[00:08] hello, is it possible for a charm to know that it is related a mysql db vs a postgresql db, or is that something that would need to be in the configuration? === spammy is now known as Guest81296 [03:07] magicaltrout: which meeting was that? [06:52] Could somebody help me juju 2.0 uninstallation steps on ubuntu? [07:24] uninstallation? [07:24] apt remove --purge juju-2.0 (ish) [07:25] then rm -rf ~/.local/share/juju if anything is left there === jamespag` is now known as jamespage [08:12] Hi all [08:13] I have bootstrapped the bootstrap node and am now trying to access the juju gui. I can reach the login page, but nothing ever happens when clicking the login button. Does anyone know, what could be the problem? [08:15] I have checked with the developer tools in firefox and I can't see any activity [08:16] Mainly I'm using Firefox, but I have the same effect with Chrome [08:17] Bonjour MonsieurBon! Hmm. Can you see anything interesting in the logs? Try "juju debug-log" while visiting the login page? [08:19] No, nothing. Should I execute juju debug-log on the host, where juju was installed initially or on the bootstrap node? [08:20] The dev tools don't show any activity at all. No logs, no network activity, nothing. I also can't see any errors there. [08:20] On the host - whereer [08:20] Oopd [08:20] gah [08:20] Wherever you are running juju status. [08:21] ok, I did that. It doesn't show anything [08:21] So you don't even see a request back to the server when you click login? [08:21] Nope, nothing at all [08:21] i've not been keeping up with the gui compatability. is there a gui <> beta offset currently or should it work? [08:22] It does call /api when loading the page, but that just returns a 101 response. And no XHR calls after that [08:23] just to let you know, I did not install the juju-gui charm manually, but am accessing the automatically deployed gui on the bootstrap node. [08:23] magicaltrout: good point - there was that breaking api change. [08:23] MonsieurBon: What version of juju are you running? [08:24] 2.0-beta7-xenial-amd64 [08:24] The bootstrap node is running trusty though. Might this be a problem? [08:25] I don't think so. That's a pretty old beta though - 12 is just out. There have definitely been some bugs with the GUI, so it would be worth upgrading and rebootstrapping. [08:25] this is just a test setup, so I can switch the juju node to trusty any time [08:27] apt-get upgrade does not show me any updates! [08:27] its probably in a ppa MonsieurBon [08:28] I'm using the ppa [08:28] 2.0 beta12 [08:29] deb http://ppa.launchpad.net/juju/stable/ubuntu xenial main [08:29] nope [08:29] devel ppa [08:29] I'll give that a try [08:30] why is stable installing beta packages anyway? [08:30] Yeah, that does seem weird. [08:30] well [08:31] you don't need to install 2.0 you can still install 1.25 [08:32] well this is just testing, so i'll give 2.0 beta 12 a go and will reinstall with 1.25 if it doesn't work [08:33] How do I "rebootstrap"? destroy and recreate? [08:33] erm [08:33] yeah juju kill-controller [08:33] Yes, unfortunately these betas aren't upgradable yet. [08:33] or something [08:33] the names keep changing:) [08:34] juju destroy-controller is the currently blessed command. [08:34] I mean, that's always been the command! ;) [08:34] I'll give destroy-controller a chance before killing it :) [08:34] * magicaltrout clearly just gets bored and goes hardcore [08:35] didn't know remove-controller existed :) [08:35] Yeah, we're trying to get people not to use kill-controller - if destroy-controller isn't working for something that's a bug. [08:35] pfft [08:35] :) [08:35] kill kill kill! [08:36] Well it's nice to at least ask politely first :) [08:36] Apparently it's too easy for kill-controller to leave some instances running in the cloud and people getting billed without realising! [08:38] Ooh, I finally see how you can use chrome dev tools to spy on websocket traffic - that might come in handy if you're still having a problem after bootstrapping on beta12, MonsieurBon. [08:39] of course now destroy-controller does not seem to do anything... :) [08:40] Uh oh. [08:40] So what's happening? [08:40] nothing [08:41] kill did the job :) [08:41] Ah well. [08:41] KILL! [08:41] Were you bootstrapping to lxd? [08:41] Unable to open API: open connection timed out [08:41] nope, an old laptop :) [08:42] But what cloud? [08:42] it's kind of a weird test setup [08:42] Using maas? Or manual provider? [08:42] juju and maas are running on vm's [08:42] maas yes [08:43] and there are two laptops commisioned through maas I'm playing around with [08:43] Right - it's worth making sure that all the nodes are released in the maas ui. [08:43] yes sure [08:43] they were [08:43] Cool cool] [09:14] and it works! [09:14] thx for your help [09:14] MonsieurBon: the later beta? [09:15] or 1.25? [09:15] beta 12 [09:15] cool [09:15] I'll make sure to use the stable once we're deploying maas/juju for production :) [09:15] this is what happens when the hackers keep changing API ! :P [09:15] \o/ [09:16] and it is blazingly fast! [09:17] if you're interested in local testing as well MonsieurBon [09:17] make sure you check out the lxd provider [09:17] +1 [09:18] well I was wondering, can I bootstrap the controller for maas into an lxd container? [09:19] Or will I always "lose" one node to bootstrapping whatever cloud I'm using? [09:22] Well, if you're using lxd you don't actually need maas - you can just deploy to lxd directly. [09:22] Or you can deploy to lxd containers on maas nodes. [09:25] yes, but I would still like to use physical machines to deploy to [09:33] I don't think you can deploy to a container on the node, sorry. [09:34] You can deploy things to the controller machine though, if you want to. [09:35] Sorry, should have said the controller model. [09:35] juju deploy -m controller --to lxd:0 [09:36] The only problem with that is that you can't make relations between applicatons in the default model and others in the controller model. [09:38] Hi I tried uninstalling and re-installing juju , but I am getting version 2.0-beta7 , I need latest 2.0-beta12 to install. for installation I am following https://jujucharms.com/docs/devel/getting-started [09:40] Rajith: you need the devel ppa [09:41] let me know the steps for getting devel ppa [09:43] http://lmgtfy.com/?q=juju+devel+ppa :) [09:43] Rajith: sudo apt-add-repository ppa:juju/devel [09:43] then: sudo apt update; sudo apt install juju-2.0 [09:44] lazy question: anyone have a one liner to shutdown all lxd containers? [09:45] $ sudo halt # ;) [09:45] I feel like this is a trick, but isn't it `sudo lxd shutdown` [09:45] ? [09:45] sorry [09:46] s/shutdown/delete [09:46] Huh, no - they come back up again. [09:46] i have a bunch of stale juju lxd images and I can't be bothered deleting them all one by one [09:46] docker you can pipe a list in [09:46] I was hoping there was a similar lxc trick [09:47] lxc delete can take multiple containers - you can pipe the list into xargs lxc delete, I think. [09:48] yeah I can't figure out how to get just a list of container names [09:48] lxc list gives you all the extra crud === anita is now known as Guest60423 [09:54] magicaltrout: `lxc list --format json | jq -r '.[] | .name'` looks like it will do it. [09:56] ah Odd_Bloke cool! I was playing around trying to find an output that I could manipulate [09:56] you beat me to it [09:56] :) [09:56] * magicaltrout will blog it for posterity [10:02] Odd_Bloke: nice - I normally just use grep and cut, but it's a bit fiddly. [10:03] Yeah, I was using jq for something else last week, so it was fresh in my mind as a possibility. [10:14] Trying to uninstall juju getting error: cannot destroy 'lxd-pool': pool is busy [10:24] kwmonroe: cory_fu FYI I deployed the hadoop processing bundle on LXD earlier and it failed miserably waiting for Java === Guest81296 is now known as spammy [12:08] hi guys [12:08] uh oh [12:08] I know now why juju 2.0 beta4 was installed even though I had the stable ppa added [12:08] because that's the version in the ubuntu repositories! [12:09] for xenial at least [12:09] same for maas: beta2 (or so...) [12:09] I think it's a bit weird to supply beta versions in an LTS server operating system! [12:11] How can I make sure, only the stable version will be installed for juju? [12:11] erm [12:11] you're on Xenial? [12:12] yes [12:12] apt install juju-1.25 [12:12] same