=== ABC is now known as Guest50105 [03:52] I have an input.yaml file like this: [03:52] http://pastebin.ubuntu.com/23346821/ [03:52] When I use the above yaml file in 'juju deploy' I get the following error [03:52] ubuntu@juju-api-client:~$ juju deploy --series trusty --to Compute-Storage-3.maas --config input.yaml ./contrail-tsn Deploying charm "local:trusty/contrail-tsn-35". ERROR unknown option "install-sources" [03:53] What is wrong with yaml format? [04:16] The above mentioned yaml passes the online yaml parsing [04:17] The above mentioned yaml passes the online yaml parser [04:46] Never mind. I fixed the issue [06:53] Good morninf Juju world! [06:53] *morning [08:41] Hello, using xenial with juju 2.0 and juju bootstrap command halts without completing. I'm running behind a proxy. Is there a way to setup proxy for consecutive LXD containers before running bootstrap command ? === jamespag` is now known as jamespage [09:58] Hi.. [09:58] I have an issue while deploying a service uisng MAAS provider. Anyone here who can hep me with this ? [10:12] if a service goes down, it's the job of update-status hook to identify and report it if it affects the unit, right? does it do that using status_set or returning something? [10:27] and what would be the appropriate status to set? the docs seem to imply that error state is only for hooks failing [11:10] cargill: yes, update-status is a good hook to report service failures. Seting the status is the right way to go. Of course the update status hook could try to restart the service. The appropriate status is to say "blocked" in case user action is required or "waiting" in case there is an automated recovery on the way [11:19] hi, is there documentation, besides the charm descriptions, on how to deploy openstack with neutron-gateway external network, and tunneling data network between neutron and compute nodes? [11:21] with OVS [11:26] i've read that i need os-data-network [11:26] but how does it corelate with bridge-mappings and data-port? [12:00] MrDan: You might get some more action in #openstack-charms [12:00] ah, ok, did not know about that channel, thanks [12:01] MrDan: The guys that developed the charms are there :) [12:51] can I combine the @hook and @when decorators? [12:52] seems quite useful for the update-status hook [12:57] cargill - its not obvious why you cant, but mixing @hook and @when will yield bad behavior. We dont currently list that as a supported way to use the reactive decorators [12:57] you can use either/or, but not both on the same method. [12:57] cargill if you have need of introspecting state in that method body, i can recommend decorating with the @hook decorator, and you can import is_state and use that to check in the method body if a state is set. [12:57] yeah, that's what I see, nothing really happening, but I think that just guarding on the desired state will do the right thing [12:58] I'm not sure the charms even see hooks anymore as such [12:58] or just decorate for the state, and handle the cases where its being called. [12:58] right, thats been the authors intent with reactive is to fully abstract teh hook mechanism and urge developers to only use reactive states. [12:59] yeah, is what I ended up doing (well, waiting for the test to finish) [13:02] seems super useful to think of the reactive states in terms of a petri net, makes it quite simple to alter the design (add new states or separate processes/relations) [13:05] kjackal_: did you recommend to me yesterday juju-relation-mysql? [13:05] cause i've got problem with it [13:06] Spaulding - yep, that was mentioned by magicaltrout. What can i do to help you? [13:06] Spaulding: I do not think it was me but what is the problem, I might be able to help [13:06] oh yes, sorry - my bad... to many nicknames to remember! [13:07] One sec, I'm creating gist :) [13:08] https://gist.github.com/pananormalny/58ee2c54acf127dc279bf34990b94238 [13:08] After I deployed my charm - I've " juju add-relation mysql sme" [13:09] In debug-log I can see that hook is executed [13:09] your includes is incorrect [13:09] interface:mysql [13:10] So I need to add mysql interface as well or I should change last line? [13:13] i don't define it in yaml [13:13] but in mine [13:13] i put [13:13] includes: ['layer:basic', 'interface:mysql', 'interface:pgsql'] [13:13] so i assume yours should say interface:mysql [13:13] s/yaml/yaml list [13:14] ok... so how to use juju-relation-mysql? [13:15] I think I should provide it as well? [13:15] But I'm not sure about that [13:17] i have no idea what juju-relation-mysql is [13:17] so i think you've made it up :) [13:17] oh its just the name of the github repository [13:17] * lazyPower smirks [13:17] you are allowed to ignore that [13:18] https://github.com/johnsca/juju-relation-mysql [13:19] magicaltrout