[00:31] edsiper, sorry, was changing locations there.. is "out_mongodb" the name of your endpoint in your layer's metadata.yaml, under requires: [00:31] ? [00:32] edsiper, for example, if I had this in my layer's metadata.yaml, https://paste.ubuntu.com/15491929/, the state name would be "database.database.available" [00:39] damn it something is wrong in my maas config :-/ [01:10] kwmonroe: I recommend using the leadership settings to distribute the masterhost:port rather than the peer relation, although the peer relation will work too. [01:12] kwmonroe: With the peer relation, after leadership failover you could end up with two units advertising masterhost:port on the peer relation (eg. when the lead unit has been powered off but the unit not removed - it still exists on the peer relation) [01:12] kwmonroe: With leadership settings, there is a single source of truth [01:43] thx stub! that's good info. [02:20] almost got it bootstrapped, getting stuck downloading the tools :-/ === natefinch-afk is now known as natefinch [02:39] am I missing something or is curl -sSfw 'tools from %{url_effective} downloaded: HTTP %{http_code}; time %{time_total}s; size %{size_download} bytes; speed %{speed_download} bytes/s ' --retry 10 -o $bin/tools.tar.gz <[https://streams.canonical.com/juju/tools/agent/2.0-beta2/juju-2.0-beta2-trusty-amd64.tgz]> not a valid curl command? [02:45] cmars, yes, it's out_mongodb [02:54] hmmm gotten hung up twice at that point :-/ [03:05] edsiper, back from dinner. what does the function that you're decorating with that @when look like? does it take just one argument? [03:06] it probably does.. in which case, i'm running out of ideas.. other than maybe checking that the latest built charm is actually what you've got deployed and are relating to [03:22] cmars, yeah, one argument and basically doing a log() [03:22] hmm [03:26] cmars, http://pastebin.com/H1eUhge7 [03:29] edsiper, can you paste me your layer.yaml as well? [03:29] ok this is messed up, how can apt-get update work but juju deploy --to 0 juju-gui cant [03:36] cmars, ooops, no layer.yaml [03:37] edsiper, aha [03:37] cmars, I will dig into it tomorrow :D , thanks for pointing the main problem [03:37] edsiper, add the interface, charm build, and you'll probably see that get called [03:37] g'night... [03:37] edsiper, night [03:37] yeah [03:50] got love running into an ipv6 vs ipv4 confusion [03:54] some days I just wonder how I go to where I am when I miss the basics [04:17] oh I do like the default juju status for 2 :) [05:55] OK so I just did juju deploy juju-gui and juju expose juju-gui, juju status shows that it has the right public address and yet I keep getting connection timed out errors [09:57] Hi, i would like to write a docker charm utilising the layer-docker charm but it is not clear to me what's the best way to do this. Do i simply clone layer-docker or can i use something from charm-tools to pass this as an argument? === scuttlemonkey is now known as scuttle|afk [11:29] deanman: You declare the docker layer in your layer.yaml. You probably want to go over https://jujucharms.com/docs/devel/developer-getting-started and https://jujucharms.com/docs/devel/developer-layer-example (the devel branch of the docs now documents layers and how to use them) [11:30] stub: thank you! [13:18] hi I am getting this issue while deploying the latest ceph charm https://bugs.launchpad.net/opnfv/+bug/1561984 [13:18] Bug #1561984: ceph-osd charm failed to deploy [13:20] can someone help me finding the solution. it seems latest changes in ceph-osd charm http://bazaar.launchpad.net/~openstack-charmers/charms/trusty/ceph-osd/next/revision/76 is causing this issue? [13:53] marcoceppi: did you forget to push your fixes to fiche? Still shows up busted for me on jujucharms.com [13:53] well not busted, just with the extra stuff [14:03] jcastro: it's pushed. [14:03] store ingestion, probably [14:04] jcastro: http://bazaar.launchpad.net/~charmers/charms/trusty/fiche/trunk/revision/2 [14:13] jrwren: which one of you guys handles ingestion? === cos1 is now known as c0s [14:37] jcastro: kind of all of us? [14:37] can you check fiche? === scuttle|afk is now known as scuttlemonkey [14:49] jcastro: looks like its been pushed to before, so ingestion is disabled for that charm. [14:58] jrwren: what does that mean? [14:58] why would ingestion be disabled? [15:00] jcastro: because someone uploaded it using charm push? [15:00] oh, so how do we fix it to make it ingestible? [15:02] jrwren: do we need to rerun a publish or something? [15:03] jcastro: You cannot. Once a charm has been published using the new charm push command, that is how it is now managed. This fits with ingestion going away. [15:04] ok so how do we fix it so that charm is