[01:05] Hey lazyPower have you run a redis cluster in k8s yet? [01:08] Budgie^Smore: Its been a minute, but yeah it was part of the petstore examples [01:09] yeah I was thinking of doing it stateful :) [01:09] although the google example only uses rcs [01:09] StatefulSet [01:09] yup [01:09] that would be a solid contribution back to their examples dir too [01:10] watching Hightower at RedisConf right now [01:11] Right on :) Got a livestream link? [01:11] just a youtube vid from last year [01:11] https://www.youtube.com/watch?v=0hlv9iIncik&list=PL83Wfqi-zYZHtHoGv3PcGQA3lvE9p1eRl&index=16 === lazyPwr is now known as lazyPower [07:58] Good morning Juju World! === frankban|afk is now known as frankban === mup_ is now known as mup [11:52] Hi! What is the username that each node uses when connecting with controller? [11:53] I can see the password on agent.conf but I don't figure out the username [12:00] tvansteenburgh: thanks, we will for sure make use of the new version of the juju lib as soon as possible, but i think that we won't be able to migrate to the new one in time for the MWC demo [12:45] Hi Folks. Does anyone know if charms for openstack release are available yet? [12:47] sorry i meant openstack ocata release [13:54] tvansteenburgh: have you seen scripts bombing out due to "OSError: [Errno 9] Bad file descriptor" before? [13:54] rick_h: yes, was just working on that yesterday [13:55] rick_h: typically happens when you don't clean up connections before closing the event loop [13:55] tvansteenburgh: any clue on what tossed it? the failures are coming out of python code vs my code so not sure wtf happend [13:55] tvansteenburgh: oic [13:55] rick_h: i can look at code or hangout if you want [13:55] tvansteenburgh: that's a good hint. I'm not cleaning stuff up. Will poke at it [13:55] rick_h: cool, lmk if i can help [13:55] tvansteenburgh: https://pastebin.canonical.com/180649/ is what I'm hacking on atm [13:56] tvansteenburgh: so assume I need to add some cleanup to my stress runs === jamespage_ is now known as jamespage [13:59] rick_h: yeah you probably want to await cont.disconnect() and await stress.disconnect() at the end of the stress func [13:59] tvansteenburgh: makes sense ty === admcleod_ is now known as admcleod === frankban is now known as frankban|afk === mup_ is now known as mup === mup_ is now known as mup [18:31] marcoceppi: ok our openstack charms release is all clear - feel free to move charm-helpers when you're ready [18:31] if there is anything I can do to help just shout [18:31] jamespage: thanks! [18:34] hi all. [18:35] after creating a juju controller via maas, when deploying a model to the same machine, the machine-0 endsup in the "down" state. How can I investigate? [18:35] I did the 2 following commands [18:35] juju bootstrap r16maas --debug --to osv16ocp1b [18:35] juju deploy juju-gui --to osv16ocp1b [18:36] I use Ubuntu Xenial, Juju 2.1.0 and MAAS 2.1.3 [18:37] when sshing into the "machine 0", the juju-machine-0 service can't be started. It says it's not found. [18:48] anyone care to explain to me the logic behind having the controller model instead of having the controller be part of the default model? [18:56] stormmore: because folks would bootstrap and start deploying [18:56] stormmore: and the controller model cannot be destroyed [18:56] stormmore: so this way, folks are put into a safe place right after bootstrap, the controller model is kept apart, and users are encouraged to use the default, create their own, and eventually remove the default model all safely [19:06] rick_h, I get that. Just seems over kill to have to assign a full machine / instance for juju. Not so big a deal with cloud instances but bare metal... [19:07] stormmore: you're completely free to switch over there and run things there [19:07] stormmore: that tends to be the use when folks use one controller per deploy, but in a multi-model world models are cheap [19:07] stormmore i usually make a VM on the server that's running MAAS and put the controller on it [19:07] so having one machine be the controller for dozens of models is what we're defaulting to [19:07] stormmore: what jhobbs said ^ [19:08] I know of folks that setup a lxd, register it in maas, and deploy to it in that way [19:08] rick_h, oh I know that too, just saying. most likely I will go multimodel at some point [19:08] stormmore: but especially if you head toward HA/etc we default to the middle ground [19:08] not too scale down, not so scale