applies to maas I guess? [12:13] dunno never used maas [12:13] probably [12:13] hum, there's no maas-version package... [12:14] there is a stable PPA [12:14] oh [12:14] no xenial build [12:15] dunno, you require the support of a canonical person :) [12:15] or from the #maas channel! :) [12:15] babbageclunk: --^ [12:15] the us folk will start appearing shortly MonsieurBon [12:15] hehe [12:15] the traffic will pick up with brainy folk [12:16] rick_h_ will know someone who knows [12:16] * rick_h_ knows nothing! [12:16] is there a problem with juju 1.25 talking to maas 2.0? [12:16] * magicaltrout shrugs [12:16] :) [12:16] 1.25 doesn't talk maas 2.0 [12:16] rick_h_ will know someone who knows ;) [12:16] only juju 2.0 [12:16] You know nothing, rick_h_ [12:17] :) [12:17] rick_h_ knows..... [12:17] and yes, juju 2.0 is in xenial as juju, and a new beta is coming today [12:17] beta12 [12:17] I feel like beta12 is still giving me troubles. [12:17] lets see how to install maas < 2.0 then... [12:17] might have the added benefit of wake-on-lan! :D [12:19] kjackal: you available? [12:24] apparently I need to use 14.04 [12:25] MonsieurBon: what with beta12 is giving you trouble? [12:25] MonsieurBon: yes, to run maas 1.9 but you can run 16.04 on top of that maas 1.9 ok. So the one machine maas runs on needs to be 14.04 if you're using maas 1.9 [12:26] don't know. mysql charm seems to deploy to lxd on machine x but state always shows as unknown [12:27] MonsieurBon: the machine state or the application status? [12:27] the application state [12:27] machine boots fine [12:27] may just be the charm isn't setting anything awesome or useful in there I guess [12:27] nothing else runs on that machine (yet) so no relations. Just the mysql charm [12:29] MonsieurBon: try the mariadb charm as well [12:29] if the mysql is giving you problems [12:30] is that completely interchangeable? [12:30] i.e. it'll talk to the likes of nova-compute? [12:30] ah you doing openstack stuff [12:31] no in that case :) [12:31] ok, I'll try a downgrad then. According to the maas channel I shouldn't use 16.04 until the point 1 release :) [12:31] wimps [12:31] they need to man up [12:32] hehe [12:33] cory_fu: as kjackal isn't talking to me, ping me when you have 2 mins, just got a quick Hadoop Q [12:33] (please) [13:03] magicaltrout: hey [13:03] let me read what's up [13:03] not the backlog [13:03] i've not asked my question yet [13:04] go on [13:05] kjackal: i just wondered, I'm finishing up this PDI charm, and the Hadoop configuration is optional but obviously I'd like it in place for interaction with your bigtop stuff [13:05] if I told it to use the hadoop-client layer stuff, I could get access to the *-site.xml config files it needs I guess, but would that make it a mandatory relation with Hadoop? [13:06] i'm trying to avoid having a hadoop charm and a non hadoop charm version [13:07] magicaltrout: cool! making a relation mandatory or not has to do with the logic you put on the reactive part of your charm [13:08] you can completely ignore a relation by not having any hook acting on joining and/or departing [13:09] there are a couple of techniques that you may find interesting in case of optional relations [13:09] first you can test for a state [13:10] and then you can build the interface from any existing relation [13:10] magicaltrout: give me a moment to find an example === gnuoy` is now known as gnuoy [13:11] well i want to act on the relation, but I don't want the charm to be subordinate to hadoop being available [13:11] basically [13:11] which maybe just how it is, i've not yet tested it [13:12] i'm just making sure PDI can talk to hadoop then i'll dump it all in the charm logic [13:13] ok, let me understand what you have right now [13:13] right now you have PDI that does not relate to hadoop, right? [13:14] it doesn't [13:14] awesome [13:14] but to do so all i need to do is copy in the hadoop configs [13:14] and set a couple of parameters [13:16] ok, cool! So you need to build on this layer: 'layer:hadoop-client' [13:17] okay cool [13:17] you would better set an option silent: True so that the hadoop client does not report "ready" [13:17] i'm sure i can manage that [13:17] have a look at