ingested? [15:04] tldr, how do we make the jujucharms.com page reflect what's actually in the vcs branch? [15:05] jcastro: AFAIK we have no means to do that. You can fix the charm in the store by pushing the new version using the charm push command. [15:05] marcoceppi: is that what you tried? [15:10] jcastro: what is wrong with it? https://jujucharms.com/fiche/ is fiche-3 and doesn't show the extra files that were removed in bzr rev-2 [15:11] huh [15:11] that directory was there an hour ago [15:11] but marco pushed that on 21 march, maybe I had a stale cache? [15:12] jrwren: ok I guess I'm all set lol, thanks. [15:12] jcastro: Sorry for the confusion. [15:13] jrwren: actually, I just realized how I did that [15:13] google for "fiche jujucharms" [15:13] the result sends me _specifically_ to revision 2 instead of trunk [15:16] msust be your tuned google. oh... 5th result for me. [15:16] I rarely look down to 5th ;] [15:17] is it rev 2 or 3? [15:17] there's a bug on this already, we need to present google with the right bits [15:17] I just should have been paying attention to the rev # [15:17] latest is 3. [15:18] yeah google shows me 2 [15:18] I'd grumble about the link that says revisions: 1 on that page, but I'm told that UI is by design, so I'm not going to grumble about it and instead suggest you not be mislead by it. [16:01] are there any known issues with bootstrapping Xenial on juju2/maas 1.9? [16:02] Maas can deploy xenial just fine by itself, but when I try to bootstrap on xenial, it fails - maas shows the install completes, but juju complains that "instance started but did not change to Deployed state" [16:02] juju bootstrap xenial dr --config default-series=xenial --debug [16:03] I can bootstrap on 14.04 just fine [16:10] does anyone have a bash prompt that shows the current juju env [16:10] it could be handy on one of the machines I work on [16:25] cory_fu might, i know he's a bit of a wizard at that stuff with liquid prompt === natefinch is now known as natefinch-afk === natefinch-afk is now known as natefinch [18:22] hi LiftedKilt, i would suspect a bootstrap timeout. ie. it may be taking slightly longer than the 10m default. see `juju help bootstrap` to learn how to increase the time that juju will wait for the bootstrap machine [18:26] beisner - the actual OS installation only takes a couple minutes, and I can see the output on MAAS that the Xenial install was successful. Are you thinking it's the tools installation that might be holding it up? [18:27] LiftedKilt, it may be a separate issue, but i also had the same "instance started but did not change to Deployed state" with xenial as it was overrunning the 10m default timeout a bit. worth noting, the machines i'm talking about are a bit slow, and take several minutes just for POST. [18:28] beisner: I appreciate the tip - I increased the timeout to 20 min and am redeploying. Fingers crossed that it works! I successfully bootstrapped Xenial a couple days ago, I just haven't been able to do it since. Been driving me nuts. [18:37] Hi [18:39] beisner: good looking out - it deployed just fine. Thanks! [18:39] LiftedKilt, \o/ awesome, happy to help. [18:45] Office Hours in about 20 mins: https://plus.google.com/hangouts/_/hoaevent/AP36tYfv6UdYcQGhr8qG33I1H0mmMK-2c5vFDcgXtVZrCEa-Tk0WuQ?hl=en - to join, otherwise tune into https://ubuntuonair.com/ [18:51] kwmonroe: cory_fu: you guys doing office hours? [18:51] someone from juju core would be great! [18:51] cherylj: how about you? [18:53] jcastro: what do I need to do? [18:53] * cherylj has nothing prepared [18:53] oh snap, we should talk about the new "admin" and "default" model stuff coming out in beta3 [18:54] jcastro: and when is it? [18:54] just come to the hangout [18:54] jcastro: linky me please and I'll join [18:54] it's in 5 minu8tes [18:54] the idea is not to be prepared, just talk [18:54] it's like a podcast [18:54] cherylj: yea just party for a bit [18:54] Office Hours in about 5 mins: https://plus.google.com/hangouts/_/hoaevent/AP36tYfv6UdYcQGhr8qG33I1H0mmMK-2c5vFDcgXtVZrCEa-Tk0WuQ?hl=en - to join, otherwise tune into https://ubuntuonair.com/ [19:00] hoa, I initially read that as home owners assocation :-) === cos1 is now known as c0s [19:09] office-hours: http://paste.ubuntu.com/15500937/ [19:09] kwmonroe: I don't if you guys have seen this https://d0.awsstatic.com/whitepapers/lambda-architecure-on-for-batch-aws.pdf - looks like a decent and clean explanation of the lambda-arch [19:12] c0s: looks like an interesting read, thanks [19:13] np [19:16] can we attach cinder vols to nova-lxd instances? [19:16] sorry [19:17] rick_h: good to hear! thanks! [19:18] marcoceppi: awesome! [19:21] bdx, cinder + nova-lxd is not supported yet [19:21] cargonza: thanks, what is