up [19:08] jhobbs, yeah that is the conclusion I came to too ;-) [19:08] for us it's a case of sane/safe defaults. There's more options available that will hopefully work for what you need [19:10] rick_h, I actually thought about doing it as an LXD on the same node I was installing MaaS on but was wondering if it would be better to have it on the node that is going to be the cluster master and run other services on it [19:10] stormmore: I guess "better" with what criteria is the question [19:14] rick_h, well the basic objective is to minimize the amount of resources required for managing a cluster. for instance I am planning on moving the maas postgres db into the cluster once it fully up [19:15] stormmore: interesting [19:15] The plan is to have a bootstrap node that it "removed" once a data center is fully functional [19:16] yes it means making sure everything is HA [19:37] rick_h, am I understanding you correctly, people are getting maas to manage some LXD containers as well as other bare-metal? [19:39] yeah, i've never heard of that :) if that's possible I'd like to know more [19:39] * rick_h checks his sources [19:40] you and me both jhobbs :) to me even VMs seems overkill for building a management node [19:40] ah, kvm is what I'm thinking of: https://blog.felipe-alfaro.com/2014/05/29/how-to-configure-maas-to-be-able-to-boot-kvm-virtual-machines/ and the like [19:41] still having issues here with the machine-0 reported down in the juju status. fresh bootstrap install via maas/physical, and trying to put the juju-gui on the machine-0 [19:41] yeah I was pondering that, just not as experienced with KVM as other virt tech :-/ [19:41] http://askubuntu.com/questions/665741/how-to-create-a-lxd-container-on-a-machine-and-register-it-in-maas for the lxd side [19:41] mimizone: what version of juju? [19:41] juju 2.1.0-xenial-amd64 [19:41] mimizone: juju gui comes bundled ootb [19:42] mimizone: just run juju-gui [19:42] sorry [19:42] "juju gui" [19:42] as for the reporting of down, is the machine down? can you ssh to it? [19:42] mimizone: this is after you "juju switch controller" and see machine 0 there? [19:42] yes I can ssh to the machine and I see jujud-machine-0 service is up [19:43] juju switch controller? I must have read the wrong documentation, never saw this step [19:43] mimizone: right, so in juju 2, there is the ability to operate multiple models [19:44] mimizone: and out of the box, you end up on an empty model w/o access to the controller itself so that it's safe [19:44] mimizone: you can switch to the controller and see machine 0 by using "juju switch controller" [19:44] mimizone: check out 'juju models' as well [19:44] mimizone: but the gui is built in so you can just run the 'juju gui' command and get it to open up. [19:45] Aaaaaaahh :) much better with the right explanation :) [19:45] good to see the gui is built in [19:45] mimizone: if you're new to juju2 I'd check out https://jujucharms.com/docs/stable/introducing-2 [19:45] rick_h: thanks [19:45] mimizone: and some videos from the past: https://www.youtube.com/results?search_query=juju+2.0 [19:46] I will check. I am doing all this for the JOID model for OPNFV [19:48] rick_h: thanks again. was completely far from understanding all the changes. [19:49] mimizone: all good, let us know if you have any questions [19:49] mimizone: hopefully some goodness in there for you to discover [19:49] https://insights.ubuntu.com/2015/01/15/virtualbox-extensions-for-maas/ is what I am using for inspiration right now [19:55] hey lazyPower so 2 days with a restart on kube-dns :) [20:35] cory_fu, kwmonroe: https://github.com/juju-solutions/matrix/pull/84 [21:39] kwmonroe: thx for the +1. I went ahead and merged the code :-) [21:42] i trust you completely petevg :) [21:42] I will strive to live up to your trust. Or something :-) [22:06] cory_fu: does this look familiar? http://paste.ubuntu.com/24061228/ i just built a local jenkins charm, now it's complaining about layer.options('basic') [22:08] kwmonroe: o_O The function is definitely there: https://github.com/juju-solutions/layer-basic/blob/master/lib/charms/layer/__init__.py [22:09] kwmonroe: Any chance you have an old copy of layer-basic checked out locally? [22:09] lol cory_fu, my bad [22:10] i'm dumb... "find . -name __init__.py -delete" is rarely what you wanna do. [22:10] ha [22:10] I was about to be worried [22:10] nothing like a catastrophic layer-basic failure at 5pm on a friday, #amirite? [22:11] :)