes exactly but should be possible | 21:43 |
| qwang | https://www.irccloud.com/pastebin/meXkts3w/ | 21:44 |
| qwang | like this? | 21:44 |
| *** sdake has quit IRC | 21:44 | |
| qwang | I'm not quite familiar with how the template works | 21:44 |
| mgiles | qwang I don't think that'll work. Taht would give you two complete service def's just appended to each other | 21:45 |
| mgiles | I'm not sure what portdirect was suggesting. Not using the common_svc template at all or if he had other thoughts how to plug it in. | 21:46 |
| *** gagehugo has left #openstack-kolla | 21:46 | |
| qwang | mgiles: I can put them into two files. That would work, right? | 21:47 |
| *** jgriffith is now known as jgriffith_AutoAw | 21:47 | |
| jascott1_ | qwang i was talking bout the deployment yaml, i thought it would take 2 there but not sure | 21:47 |
| jascott1_ | the svc can be whatever port but proabably 443 | 21:48 |
| jascott1_ | idk do we need both? | 21:48 |
| sdake_ | sup peeps | 21:48 |
| qwang | I think 80 for http and 443 for https | 21:49 |
| jascott1_ | sdake whassup | 21:49 |
| mgiles | hey sdake_ | 21:49 |
| *** sdake_ is now known as sdake | 21:49 | |
| sdake | sup mgiles | 21:49 |
| v1k0d3n | kfox1111: can you do me a favor? | 21:49 |
| sdake | had to collect my freshman from school | 21:50 |
| mgiles | things quieted down quite a bit after you left. Coincidence, I think. :) | 21:50 |
| v1k0d3n | i think i have a good contact for helm, who can explain things a bit better. (been helping us). | 21:50 |
| sdake | mgiles must be | 21:50 |
| mgiles | qwang You could do two files, but that's two services each would have to have different names. I think logically this is a single service with two ports | 21:51 |
| mgiles | which nobody else has hit specifically yet | 21:52 |
| *** jgriffith_AutoAw is now known as jgriffith | 21:52 | |
| qwang | mgiles: then let me try to tweak _common_svc template a little bit to support multi ports | 21:53 |
| mgiles | qwang I also thing jascott1_'s question is legitimate. I'm not sure if they both are used simultaneoulsy or if it depends on how horizon is configured in which case we could make the port configurable and just expose one. I don't know enough about horizon to know the answer though | 21:55 |
| *** lrensing has quit IRC | 21:55 | |
| jascott1_ | tempting to just say 80 on container and 443 on service but some orgs want TLS all the way | 21:56 |
| Pavo | has anyone used linuxbridge instead of ovs with kolla yet? | 21:57 |
| portdirect | hey sorry - got sidetracked, having thought about it i think just exposing on port 80 for now makes sense? as it should be configurable intill we have a way for adding tls 443 would not be required? | 21:58 |
| *** Satya_ has quit IRC | 21:58 | |
| sdake | we can really tackle tls at a later date I think unless soemone is keen to write the code for it now, then feel free ;) | 21:58 |
| * jascott1_ hides behind lamp | 21:59 | |
| sdake | tls is a real requirement medium term | 21:59 |
| sdake | right now feels a bit premature to tackle ;) | 21:59 |
| Pavo | anyone? | 22:00 |
| sdake | jascott1_ we can see your shadow ;-) | 22:00 |
| mgiles | I hate to default to insecure, but I think that makes sense anyway | 22:00 |
| sdake | mgiles it took kolla-ansible about 12-15 months to get tls implemented | 22:00 |