you dirty devil! [13:19] * Spaulding hides [13:20] https://github.com/johnsca/juju-relation-mysql/blob/master/interface.yaml#L1 [13:20] thats the bit you're interested in Spaulding [13:20] exactly [13:20] you proposed that yesterday [13:20] Spaulding - i think he's calling out that in line25 of your gist, you're including 'interface:juju-relation-mysql' [13:20] which will never resolve [13:21] it should instead be 'interface:mysql' [13:22] surprisingly it's resolved to mysql [13:22] build: Processing interface: mysql [13:22] build: Processing interface: mysql [13:22] its like magic [13:23] magicaltrout - did you see the recent announcement that elastic is teaming up with prelert for machine learning against operational data? [13:25] https://www.elastic.co/blog/welcome-prelert-to-the-elastic-team -- for context [13:26] very interesting [13:27] this is what merlijin was talking about a while ago right? [13:27] this is the future for log analytics stuff, where you have beats, splunk, graylog whatever [13:27] just pumping logs through an indexer is okay, but where is the added value? [13:27] it needs so magical powers to enhance the logging to do more and tell you stuff you didnt' already know [13:28] doing deep machine-learning analysis on log data to identify system issues and auto-diagnose problems... seems like the first step to yet another layer of auto-healing [13:28] yup [13:28] s/auto/self/ [13:28] kwmonroe cory_fu - that article above is probably of interest to you two as well. [13:29] the only articles kwmonroe is interested in are swimming pool enhancements [13:30] well, i used to install those too. if he wants a discount on hayward/jandy hardware i can probably put him in touch with the right people [13:31] http://www.choicepoolsandhardscapes.com/ -- some of my previous work (both the site and the construction projects used as illustration) [13:31] lazyPower: sounds awesome. with enough security data it would make a sweet platform for a SIEM or ESM system. [13:31] i dont know what those acronyms are, but they sound very official [13:32] Security information and event management / Enterprise Security monitoring. [13:32] oh right, i knew that [13:33] indeed jrwren, indeed [13:33] sorry lazyPower, i previously worked in security. [13:33] lazyPower previously worked in swimming pools [13:33] and very nice they look too! [13:33] ^ [13:34] thanks :) [13:34] i'm happy i'm done with what we dubbed as hebrew labor, carting stones across someones bakc yard to build those retaining walls. [13:34] not sure if thats offensive, so if it is, i apologize in advance [13:35] probably :P [13:35] i dug out my telescope foundations by hand which were a 3m cubic hole, that was the hardest thing I've ever one [13:49] yeah, thats hard work man. anybody that said ditch digging is easy is either a liar or has a bobcat [13:58] * magicaltrout on NASA call "we've written our UI and middleware in Python... do you mind writing the backend in python"..... "sure.... not a problem" [13:58] * magicaltrout furiously googles python stuff [14:00] magicaltrout welcome to the painnnnnnnn ;) i kid i kid, python isn't so bad [14:01] na [14:01] this flask stuff looks pretty straightforward [14:01] flask is very simple, and very extensible [14:01] have to write a dockerfile generator for machine learning toolkits [14:02] yeah, combine that with some jinja templating logic and you've got yourself a web service to spit out dockerfiles [14:32] I need advice to know if it is illadvised to try and make a charm that configures lxd for me [14:33] and also, how the heck I'd do it [14:34] well the lovely lazyPower maintains layer-docker i believe [14:34] I've got lxd configured by hand on my laptop and I have some scripts that make containers for me and run through tests to check if my code it good to merge [14:34] so I don't see why you wouldn't have layer-lxd [14:35] and a coworker has a charm that they use for checking out repos and running scripts that happent o use old lxc commands [14:35] thats tricky, its fine if the charm is going to maintain its own lxd container and its setup, but wholesale reconfiguring the lxd daemon would likely disrupt the native juju lxd components [14:36] this is the charm that works with all the old lxc stuff https://github.com/checkbox/pmr-configs [14:37] my other coworker just uses travis, but his repo is on github [14:37] mine's on lp [14:38] the lp teammates have that charm to set up a little ci environment for us [14:40] skay - i think its fine to manage a set of lxd containers on a host with a charm [14:41] skay i just wouldn't wholesale reconfigure the lxd service on the unit if you ever plan on using juju's native lxd bits on that unit, eg: --to lxd:# [14:41] marcoceppi or others may disagree. but i'm in favor of not limiting what you