/srv/irclogs.ubuntu.com/2017/08/25/#juju.txt

=== frankban|afk is now known as frankban
armaanjamespage: hello, could i ask your opinion on https://github.com/juju/1.25-upgrade ? Can i use this for upgrading OpenStack environments?09:02
Ting_Hi, could people use other authentication-type for juju aws instead of access-key? We are using aws federation account seems very hard to get access key...09:13
rick_hTing_: we need to support that. There's an open bug that could use some real user pressure. Let me see if I can find it.11:57
Ting_rick_h: thx :)11:59
rick_hTing_: on my phone ATM and not able to find it. Let me look when I get to the office.12:00
Ting_rick_h: sure, no hurry12:04
tychicusjust deployed the microbot application in CDK 1.7 to get a better understanding of how the ingress controller and load balancer work.13:45
tychicuskubectl get ingress does not return an address13:45
tychicusthe only values returned are NAME, HOSTS, PORTS, AGE13:46
tychicusI can validate everything is working correctly by creating a /etc/hosts entry for microbot.10.148.0.105.xip.io and using the IP address provided in the hostname,13:48
SimonKLBtychicus: did you create it using the microbot action?13:51
tychicusSimonKLB: yes juju run-action kubernetes-worker/0 microbot replicas=313:52
SimonKLBdid you check the action status and/or output?13:52
SimonKLBalso, is it only the ingress that is missing or is the deployment also not there?13:53
tychicushttps://gist.github.com/roll4life/2cbaad645379f3db6998d2e08f70f66413:54
tychicushere is the output of kubectl get services,endpoints and kubectl get ingress13:55
SimonKLBah my bad, i thought you meant the ingress was never created13:55
tychicussorry, ingress is created, but the address is not displayed13:56
SimonKLBis the kubernetes-worker exposed?13:56
tychicusyes13:57
tychicuskubernetes-worker      1.7.0    active      3  kubernetes-worker      jujucharms   40  ubuntu  exposed13:57
SimonKLBhmm, can you check the logs of the ingress-controller?13:58
tychicusSimonKLB: updated the gist with the ingress-controller log output https://gist.github.com/roll4life/2cbaad645379f3db6998d2e08f70f66414:12
tvansteenburghtychicus: what's the output of kubectl get pods14:12
tvansteenburghis there an ingress controller running?14:12
tychicusyes 314:14
SimonKLBtychicus: just tried it out on my end and works fine, wierd..14:15
tychicusok gist updated14:15
SimonKLBcan you check which version of the ingress controller youre running? `kubectl get rc nginx-ingress-controller -o yaml | grep image:`14:16
tychicusimage: gcr.io/google_containers/nginx-ingress-controller:0.8.314:16
SimonKLBsame here14:17
SimonKLBso it's definately that "has no active endpoints" that's the issue14:17
SimonKLBor atleast that is what looks different from mine14:18
tychicusok thanks I'll look into that and report back what I find, thank you for your help14:20
SimonKLBtychicus: which provider are you running on btw?14:22
tychicuslocal openstack juju/maas deployment14:23
tychicusupdated the gist with the output of kubectl get ing microbot-ingress -o yaml14:29
SimonKLBthat's wierd, it looks like it's adding ips, but they are empty14:30
SimonKLBactually, looking at the log again it says "Updating loadbalancer default/microbot-ingress with IP *nothing*"14:30
tychicusright, just noticed that as well14:31
SimonKLBtychicus: can you paste your juju status output?14:34
SimonKLBif the workers doesn't have public ips that might be it14:34
tychicusupdated, they do have "public addresses"14:35
tychicusthey are rfc 1918 addresses, but they a "public" internally14:36
SimonKLByea14:36
SimonKLBtychicus: check if kube-proxy is running ok on the workers14:44
SimonKLByou can run this for example: `juju run --unit kubernetes-worker/0 systemctl status snap.kube-proxy.daemon.service`14:45
