=== frankban|afk is now known as frankban | ||
assaf | hi guys , im trying to deploy the canonical distro of kubernetes in an offline enviorment , currently im stuck with deploying kube-proxy snaps , i tried manually downloading and attaching the snaps to the kubernetes-worker resource but they serivces fails to start | 17:22 |
---|---|---|
=== frankban is now known as frankban|afk | ||
ryebot | assaf: Hey, I can try to help you with that. Can we start with which snaps you downloaded and attached? | 17:51 |
ryebot | assaf: I'd also like to refer you to this documentation: https://github.com/juju-solutions/bundle-canonical-kubernetes/wiki/Running-CDK-in-a-restricted-environment | 17:52 |
ryebot | assaf: In particular, you may find the cdk-shrinkwrap tool helpful in your case. | 17:54 |
torontoyes | Is it possible to use juju to deploy windows 10 using Maas? | 18:43 |
jose-phillips | hi | 20:04 |
jose-phillips | is posible remove ceph from openstack juju charm? | 20:04 |
jose-phillips | and where the charms are stored? | 20:04 |
jose-phillips | to perform a modification? | 20:04 |
jamesbenson | stokachu:ping | 21:18 |
jamesbenson | ryebot: ping | 21:19 |
ryebot | jamesbenson: Something I can help with? | 21:20 |
jamesbenson | ryebot: I was wondering if you could provide some input on canonical k8s deploy | 21:21 |
ryebot | jamesbenson: Yes, definitely | 21:22 |
jamesbenson | ryebot: It's completed (I think) the conjure up, all services are active, and in juju status everything is active and idle, however, the conjure up window hasn't gone away yet... I thought there was post conjure things that happened. | 21:22 |
jamesbenson | ryebot: I can send screen shots if you want | 21:22 |
jamesbenson | ryebot: In conjure-up it has a "Setting relationship kubeapi-load-balancer:apiserver <-> kuberneters-master:kube-api-endpoint" | 21:23 |
jamesbenson | ryebot: But has been there for a while now.... 1+ hr. | 21:24 |
ryebot | alright | 21:24 |
* ryebot thinks | 21:24 | |
ryebot | jamesbenson: can you open up another terminal and take a look at `juju status` | 21:25 |
jamesbenson | ryebot : yeah, I'm actively looking at that as well | 21:26 |
ryebot | jamesbenson: okay great, can you send me a pastebin of that? | 21:27 |
jamesbenson | http://paste.ubuntu.com/26060543/ | 21:27 |
ryebot | perfect, one sec | 21:27 |
ryebot | jamesbenson: it appears that everything has deployed successfully. | 21:28 |
ryebot | jamesbenson: There may have been a hiccup in the conjureup output, but that looks good to me. | 21:28 |
jamesbenson | https://snag.gy/liuxHj.jpg | 21:29 |
jamesbenson | ryebot: so where do I go from here then? | 21:29 |
=== petevg_afk is now known as petevg | ||
jamesbenson | ryebot : kubectl get nodes | 21:29 |
ryebot | jamesbenson: yeah, you can log into one of the boxes and go to town with kubectl | 21:30 |
ryebot | jamesbenson: or copy out the kubeconfig and use kubectl locally | 21:31 |
jamesbenson | ryebot : do you have a command for that? which node is it in? | 21:31 |
ryebot | jamesbenson: all workers and masters will have a copy | 21:32 |
jamesbenson | okay | 21:32 |
ryebot | `juju ssh kubernetes-master/0` | 21:32 |
jamesbenson | yep, found the config file | 21:32 |
jamesbenson | copied it locally | 21:32 |
jamesbenson | ryebot: having issues the command: kubectl --kubeconfig=~/.kube/config (I made the dir & copied the config file here). The command isn't valid it seems... | 21:36 |
ryebot | jamesbenson: do you have kubectl in your path? | 21:37 |
ryebot | ie can you kubectl --help ? | 21:37 |
jamesbenson | yes | 21:37 |
jamesbenson | http://paste.ubuntu.com/26060596/ | 21:38 |
jamesbenson | ryebot: that's the output when I issue the kubeconfig command above. | 21:39 |
ryebot | jamesbenson: oh, it may not expand ~ | 21:39 |
ryebot | jamesbenson: try it with the absolute path | 21:39 |
jamesbenson | ok | 21:39 |
jamesbenson | ryebot: same issue | 21:40 |
jamesbenson | ubuntu@k8s:~/.kube$ kubectl --kubeconfig=/home/ubuntu/.kube/config |pastebinit | 21:41 |
jamesbenson | http://paste.ubuntu.com/26060615/ | 21:41 |
ryebot | jamesbenson: alright hmm | 21:42 |
jamesbenson | can I safely kill the conjure-up screen? | 21:42 |
jamesbenson | ryebot: I actually deployed twice on two different machines... both are stuck on the conjure up screen with different status' but same juju status, active/idle | 21:43 |
jamesbenson | ryebot: The other machine says it's stuck at "Setting relation easyrsa:client <-> kubernetes-master:certificates | 21:43 |
jamesbenson | " | 21:43 |
ryebot | jamesbenson: I -think- so, but am not entirely sure - maybe stokachu can weigh in | 21:45 |
jamesbenson | https://snag.gy/rUToJM.jpg | 21:45 |
ryebot | jamesbenson: oh, I'm being silly - you'll need to tack a command onto that | 21:45 |
jamesbenson | ? | 21:45 |
ryebot | jamesbenson: so, `kubectl --kubeconfig=... get po` | 21:45 |
jamesbenson | ok | 21:45 |
stokachu | relations are there | 21:45 |
stokachu | you can kill conjure-up | 21:45 |
ryebot | stokachu: thanks :) | 21:45 |
jamesbenson | ah yeah, get nodes worked | 21:46 |
stokachu | cory_fu: ^ i see this on internal testing ci as well, where the deployment in conjure-up will stop at setting relations | 21:46 |
jamesbenson | is there a way to set the config so I dont' always have to include that? | 21:47 |
jamesbenson | ah, I guess when I moved it into that dir, it did it... | 21:47 |
stokachu | i think there is an environment var | 21:47 |
jamesbenson | kubectl get nodes works now without issue. | 21:47 |
stokachu | we also want to do some magic later on to merge the configs and have you switch by context | 21:48 |
stokachu | just not done yet | 21:48 |
jamesbenson | sweet, I'm still pretty new to k8s, so learning a lot. | 21:48 |
jamesbenson | thank you stokachu and ryebot | 21:48 |
stokachu | np | 21:48 |
stokachu | jamesbenson: hit us up in #conjure-up if you have any more specific questiosn around that | 21:48 |
jamesbenson | out of curiousity, where are you guys based out of? I'm always curious how ^_^ | 21:48 |
cory_fu | stokachu: conjure-up stops, but the relations exist? That sounds like the AllWatcher not seeing the relations come through. Not sure if that we could handle that in libjuju in a reasonable way. Maybe a timeout after which it re-inspects the complete model status? | 21:49 |
cory_fu | Seems heavy-handed | 21:49 |
cory_fu | We've already | 21:49 |
stokachu | cory_fu: yea, it's why majority of the auto ci runs fail | 21:49 |
cory_fu | *gotten complaints about pulling in the full model state on connect | 21:49 |
stokachu | jamesbenson: im EST | 21:49 |
cory_fu | "complaints" | 21:49 |
cory_fu | comments | 21:49 |
stokachu | "suggestions" | 21:49 |
jamesbenson | stokachuL nice I used to be in NY, now in Tx | 21:50 |
=== admcleod_ is now known as admcleod |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!