can do, just so long as you're aware of the potential caveats [14:44] lazyPower: magicaltrout: thank's y'all. [14:44] whoops that ' cloned its self and hoped back 2 [14:44] needs more coffee [15:06] is there a way to remove charms where the agent is “lost”? remove-unit and remove-machine are not doing the job. i do want to remove it, i’ve already replaced the “lost” unit with a new one [15:07] hml: don't suppose you tried remove-machine with --force? [15:08] ryebot: trying now, i missed the —force option [15:09] ryebot: that did it - thanks. [15:09] could someone look at it and tell me why it stuck on "Please add relation" [15:09] https://gist.github.com/pananormalny/aa3be95bfb285ecaa2d0b3e2486101e4 [15:09] hml: awesome! np [15:09] Even when I've added a relation... [15:09] I followed the layer-vanilla... [15:11] jose: ping [15:19] marcoceppi: pretty sure your response to my email distills down to "don't use colocation or manual provider" :) Which, btw, I agree with :) [15:19] natefinch: yeah, but we're forced to do these things because not everything works in lxd machines [15:20] marcoceppi: yeah, I really wish we could just default to tossing everything in a contain [15:20] lazyPower mbruzek I wonder if we can explore using --to kvm: for some of these workloads? [15:20] er [15:20] natefinch: ultimately, that would be the best thing [15:20] marcoceppi - we're going to have the same networking issues. [15:20] lazyPower: ack, ta [15:20] which is why we've modeled our dense deployment as is today [15:21] marcoceppi: btw, on the stop hooks, the docs are pretty clear. They might possibly be wrong, but they're clear at least: https://jujucharms.com/docs/2.0/authors-charm-hooks#stop [15:21] natefinch: I just looked them up, that's not what they were always set to [15:21] maybe this is just a failure to name things appropriately [15:22] but stop =/= uninstall to me [15:22] i wrote that docpage [15:22] if its wrong, its my fault [15:22] marcoceppi: certainly. start and stop vs. install and uninstall [15:22] I guess we have an install hook [15:22] natefinch: I also think stop hook is run on all containers where the parent machine is about to be rebooted because of `juju-reboot` [15:23] natefinch: I will experiment [15:23] marcoceppi: would be good to know. I'd be surprised if it ran on reboot, but only slightly surprised :) [15:43] marcoceppi: could you help me with interface:mysql? I can't get it working... :( [15:43] Spaulding: absolutley, what's not working? [15:44] it stuck on "add relation" [15:44] but relation is there... and i can see creds in debug-log as well [15:44] https://gist.github.com/pananormalny/aa3be95bfb285ecaa2d0b3e2486101e4 my reactive [15:44] database stuff is at the bottom of file [15:45] Spaulding: what does juju status say atm? [15:46] Spaulding: and can I see your metadata.yaml? [15:46] blocked "Please add relation to mysql" [15:46] sure [15:46] here you have: https://gist.github.com/pananormalny/68680a0a104c881506ee0f6881103ca8 [15:47] Spaulding: so, this is just a mismatch in naming [15:47] Spaulding: you named the relations "db" but you refer to it as "database." in reactive. Either change metadata.yaml to be database or change your decorators to be db.connected and db.available [15:47] ok... so this db should be mysql? [15:47] yeah [15:47] now i got it! [15:48] Spaulding: yeah, thta's probably not highlighted very well in the readme, I'll make sure we update that [15:48] Spaulding: everything else looks great! [15:48] and i spent like 3 hrs [15:48] changing when and when_not [15:49] thx again marcoceppi ! [15:49] Spaulding: sorry :( hopefully this gets you sorted [15:49] yeah, now I can proceed [15:50] I don't have much time today but tomorrow I hope I'll be able to restore sql dump finish this charm finally [15:50] and finish*** [15:52] Spaulding: awesome, feel free to ping here if you have any more questions! [15:53] will do - for sure! === arturt___ is now known as arturt [16:32] bdx: yes? [16:47] jose: https://github.com/jamesbeedy/charm-mailman [16:47] bdx: sorry, I don't work with layers, but if you want to push that, feel free to [16:48] jose: oh, I thought you wrote the OG mailman charm [16:48] ? [16:48] yes, but again, I don't work with layers, if you want to push a trusty/xenial version for it, feel free to [16:48] I probably won't get around to it until a month from now at least [16:49] maybe even two, considering my schedule [16:49] no worries, just thought you would like to have a looksie [16:49] I'm not pressing anything lol .... just thought I would share w/you :-) [16:49] cool [16:55] Hi all, a question about running Juju (2.0) on GCE: is it possible to configure Juju so that it does _not_ allocate a GCE External IP for each machine? (I also asked this yesterday morning, but hopefully there are