| mgiles | ugh | 22:01 |
| sdake | the actual work was probably 2-3 weeks | 22:01 |
| sdake | had to have the framework in place first | 22:01 |
| sdake | tls is easy with haproxy | 22:01 |
| sdake | we have code patterns for it already | 22:01 |
| mgiles | hopefully will be easier in kolla-kubernetes with existing patterns in place | 22:02 |
| portdirect | even easier with ingress controllers :) | 22:02 |
| sbezverk_ | why do not we put haproxy as a side car to horizon and terminate 80 and 443 on ha proxy | 22:02 |
| *** sayantan_ has quit IRC | 22:02 | |
| sbezverk_ | in this case we could re-use | 22:03 |
| sdake | mgiles my thinking as well | 22:03 |
| sbezverk_ | what kolla uses | 22:03 |
| jascott1_ | what do I put in PS to let everyone know tis just a check point commit and not to go nuts on reviews yet? "checkpoint"? | 22:03 |
| sdake | sbezverk_ wfm :) | 22:03 |
| sdake | jascott1_ WIP: title | 22:03 |
| sdake | WIP = work in progress | 22:03 |
| jascott1_ | oh ok that hasnt had that effect but alrighty | 22:03 |
| sdake | another one people use is "DNM" | 22:04 |
| sdake | do not merge | 22:04 |
| sdake | jascott1_ ya people keen to review to learn from each other | 22:04 |
| mgiles | jascott1_ I've found giving it workflow -1 seems to keep people from looking at it | 22:04 |
| sdake | jascott1_ thats the beuty of gerrit :) | 22:04 |
| sdake | mgiles ya that works well | 22:04 |
| mgiles | at least too seriously | 22:04 |
| jascott1_ | cool thx guys | 22:04 |
| sdake | people see that big X and ignore it :) | 22:04 |
| *** diablo_rojo has quit IRC | 22:05 | |
| mgiles | sdake since i'm new here maybe you can explain. What's the practice with +2 vs workflow +1? I assumed it was two reviewers but sometimes seems to be a single reviewer. | 22:06 |
| sdake | mgiles ya - since our core team is small on kolla-kubernetes, we typically have 1 +2 and 1+w | 22:06 |
| sdake | mgiles once the core team is sufficient in size, we will be using 2 +2 1 +w | 22:06 |
| mgiles | are the +2 and +w different reviewers or the same? | 22:07 |
| sdake | mgiles the reason for 2+2s is to make sure the code is sufficient quailty | 22:07 |
| sdake | mgiles the first core reviewer applies the first =2 | 22:07 |
| sdake | +2 | 22:07 |
| sdake | the second core review applies the second +2 and +w | 22:07 |
| sdake | gotta pick up son from school bbi15 :) | 22:07 |
| mgiles | sdake thanks - see you later | 22:08 |
| Pavo | has anyone used linuxbridge instead of ovs with kolla yet? | 22:15 |
| openstackgerrit | Merged openstack/kolla: Add zookeeper image https://review.openstack.org/406878 | 22:19 |
| Pavo | guess not | 22:19 |
| *** rhallisey has quit IRC | 22:20 | |
| *** bmace has quit IRC | 22:20 | |
| *** bmace has joined #openstack-kolla | 22:21 | |
| openstackgerrit | Justin Scott proposed openstack/kolla-kubernetes: WIP Helm-ize Heat https://review.openstack.org/409959 | 22:21 |
| *** jascott1_ has quit IRC | 22:21 | |
| *** jascott1_ has joined #openstack-kolla | 22:21 | |
| *** jgriffith is now known as jgriffith_AutoAw | 22:23 | |
| *** matrohon has quit IRC | 22:24 | |
| *** lbeliveau_ has joined #openstack-kolla | 22:25 | |
| *** lbeliveau has quit IRC | 22:26 | |
| *** dmsimard has quit IRC | 22:26 | |
| *** dmsimard has joined #openstack-kolla | 22:26 | |