tychicusok, thanks, I was just getting ready to ask how to check the status :)14:46
tychicusthey all show as running all have these errors with ID 1,5,7,9,1114:49
tychicusUnable to decode an event from the watch stream: stream error: stream ID 1; INTERNAL_ERROR14:49
SimonKLBtychicus: you can get the full log if you run `juju run --unit kubernetes-worker/0 "journalctl -u snap.kube-proxy.daemon.service"`14:51
SimonKLBor ssh into the machine14:51
tychicusgladly14:51
Cynervatychicus: i think the behavior you're seeing with microbot is pretty typical15:45
Cynervaseems like ingress only gets an address assigned to it if you have external loadbalancer support15:45
Cynervawhich AFAIK requires cloud integration15:46
Cynervawe only get that on AWS when deployed via conjure-up, i think15:46
Cynervabut, the ingress is still usable without a loadbalancer15:46
tychicuscynerva: ok, thanks15:47
SimonKLBCynerva: wierd, because it's working fine here and im running it on LXD15:51
Cynervahuh15:53
Cynervathat is weird15:53
SimonKLByea :D15:53
Cynervai just had a go on AWS w/o native integration, got no address but it's working the way i'd expect15:56
SimonKLB% juju status | head -n 215:58
SimonKLBModel                        Controller           Cloud/Region         Version       SLA15:58
SimonKLBconjure-kubernetes-core-531  localhost-localhost  localhost/localhost  2.3-alpha1.1  unsupported15:58
SimonKLB% kubectl get ing15:58
SimonKLBNAME               HOSTS                           ADDRESS            PORTS     AGE15:58
SimonKLBmicrobot-ingress   microbot.10.212.38.141.xip.io   10.212.38.141...   80        1h15:58
SimonKLB:o15:58
tychicusyeah, that is pretty much what I expected to see :)15:59
SimonKLBwhats even more interesting is that my nodes does not have external ips:16:02
SimonKLB% kubectl get no -o wide16:02
SimonKLBNAME            STATUS    AGE       VERSION   EXTERNAL-IP   OS-IMAGE             KERNEL-VERSION16:02
SimonKLBjuju-cf3e22-3   Ready     16h       v1.7.0    <none>        Ubuntu 16.04.3 LTS   4.10.0-21-generic16:02
SimonKLBjuju-cf3e22-4   Ready     16h       v1.7.0    <none>        Ubuntu 16.04.3 LTS   4.10.0-21-generic16:02
SimonKLBCynerva: do you know if kubernetes or the ingress-controller somehow aware of it's environment and accknowledge the internal ips as external when running inside LXD or something?16:03
SimonKLBsince the ingress controller is actually adding the interal ip of the node:16:04
SimonKLB% kubectl get no juju-cf3e22-3 -o jsonpath='{.status.addresses}'16:04
SimonKLB[map[type:InternalIP address:10.212.38.141] map[type:Hostname address:juju-cf3e22-3]]%16:04
CynervaSimonKLB: not that i'm aware of, but could be16:06
=== frankban is now known as frankban|afk
bdxrick_h: sup17:59
bdxrick_h: is there anything on the roadmap for JAAS to support team owned models?17:59
rick_hbdx: not direction at the moment. There's some precursor work going on though18:00
rick_hbdx: e.g. not in the next release of work but it's on the radar and some groundwork is landing/going on18:00
bdxnice nice18:00
bdxthats good to know18:00
bdxI'm leaving a trail of models owned by myself across the technology community :)18:01
rick_hbdx: lol18:01
rick_hbdx: it's not been pushed because you can add/remove users and such so there's some level of control.18:02
rick_hso it's longer to do since you have to add each user vs a single team18:02
bdxI see18:02
kwmonroebdx: just share your creds with people you trust.  instant team!  we can put them in relevant irc /title bars if it makes life easier for you.18:03
rick_hbdx: and juju doesn't really have solid "group" idea on a self-operated controller so it needs some smarts18:03
rick_hhah18:03
rick_hkwmonroe: always here to save the day!18:03
kwmonroe:)18:03
bdx^ perfect example of what not to do18:07
bdx:)18:07

Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!