=== nevermind is now known as Guest85527 [00:48] @bdx Number 2 is asking me to log in as a specific user (you?) it didnt work for me. As for number 4, it gave a 404 not found error [00:49] fallenour: `juju deploy cs:~jamesbeedy/owncloud-4` ? [00:49] @bdx OOOOO!!!! [00:49] @bdx MAAYBE! [00:50] @bdx I will say that the charm page for it gave a 404 on the /4 page, so maybe an issue? im not sure? [00:50] hmm, thats odd [00:51] fallenour: this https://jujucharms.com/u/jamesbeedy/owncloud/4 ? [00:51] @bdx I do know that if this works, Ive gotten cloud data, pictures are next, and then the FreeIPA box, and then I gotta turn on SSL on everything [00:51] @bdx that loaded, much appreciated my friend [00:52] np [00:52] @bdx now lets just hope that DDNS with Afraid works well, and then HAProxy fine tuning, and then hopefully I can finally get some damn sleep. Its been 16 hour days for....what month is it? [00:55] hah - yea [00:59] fallenour: I should document in the readme, the owncloud charm exposes the http endpoint [01:00] @bdx yea, I was hoping for a simple juju expose owncloud && juju add-relationship owncloud haproxy [01:00] fallenour: so you could `juju deploy cs:~jamesbeedy/owncloud-4; juju deploy haproxy; juju relate haproxy owncloud;` [01:00] to get the ssl termination/front end [01:01] thats kindof the *basic* workflow as far as relating things to haproxy goes [01:01] @BDX YEAAA! Thats what im talkin' bout, nice and easy. im hoping with ssl itll be a smooth transition. FreeIPA lets me do PKI, and generate my own CA infrastructure. Im hoping it just accepts life and lets me do whatever I want without resistance. [01:04] fallenour: the most simple way to get the haproxy charm to terminate ssl for you, is to deploy it similar to this https://gist.github.com/jamesbeedy/d587cbf048038fb274ef4cd55c4ee3dd [01:05] fallenour: where the ssl_key, and ssl_cert are base64 encoded strings e.g. (`cat mycert.crt | base64; cat mykey.key | base64;`) [01:07] fallenour: once you fill in the ssl_key and ssl_cert configs with your own cert/key you can put ^ in `haproxy.yaml`, then deploy haproxy with `juju deploy haproxy --config haproxy.yaml` [01:07] fallenour: then when you `juju relate owncloud haproxy` haproxy will redirect to https and terminate the ssl for you [01:07] if you run into issues just give a shout [03:22] Question: is there a way to change cloud controller endpoint url after juju controller has been bootstrapped? [03:24] I've got a testing environment where I happened to change maas dns settings some and that broke juju environment that was using old dns names... [03:25] In this environment I could just wipe the controller, but out of interest, I would assume there is a way to update controller endpoint uri without wiping the controller and redoing bootstrap [03:33] can someone explain how your supposed to log in to a fresh keystone application? [04:00] Well, jury-rigged it by adding additional dns names with: "maas admin dnsresources create name=@ domain=test.domain.tld ip_addresses=a.b.c.d" === frankban|afk is now known as frankban [11:36] @bdx @rick_h @catbus It is Time o.o [11:37] * rick_h ponders what that means? [11:37] Wait! DDNS! and Then! It is Time o.o [11:39] oh shit :( [11:39] I totally forgot to FQDN everything in maas o.o [11:40] So....if I randomly start changing hostnames all over the place in systems, is that gonna kill MaaS and its ability to communicate with the hardware? [11:40] or will it ignore it, and continue to utilize IPMI? [11:43] hello. I'm dealing with a production incident, experiencing high load on our juju controller. It looks like: https://bugs.launchpad.net/juju/+bug/1703675 [11:43] Bug #1703675: PingBatcher could gracefully handle "DuplicateKeyError" [11:43] is there any workaround? [11:47] @jacekn you can build additional juju controllers to share the load with across multiple models, or put them in a shared load configuration [11:47] fallenour_: we already have 3 controllers [11:48] @jacekn add 3 more: https://jujucharms.com/docs/2.1/controllers-ha [11:49] @jacekn I know it sounds crazy, but believe in the Juju, it works. [11:52] fallenour_: is there any official documentation on this? ISTR that last time I checked juju HA did not cause it to scale horizontally (mongo writes wre still going through single node) [11:53] @jacekn "You cannot repair the cluster as outlined above if fewer than half of the original controllers remain available because the MongoDB replica set will not have the quorum necessary to elect a new master. You must restore from backups in this case." [11:53] have you verified your original controllers in the cluster are available? [11:54] fallenour_: my problem is not to do with quorum, all 3 controllers are still available it's just that one has high load [11:54] otherwise tear down your cluster, rebuild your cluster, resync the HA, that may fix your loadbalancing issue on your controllers [11:54] @jacekn I know that, but what that tells me is that if only 1 of them has a high load, its not loadbalncing across the controllers. [11:55] @jacekn hang on, I foudn your issue [11:55] "multiple redundant controllers are instantiated while the single active controller is defined as the master. Automatic failover occurs should the master lose connectivity." [11:56] @jacekn I think this means its not actually an "HA Cluster" Its like more like a hot spare. [11:56] its a feature? [11:57] @jacekn a work around would be to configure your systems to use a specified controller by DNS name, and then use HAProxy to loadbalance between the controllers that way. Im not sure if thatll work though. @stokachu @catbus @bdx @rick_h Your Thoughts? Is it possible to load balacne between controllers via hostname? [12:01] tinwood: You available? [12:01] @cory_fu whats up? [12:01] cory_fu, omw [12:04] fallenour_: Hey, what's up? [12:23] Hello there. Is it possible for juju, if i have all the charms locally available to install charms when fully offline (no connection to the internet)? [12:23] I think that i need to have a apt-mirror for the ubuntu version i want, i also need a local mirror of the cloud repo [12:23] but what i don't know is if and how it should be possible for the LXD images to be downloaded correctly [12:24] @blackdex make a localized repo copy [12:24] @blackdex I do localized repos due to my 6/1 connection at my current location. I carry servers over to a friends house, download from his 1Gb connection, and then carry the servers back. [12:26] @blackdex There is a juju charm for a localized repo, please see here: https://jujucharms.com/docs/1.21/howto-offline-charms [12:28] correction: That is what you are looking for, there is also a charm for a localized repo for debian/ubuntu packages, which I recommend you also get. [12:32] but stuff for OpenStack like neutron, nova etc... will not support that i think [12:33] @blackdex it will, just use HAProxy or Nginx with proper DNS And FQDN practices [12:35] so you say that i need to catch the fqdn's of the repo's to point to a local mirror :) [12:36] so if i use maas, and i have a local repo on the maas server available, i just point archive.ubuntu.com to 127.0.0.1 ;) [12:38] tinwood: proposed tag updated [12:39] thanks cory_fu :) [12:57] what's the equivalent of `juju unregister` for a model? [12:58] @blackdex I would recommend you simply use local.repo.local, and use your dns/dhcp server to point it to a local 10. ip space [13:08] tvansteenburgh: You can `juju unregister` a controller, but I don't think you can do it on a per-model basis [13:09] Maybe ungrant yourself permissions? [13:12] tvansteenburgh: there isn't. bdx and I were lamenting that and there's conversations with core about how to deprecate the idea of "owner" that users can't change or work around [13:17] rick_h: cory_fu: i just want a way to clean up my `juju models` list when i know a thing doesn't exist any more, but juju thinks it does [13:22] can someone remind me where I get the charm build plugin from? [13:23] cory_fu, tvansteenburgh can you remember ^^? [13:25] mattyw: snap install charm [13:25] tvansteenburgh, https://pastebin.canonical.com/198727/ [13:26] tvansteenburgh, ah right, it's a problem related to snap/ ppa interaction [13:26] sigh [13:26] tvansteenburgh, ok thanks [13:26] np === andreww is now known as xarses_ [15:36] tvansteenburgh, if multiple set_states are called in a reactive charm do you know which order they get executed in? stack or queue? [15:42] mattyw: cory_fu might be able to help with that ^^ [15:43] mattyw: The order is specifically undefined. [15:44] mattyw: If you need to enforce an ordering for handlers, you should do so by chaining flags (states). [15:45] That is, handler 1 reacts to flag A and then sets flag B. Then, handler 2 reacts to flag B. === frankban is now known as frankban|afk [16:56] months of working, years of planning, boiled down to 20 lines of code. Either itll work, or itll work when I threaten it with a hammer. [16:59] o/ juju world [18:48] ok, down to the last item [18:48] I'm attempting to destroy a model that has machines that failed to deploy, and the destroy-model is stuck waiting for the machines, is there a way to kick this down the road? [18:48] I need to get nginx to forward from internal IP to internal server with relelvant service, and I feel like im missing something [18:49] @xarses_ @catbus @rick_h @bdx ive got DDNS pointed with all my subdomains to my WAN IP. My Router forwards all 80 and 443 stuff to my Nginx box. [18:49] ? [18:49] If I use my browser to go to panda.eduarmor.com, it loads my gninx server html welcome message, but not the application its supposed on panda.eduarmor.com/horizon [18:51] example: http://panda.eduarmor.com [19:02] @rick_h: https://jujucharms.com/docs/search/?text=zone is giving an error, but something like https://jujucharms.com/docs/search/?text=charm works fine [19:41] good, god, why is the handeling of availability-zone so frigging hard [19:49] xarses_: filed a bug on that. Not sure but clearly something in the search for zones has caused the app to go ape [19:50] xarses_: and AZ are supposed to be automatically handled across mulitiple units. It's a bit too fine grained for most use so I don't think a lot of work's been put into exposing those knobs in a clean way. [19:50] 'asdf' has a similar response, but maybe because there are no responses [19:50] xarses_: what are you trying to do? [19:50] xarses_: yea, but zone should have stuff in there I'd imagine [19:50] ya, thats why I pointed that one out [19:50] xarses_: https://github.com/juju/docs/search?utf8=%E2%9C%93&q=zone&type= heh [19:51] rick_h: my network and availability zone must be passed together (provider networks before routed-provider-networks support) [19:51] network is set on the model [19:51] there is no valid constraint for zone [19:51] zone can be passed on --to [19:52] or as zone= on a add-machine [19:52] but there is no way to specify a zone when there is more than one machine being asked for [19:52] or the charm wants to spin up multiple members [19:53] or when deploying a bundle [19:53] so the --to constraint works for machines and you can specify them in the bundle and then the application to go on which machine [19:54] but yea, if you ask for 5 things then it splits one into each AZ [19:54] ya, thats invalid for my network [19:55] (zone:one, nets: [one-vlan11, one-vlan12])(zone:two nets:[two-vlan11, two-vlan12]) [19:56] you can not specify them in the bundle [19:56] there is no respected arg for that [19:56] end exporting it always looses what ever was manually set when deploying [19:57] it just keeps coming back to, I have a constraint of network and az pair, and I can't specify it [20:28] hmm... [20:28] certainly seems like a failing in the ability to easily place things [20:37] hi, I'm currently having issues in scaling the charm units.. When I scale any app with add-unit command, new unit gets deployed but it goes in error state. [20:38] That makes all of the subordinate charm deployment to wait until the parent charm error is resolved [20:38] is there any way to force the subordinate charm deployment even if the parent is in error? [21:53] so weird question of he day, when looking at juju status is the relationship section suppose to only show the relationships for ? [21:56] * stormmore is looking at the output from juju status in 2.3 from the edge snap [22:08] thumper: certainly is, won't be resolved anytime soon, the network is too complex. at a minimum have to switch to a new SDN, or upgrade from mitaka to at least okata [22:09] and I can't switch to the new SDN until I unwind and change some of these juju charms, which I cant do in my lab easily because I have to sepecify the zone in odd ways because it isn't treated as a constraint