| *** dmsimard has quit IRC | 22:27 | |
| *** sdake_ has joined #openstack-kolla | 22:27 | |
| *** sdake_ has quit IRC | 22:27 | |
| *** dmsimard has joined #openstack-kolla | 22:28 | |
| *** sdake_ has joined #openstack-kolla | 22:28 | |
| *** Pavo has quit IRC | 22:29 | |
| *** dmsimard has quit IRC | 22:29 | |
| *** Pavo has joined #openstack-kolla | 22:29 | |
| *** sdake has quit IRC | 22:30 | |
| openstackgerrit | Qin Wang (qwang) proposed openstack/kolla-kubernetes: WIP - Hel-ming horizon components https://review.openstack.org/408975 | 22:32 |
| *** msimonin has quit IRC | 22:32 | |
| *** dmsimard has joined #openstack-kolla | 22:37 | |
| openstackgerrit | Merged openstack/kolla-kubernetes: Helm creation of Nova services https://review.openstack.org/409473 | 22:42 |
| *** msimonin has joined #openstack-kolla | 22:43 | |
| openstackgerrit | Merged openstack/kolla-kubernetes: Helm for installing other nova services https://review.openstack.org/409486 | 22:43 |
| *** sayantan_ has joined #openstack-kolla | 22:46 | |
| *** lbeliveau_ has quit IRC | 22:47 | |
| *** sayantan_ has quit IRC | 22:48 | |
| *** sayantan_ has joined #openstack-kolla | 22:48 | |
| openstackgerrit | Mark Giles proposed openstack/kolla-kubernetes: Helm database creation for Nova https://review.openstack.org/409474 | 22:49 |
| openstackgerrit | Steven Dake proposed openstack/kolla: Enable zookeeper for CentOS/OracleLinux https://review.openstack.org/409965 | 22:51 |
| *** lbeliveau has joined #openstack-kolla | 22:53 | |
| *** msimonin has quit IRC | 22:54 | |
| *** msimonin has joined #openstack-kolla | 22:56 | |
| *** jgriffith_AutoAw is now known as jgriffith | 22:57 | |
| *** jheroux has quit IRC | 23:01 | |
| kfox1111 | v1k0d3n: sorry. was in a metting, and headed to another. | 23:03 |
| kfox1111 | v1k0d3n: yeah. would be happy to talk to folks with more helm experience. | 23:03 |
| *** jascott1_ has quit IRC | 23:04 | |
| *** jascott1_ has joined #openstack-kolla | 23:04 | |
| openstackgerrit | Steven Dake proposed openstack/kolla: Enable zookeeper for CentOS/OracleLinux https://review.openstack.org/409965 | 23:05 |
| *** jascott1_ has quit IRC | 23:06 | |
| *** jascott1_ has joined #openstack-kolla | 23:06 | |
| *** jascott1_ has quit IRC | 23:11 | |
| *** yuanying has joined #openstack-kolla | 23:13 | |
| *** msimonin has quit IRC | 23:15 | |
| *** srwilkers has joined #openstack-kolla | 23:17 | |
| *** diogogmt has quit IRC | 23:18 | |
| *** yuanying has quit IRC | 23:20 | |
| *** sayantan_ has quit IRC | 23:24 | |
| *** sayantan_ has joined #openstack-kolla | 23:24 | |
| *** mgiles has quit IRC | 23:25 | |
| *** yuanying has joined #openstack-kolla | 23:25 | |
| *** TxGirlGeek has quit IRC | 23:28 | |
| *** vhosakot has quit IRC | 23:36 | |
| *** mkoderer has quit IRC | 23:36 | |
| *** sp_ has quit IRC | 23:39 | |
| *** yatin_on_leave has quit IRC | 23:39 | |
| *** adrian_otto has joined #openstack-kolla | 23:40 | |
| *** JRobinson__ has joined #openstack-kolla | 23:42 | |
| *** yatin_on_leave has joined #openstack-kolla | 23:51 | |
| *** dave-mccowan has quit IRC | 23:53 | |
| *** sp_ has joined #openstack-kolla | 23:53 | |
| *** rhallisey has joined #openstack-kolla | 23:58 | |
| *** Bico_Fino has quit IRC | 23:58 | |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!