what spark is doing: https://github.com/juju-solutions/bigtop/blob/spark/bigtop-packages/src/charm/spark/layer-spark/layer.yaml [13:18] spark will deploy as standalone and if you relate it to a hadoop-plugin it will be able to use hadoop and hdfs [13:18] ah [13:18] marvelous [13:18] just what i need [13:18] so you add the hadoop-client layer [13:19] then you react to whatever might change your configuration: https://github.com/juju-solutions/bigtop/blob/spark/bigtop-packages/src/charm/spark/layer-spark/reactive/spark.py#L104 [13:21] magicaltrout: the fact that hdfs is available is spotted by https://github.com/juju-solutions/bigtop/blob/spark/bigtop-packages/src/charm/spark/layer-spark/reactive/spark.py#L123 [13:23] magicaltrout: there is also the option to create an interface object from a state: https://github.com/juju-solutions/bigtop/blob/zookeeper/bigtop-packages/src/charm/zookeeper/layer-zookeeper/lib/charms/layer/zookeeper.py#L81 [14:46] Hi marcoceppi: do have any updated on this issue https://github.com/marcoceppi/charm-mysql/issues/4 [14:53] Hello Team, Could someone help me with Juju 2.0 Uninstallation steps? [15:00] juju deploy juju-gui returns cannot retrieve charm network is unreachable [15:00] But I can telnet to the displayed IP address / port [15:00] what might be wrong? [15:00] routing, dns, something else? [15:01] I can resolve api.jujucharms.com [15:01] from the controller? [15:02] this is juju 1.25 now :) [15:02] not from the bootstrap node [15:03] should I try this from the node I'm deploying juju-gui to? [15:03] i.e. the bootstrap node? [15:03] yeah [15:04] how do I connect to the node? [15:04] what ssh user? [15:05] oh, found juju ssh [15:06] nope, no connectivity on that host. I'll go and check the FW === zz_CyberJacob is now known as CyberJacob [15:40] magicaltrout, stupid me: missing default GW in maas interface configuration! [15:40] \o/ [15:42] I should have noticed this before, as bootstrapping had some troubles already! [15:42] MonsieurBon really glad you got it figured out though :) [15:43] he's no MonsieurBon, dont believe a word lazyPower says! [15:43] O_o [15:43] lazyPower, me too :) [15:43] magicaltrout - who gave you access to the peanut gallery today? [15:43] * lazyPower throws peanuts @ magicaltrout [15:43] hehe [15:43] magicaltrout, no I believe it. Because if I hadn't he would have to try to support me. So I'll bet he's glad :) [15:43] i'm allowed, I've been trying to finish my pdi charm all afternoon [15:43] i believe that buys me peanut credits [15:44] magicaltrout - http://traefik.dasapp.co -- i've managed to figure out proper reverse proxying and networking in our k8s bundle(s) [15:44] and you bet, I'll give you 36 hours of peanut gallery access for hte mention of completion of the current iteration of the pdi charm(s) [15:44] in other words, i have no idea what i am talking about [15:44] hehe [15:45] magicaltrout - how have you been addressing networking in mesos? [15:45] i've not (yet) [15:45] does marathon/dcos bring with it, its own flavor of SDN business? [15:45] ah ok [15:45] this wasn't nearly as painful as I was expecting. [15:46] the problem is, you dump stuff into the cluster, and locally it will spin up the stuff I need and create endpoints [15:46] granted its not fully spaces aware so its completely possible to model yourself into a corner as it stands today [15:46] but there is the conundrum of the juju managed firewall [15:46] exactly [15:46] juju run --service foobar "open-port 3600" [15:46] thats schenanigans [15:46] i have spoken with marcoceppi about spiking on a daemon that runs to handle that on behalf of hte user. Enable it via config/etc. [15:46] yeah, but its not nice and doesn't close if you undeployed a serive [15:46] service [15:47] well the idea is the daemon listens on the service level and opens whats defined ther [15:47] that said, dc/os has a command line util I believe you can use to deploy stuff, so if you didn't use the dashboard and instead used a juju action and cli [15:47] that could manipulation the fw# [15:47] or closes respectively [15:47] like every minute, scrape the api, and determine if an action is required [15:47] yeah [15:48] that would be good [15:48] ok