the status of that? [19:22] still working the priorities and plan. we're finishing up the target goals for 16.04. We'll consider it for future planning. [19:23] cargonza: awesome! thanks! [19:23] marcoceppi: great, will do! [19:26] LiftedKilt: heya, have you seen the layered charm bits? [19:26] jcastro: I haven't [19:26] https://jujucharms.com/docs/devel/developer-layers [19:26] so basically [19:26] I've heard it mentioned a few times, but haven't seen it yet [19:26] in the past, we made you do things the hard way [19:26] which is horrible [19:26] now you compose a charm out of layers [19:26] jcastro: awesome - looking at link now. Thanks [19:27] http://interfaces.juju.solutions/ [19:27] also a good link is https://d0.awsstatic.com/whitepapers/lambda-architecure-on-for-batch-aws.pdf [19:27] buffer! [19:27] so like, you could reuse a java layer the big data team uses, for example [19:27] * arosales meant https://jujucharms.com/docs/devel/developer-getting-started [19:27] before everyone had to do things individually [19:27] we should like /developer-getting-started to developer-layers [19:27] jcastro: this will be awesome for creating the liferay charm, for example [19:28] s/like/link [19:28] yes [19:28] I can just hand this to our devs [19:28] LiftedKilt: in fact, just pretend the old way doesn't exist [19:28] jcastro: gladly haha [19:28] so basically all the old charms weren't as testable or reusable, so like say we find a bug in one the java layers [19:29] you'd just pick up the bug fix [19:29] instead of "hey everyone for every charm using java, please fix foo." [19:29] oh that's clean [19:33] charm push looks awesome [19:38] * lazyPower raises the roof at marco's ACLs [19:38] i totally did this [19:38] haha [19:42] arosales: I think if you replace Kinesis in that paper with Kafka - you'll get pretty much the same result ;) [19:42] but not AWS-locked [19:42] lazyPower: do you know if on "push" if one doesn't specify the series does it error if the charm doesn't include a series metadata? [19:43] i do not know, i haven't tried [19:43] c0s: ok [19:47] arosales once i'm done restoring my phone and get my MFA restored, i'll let you know :P [19:47] lazyPower: no rush, just wondering [19:52] aweee [19:52] oh no [19:52] stream cut out [19:53] :-) [19:53] hit wrong button [19:53] haha, I pressed "stop stream" instead of "stop sharing" [19:54] sorry everyone [19:54] we were about to wrap up, it'll be ok [19:54] we'll just give marcoceppi a hard time for a while :P [19:54] <3 [20:06] lazyPower: rick_h_ was saying that if you try to push a charm and it has no series in the metadata you should get an error back [20:18] ah [20:18] I've only pushed to implicit series locations in the store [20:18] so, good question and good follow up :D [20:40] if we were to deploy openstack with juju and use lxd for nova, would we have no persistent storage? Is that what cargonza meant when he said cinder + nova-lxd is not supported? [20:55] falanx: My understanding is that it is being considered as a future feature ... [20:55] cargonza:^ [21:03] falanx: see cargonza's message above [21:17] falanx, the openstack lxd charm deploys ceph for storage. cinder is another type of storage that we still need to support in nova-lxd [21:36] cargonza: so by cinder you mean the reference architecture cinder using lvm, right? [21:37] cargonza: the openstack lxd charm integrates with ceph allowing containers to mount ceph volumes? [21:41] LiftedKilt, yes on both [21:43] LiftedKilt: part of the purpose of cinder is to abstract you from the backend storage implementation [21:44] cargonza: awesome [21:44] bdx: my understanding is that cinder is both the reference lvm implementation as well as the name for the general block storage api integration for openstack. [21:45] LiftedKilt: It doesn't matter to the openstack instance what the underlying storage may be ...as it is just consuming a cinder storage resource ... it doesn't matter what backend you use [21:45] or it shouldn't at least [21:46] thats the whole point right? [21:49] indeed [23:08] ping admcleod- coreycb kwmonroe kjackal [23:09] pong A-Kaser [23:09] A-Kaser: i'm about 200% sure you meant cory_fu instead of coreycb ;) [23:09] right :) [23:09] bad tab [23:09] np, what's up? [23:10] I'm trying to set my credentials to test on aws [23:11] I would like to use eu-central-1 or eu-west-1 [23:11] with devel documentation example it seems to be ignored my default-region [23:12] and the link to have more information credentials is https://jujucharms.com/docs/devel/credentials/ 404 [23:12] could you help this little point :) [23:13] A-Kaser: i noticed that too with an azure deployment, but assumed i did something wonky.. i'll check again and open a bug (unless rick_h_ knows