[00:09] kwmonroe: i just added an outbound static route [11:55] zeestrat: it is only for [General] section - this item should go to [libvirt] section so I tried even injecting this but does not work [11:58] i have some units stuck with the agent lost and workload unknown [11:58] is is possible to remove them without destroying the application [11:58] ? [11:58] goooooo kjackal [12:19] oh remove-machine --force did it [12:19] cancel that [13:28] hi magicaltrout, destroy the machine holding the broken agents? [17:04] question. what happens if my controller is unavailable or dead and gone forever. and i dont have a second one. will there be a chance to manage the models still? [17:04] is it possible to create a new one and say that this controller should manage model xxx [17:05] those are questions i took from our architecture meeting [17:05] since controllers are not HA when is it planned to implement this [17:38] ybaumy, the model is lost if its controller is gone [17:42] can anyone out there help with a juju charm deployment issue? [17:43] ceph charm deployment issue === frankban is now known as frankban|afk [17:49] pmatulis: hmm what is the recommandation of how many controller units i should have [17:51] does it even make sense to distribute controllers accross zones or is this only good for datacenters in vsphere [17:54] also what i noticed is that when deploying machines are setup in sequential order [17:54] not parallel [17:54] why is that [17:54] when i do juju status i see only one doing something [17:55] i'm sorry but i'm not fluent in juju ha but i gather the more ha units, and distributed, the better, in a sense. [17:55] not sure what you mean by 'only one doing something' [17:56] pmatulis: i see only one VM donwloading a image [20:11] How do I successfully deploy my first juju charm? [21:41] akern07, with the 'juju deploy' command. `juju deploy mysql` [22:00] akern07: is this a charm you wrote or just any charm? [22:04] akern07: the best thing to do is walk through the first tutorial: https://jujucharms.com/docs/stable/getting-started [22:13] Hello [22:14] howdy elmaciej [22:15] Does anyone knows how to add cinder volumes with different volume-type thru juju ? so I have profile ssd and hdd but want to do it thru juju [22:21] also simpler question - how to block juju from rolling back my changes in config file for example nova.conf [22:22] elmaciej: so Juju has a config for the charm that as it's updated it writes out the file. If you want to override it you'll have to see if the charm allows special injection as well or if there's a second file you can edit/leave that is slurped into the generation of the final nova.conf [22:25] so I have to modify the charm if it's not provided? I'm talking about situation I posted before - deploying nova-compute, change instance disk type to lvm. this charm do not provide the images_volume_group property. you can inject only to GENERAL section thru charm [22:25] and when I add this manually juju reverts the nova.conf after some time [22:25] so have to either modify the charm or after deployment shutdown juju [22:26] elmaciej: right, because juju is syncing things when the charm config is changed and it's written to rewrite the file. The best bet would be to chat with the #openstack-charmers and see what provisions they've got in the charms to help with what specific changes you want to make around it [22:26] elmaciej: Juju can't know the current state and make things repeatable (e.g. dump out the same config that should work again on a second install) if it's done manually around Juju's control. [22:27] ok, thanks for confirming this - I was not sure . it would be good if you can for example set kind of unmanage flag for specific charm in the model [22:27] :) but it's just a wish to wishlist [22:29] elmaciej: yea, I don't have insights into the charm and the various control knobs it exposes but I know we use these charms for real prod customer openstack deploys so I'm assuming there's some method to the madness there just might take an ask to the mailing list or irc for the openstack folks to find it. [22:31] well it's so many variations and in my case is rare situation. Mostly people use ceph for storage and here I have just lvm on SAN. so it's hard to predict everything. but as I used mirantis before in production +100nodes clusters and also redhat I can say that juju is far way better and user firendly then all others distros. [22:31] and debugging is so much easier [22:32] so I love it really [22:37] elmaciej: nice to hear, sorry I don't have a firm answer for you there on the last bit you're hitting.