i'll draft that up to the mailing list so we can get more feedback on this. [15:48] cool [15:48] thanks for riffing :D [15:48] no worries, i'm working on cleaning stuff up this week when i'm not being annoyed by NASA bods [15:48] so I'm gonna try and get PDI & DC/OS Master scalaing done [15:49] once the master scaling is done, the rest of the basic stuff is pretty straightforward [15:49] just need a few actions to sort out logins and automate the cli installation [15:51] I'm sat in the Snappy Sprint roundup and there is a guy with a far too sculpted beard for a developer [15:51] I'm not sure whats going on in the world any more ..... ;'( [15:52] MonsieurBon do you sport a beard? [15:54] magicaltrout, sort of. you? [15:54] only when I've woken up too late to shave. [15:54] why are you asking? [15:54] beards are acceptable, but if you have time to sculpt your beard you're clearly not overworked enough ;) [15:55] hehe :) [15:55] but wouldn't being shaved hint into the same direction? [15:55] magicaltrout, I think you're onto something there. Beard-groomedness could be a positive indication of employee health [15:56] lol [15:56] maybe [15:56] following the same logic, all developers should grow beards [15:56] mskalka, or that it's time for some downsizing :) [15:57] downsize the beards or the employees? [15:57] when there's too much beard downsizing then there's not enough employee downsizing, I'd say! [15:58] i can see a gap in the silicon value market [15:58] sod onsite nurseries and basketball courts [15:58] get a beard groomer on the google campus and I'm there! [15:58] all I'm getting out of this is to let me beard run wild. That way my employer thinks I'm overworked and thus productive, and I get to keep my beard and my job [15:59] I'm not sure Silicon Valley needs any more ways of signalling to women that they aren't welcome.:p [15:59] women can sport beards if they so choose..... [15:59] women are welcome to sport beards as well [16:00] well you could always include the beard groomer in some general wellness facility that would supply other hairy services [16:00] ... this went in the weeds pretty quickly [16:00] lol [16:00] we're on the cutting edge of HR here lazyPower [16:01] * lazyPower silently closes the window and resumes hacking on a blog post [16:01] * lazyPower shakes head [16:01] lazyPower is jealous [16:01] because he doesn't have a groomed beard [16:01] and I still don't know why exactly magicaltrout asked the initial question :) [16:01] considering how disshelved i look all the time, i think i'm qualified to no-op on this one. [16:01] cause i'm sat in a sprint roundup and someone with a groomed beard stood up [16:02] thats probably stokachu [16:02] o/ [16:02] stokachu: got a beard? [16:04] well, he did anyway :) [16:04] not sure if thats still the case [16:11] I have to say, all the stable version work much better, than the beta ones! :) [16:13] haha [16:19] thank guys for all you help today [16:19] cu later [16:19] no problem [16:19] happy grooming! [16:19] yes, same to you! [16:40] Hi marcoceppi: do have any updated on this issue https://github.com/marcoceppi/charm-mysql/issues/4 [16:41] Prabakaran: no. When I do I will reply to that thread. [17:32] thanks marcoceppi :) i am struct on this point ... please confirm to me once u r done making these changes at your ealiest convenience [17:40] *chuckle* [17:56] Are there any charm helpers that grab all network interface names, and their assigned IP's? [17:58] Or any charm helpers that parse /etc/network/interfaces [17:58] what language are you working in? [17:59] Python [17:59] mskalka: ^ [18:01] x58: https://pythonhosted.org/charmhelpers/api/charmhelpers.contrib.network.html might lead you somewhere [18:01] Perfect. [18:01] I ran into the same issue developing a charm in Rust, not super familiar with the python charm libraries [18:06] mskalka: Looks like it has part of what I need. Will just go steal some code :P [18:07] x58, glad to help. Might also want to look into python's sh package [18:19] kjackal: I think we've had this convo beforee [18:19] does hadoop bundle stuff run ZK or not? [18:19] i vagely remeber some HA thing [18:32] magicaltrout: kjackal is EOD, but the answer to