whether or not default_region is not always honored off hand). anyway, you can force the bootstrap to use a region like this: [23:13] (/me looks for a minute) [23:13] and just in case, I'm the dude which sent you an email right now :) [23:13] no pb :) [23:15] A-Kaser: i think this will work for ya: juju set-default-region aws eu-central-1 (or west) [23:16] A-Kaser: then proceed with your bootstrap. alternatively, you can set the region during bootstrap with " juju bootstrap [/region]" [23:17] so like, juju bootstrap kevin-is-amazing aws/eu-central-1 [23:17] or whatever... [23:17] that would be a lie though [23:17] dag nab it magicaltrout - i thought you'd be in bed by now. [23:17] I did "juju set-defatul-region" and that worked for me [23:17] hehe [23:18] arosales: s/defatul/default, but cool. thx for the confirmation. [23:18] juju like to talk : 20 x WARNING could not delete security group "juju-4520f63e-51f6-43bd-8b89-336a0190eeef-0": resource sg-15670d6d has a dependent object (DependencyViolation) [23:19] kwmonroe: yes what you said [23:19] :) [23:19] kwmonroe: default-region is ok now [23:19] thx [23:19] cool, np [23:20] A-Kaser: can you tell me what page linked you to your /credentials 404? we need to fix that up. [23:22] https://jujucharms.com/docs/devel/getting-started [23:22] link just before chapter 4. Bootstrap [23:23] -> please see this guide to credentials [23:23] ahhh, yup A-Kaser, thanks [23:26] I suppose it's possible to have more than one controler, so how can I specify which controler used with my deploy command ? [23:27] (i'm trying to read "juju help deploy") [23:27] A-Kaser: firstly, we'll get the docs fixed up with this https://github.com/CanonicalLtd/jujucharms.com/issues/235, thanks for the report [23:28] A-Kaser: secondly, you choose your controller with the "juju switch" command [23:29] so if you have more than one, for example, an aws controller and a gce controller and an azure controller, you would "juju switch gce" to make that the controller that fulfills deployment requests. [23:30] oh okay, so it's not recommanded to have 2 terminal using juju on the same time with different controller I suppose [23:36] yeah A-Kaser, i don't recommend that. you could probably do it if you were vigilient in "juju switch " before each juju interaction, but i'm not that multi-taskable. [23:36] FYI on AWS deploy wordpress don't open 80 in security group, so it's not possible to open it in a browser [23:37] kwmonroe: ok no problem I will have only one instance :) [23:37] A-Kaser: i do, however, have long-running deployments and controller, so for example, once a month, i switch to my openstack controller, juju ssh over to do something, but then switch back. it's just not something i would do very often ;) [23:39] yes, I thought it was possible to add something like "-controler=master1" in deploy command; but switch will be fine [23:39] A-Kaser: can you "juju expose wordpress"? [23:39] i'm not certain about the correlation between juju expose and aws security groups, but if you expose wordpress, i'm curious if that lets you get to it in a browser. [23:42] I have juju 1.25 and 2.0 ... my fault, i'm removing 1.25 [23:45] A-Kaser: i think they can co-exist pretty well.. just need to run 'update-alternatives --config=juju" to set whichever one you want. [23:45] or removing 1.25 and going all-in with 2.0 is cool too :) [23:46] or you could pull a magicaltrout and run from master every 2 hours. [23:48] kwmonroe: I was looking at some of the charms description like flume-kafka and whatever.... And something caught my attention [23:48] namely: the section on smoke testing of the deployed bundle [23:49] like every single bundle has it. Is there a concept of auto-smoking of a unit at the end of the deployment? [23:51] c0s: kinda -- each charm in a bundle gets all the ./tests/* run when it is bundletested, so in that regard, each unit is smoke tested [23:51] one sec, i'll find an ex [23:52] I know what you're talking about [23:52] I have a bit different angle: instead of asking a user to smoke test a hadoop cluster, something simple and automatic could be done by Juju itself. [23:53] like explained here [23:53] hdfs dfs -chmod -R 777 /tmp/hdfs-test [23:53] hdfs dfs -ls /tmp # verify the newly created hdfs-test subdirectory exists [23:53] hdfs dfs -rm -R /tmp/hdfs-test [23:53] hdfs dfs -ls /tmp # verify the hdfs-test subdirectory has been removed [23:53] and then for YARN, and so on [23:53] I think user shouldn't even see this... like ever [23:56] does it make sense, kwmonroe? [23:57] yeah, i hear ya c0s [23:58] i guess the point of that "hdfs dfs -xxx" in the readme is only there to tell the user how to manually smoke the deployment in case they didn't use a promulgated version