your question depends on which Hadoop bundle you mean. The Bigtop one (https://jujucharms.com/hadoop-processing/) doesn't currently include ZK because it doesn't currently include HA, but kjackal is currently working on adding that back in. The slightly older vanilla Apache bundle (https://jujucharms.com/apache-processing-mapreduce/) does include HA and thus does include ZK [18:33] cory_fu no worries just filling in the blanks in the PDI charm, thanks === wolverin_ is now known as wolverineav [18:39] whats up all? [18:39] has anyone used https://jujucharms.com/u/stub/storage with aws successfully? [18:40] I'm trying to use the storage charm to get my elasticsearch datadir mounted on an ebs vol ... [18:40] it seems the storage charm should be capible of this, but I can't seem to get it to attach a pre-existing vol, or to create and attach a new vol [18:41] anyone use this yet, at all? [18:41] capable* [18:41] can spell [18:43] stub: ping [18:44] bdx: Unfortunately, it's almost 2am stub's time, so you likely won't get a response from him for a few hours. [18:44] Also unfortunately, I haven't used that charm, so I can't really offer any advice. [18:44] lazyPower: I don't suppose you've used it before? [18:44] timezones are for wimps [18:44] * lazyPower reads backscroll [18:45] bdx ah - yeah [18:45] bdx - you need to pair that charm with the block-storage-broker charm [18:45] however, that stack is quite old, and is deprecated in favor of modeled storage with juju [18:45] ooooh [18:45] it was based on the euca2tools python lib, which targets a very old inmplementation of the AWS API [18:45] i see [18:46] its rife with mismatches that may potentially cause headachs, i'm not certain if we're using that anymore. I should really mail the list about it [18:46] lazyPower: But does Juju storage support connecting to an existing EBS volume? [18:46] cory_fu - only ebs volumes it has provisioned [18:46] * lazyPower will craft a quick post to see if anyone is still using it, and what the status/feedback is [18:47] cory_fu, lazyPower: how are you guys dealing with elasticsearch datadir mounts then? [18:47] bdx: Re-reading your ask, I'm not clear if you're trying to connect to a pre-existing, non-Juju provisioned EBS volume or just want it stored on EBS instead of transient storage [18:47] bdx - at the moment, the charm doesn't support storage. So its per-instance. it would be a good idea to patch the charm to work with storage however. [18:47] cory_fu: in all reality, I just need my datadir on a separate vol [18:48] bdx - add to matadata the storage bits, and give it a storage-attached hook to format and mount the volume accordingly [18:48] lazyPower: If Juju provisions the storage mount, and you need to reconnect it for some reason, is that possible? [18:48] cory_fu - i'm pretty sure that is TODO. you can attach it manually [18:48] but the storage feature itself needs modifications to support that [18:49] I've provisioned all my es instance data dirs manually so far ... I'm going to be spinning a bunch of these up and down ... I stumbled across the storage charm and respective hooks in the es charm over the weekend [18:50] lazyPower: That charm was updated by stub not too long ago, so it seems he's still maintaining it. Perhaps it has features that Juju storage don't yet support? [18:50] I was super pumped to spin up my next few elastic clusters using it ... bah [18:50] yeah - it seems fairly recent [18:50] well thats promising [18:50] yeah the storage stuff is pretty cool, we've been doing a bunch of stuff at JPL in docker which I'd like to port to juju [18:50] * lazyPower stops composing mail [18:50] makes sense to keep persistent storage though [18:51] magicaltrout: what do you do for your es data dirs? [18:51] Does config.changed in the reactive always fire after the service is installed? [18:51] magicaltrout: manual prov? [18:51] yeah bdx hacked up mounts [18:51] shnacks [18:51] https://jujucharms.com/u/stub/storage [18:51] looks so promising [18:51] lol [18:52] great idea [18:52] well the native storage stuff in juju makes sense [18:52] depends on charms supporting it though [18:53] magicaltrout: ya - what do you spec your elasticsearch instances at? [18:53]