different folks around now...) [16:55] dimitern: any idea? ^ [16:56] neiljerram: I don't think so, but let me double check.. [16:57] hey guys, are the bootstrap configs/constraints enumerated anywhere in the docs .... can't seem to find anything really ... [16:57] rick_h_, dimitern - thanks! [16:57] bdx: not at the moment. We need to build one. Most of it is there if you dump the list of controller config. [16:58] rick_h_: thanks [16:58] bdx: or honestly I poke at code most of the time. We've got a todo to dump things out into an official doc list [17:00] rick_h_: I remember hearing about a new ssl config that can be specified at bootstrap to allow proxying to the gui ..... I'm about to bootstrap for the long haul ... trying to make sure I have all of the relevant configs in there [17:00] neiljerram: no way to disable this at the moment, sorry :/ [17:01] rick_h_: from the user perspective, its a bummer I have to go searching to this extent to find the pieces I need to get up and going [17:01] dimitern, Thanks for checking that. Would it be of useful or of interest for me to register a wish for that somewhere? [17:02] neiljerram: of course! :) I see a bunch of TODOs in the source about "let's make this configurable", so it was always planned to be possible, just not yet [17:04] dimitern, Do you mean in network.go after "Type: NetworkAccessOneToOneNAT"? I was just looking there too :-) [17:04] neiljerram: yeah, that's just one of many apparently :) [17:05] dimitern, So where is the right place to request this? [17:14] neiljerram: I'd suggest filing a bug against lp:juju https://bugs.launchpad.net/juju/+filebug [17:14] neiljerram: that way we can schedule it and track it better [17:14] dimitern, Thanks, I'll do that. [17:15] neiljerram: awesome! thank you ;) [17:16] when using the openstack provider, how do I tell juju at bootstrap time which neutron network to use for the instance, if there are multiple? [17:16] error info: {"conflictingRequest": {"message": "Multiple possible networks found, use a Network ID to be more specific.", "code": 409 [17:16] what's the parameter name for --config? [17:16] ahasenack: --config network=xxx.xxx.xxx.xxx [17:19] bdx: thx [17:41] ahasenack: np, too bad its not an available bootstrap config for aws ;-( [18:01] hi, in juju-deployer, I used to use "overrides" for config variables that existed in multiple services (now applications) -- I haven't found such an option on juju2 bundles -- would I need to repeat the same key-value on all charms that have it as config param? [18:02] that's what i've done aluria [18:02] afaik that's the only way [18:02] jhobbs: ack, asking just in case -- thanks === scuttle` is now known as scuttle|afk [18:30] hey guys, I'm getting this error intermitently when sending off deploys -> http://paste.ubuntu.com/23349938/ [18:31] not sure if its been noted yet [18:34] rick_h_: ^^? [18:34] bdx: looking [18:35] bdx: hmm, looks like a header auth issue. [18:35] bdx: need to file a bug on that to get the charmstore folks looking at it please. github.com/canonicalltd/jujucharmscom [18:35] bdx: quick fix is probably to logout/back in and maybe to remove cookies. [18:35] jrwren: where are the go cookies these days? [18:35] jrwren: it's not .go-cookies any more right? [18:36] rick_h_: it is still that. [18:36] jrwren: ah ok [18:36] bdx: clear ~/.go-cookies [18:37] rick_h_: -> http://paste.ubuntu.com/23349960/ [18:38] jwren: ^ [18:38] bdx: juju login? [18:38] oooh [18:40] rick_h_: wow, .local/share/juju/accounts.yaml doesn't have my most current controller in there ... I may as well go back to sleep today [18:41] now that I've logged out [18:41] I have no record of the password [18:41] foolish [18:41] oh well [18:41] redo [18:42] bdx: ?! [18:42] rick_h_: http://paste.ubuntu.com/23349979/ [18:42] well that seems scary if you can dupe that I'd be curious what happened to lose a controller from there w/o an unregister command. [18:43] rick_h_: neither of those accounts/controllers exist anymore [18:43] rick_h_: and my recent bootstrap to aws isn't listed [18:43] bdx: oh, so those were there and have been torn down or the like? that's what unregister is for cleaning up [18:44] rick_h_: yeah, but accounts.yaml doesn't show my aws controller [18:44] the one I just logged out of trying to get charmstore to work [18:44] ahhh ..... logout removes the entry from accounts.yaml [18:44] I didn't previously know that [18:44] bdx: hmmm, but it's still in the controllers list? [18:45] bdx: e.g. you can juju controllers and see ones you're not logged in with? [18:45] rick_h_: yes -> http://paste.ubuntu.com/23349994/ [18:46] bdx: ah ok, that makes me feel less freaked out [18:46] but since I neglected to grab my account details before I logged out .... I'm out of luck [18:46] not that I had anything deployed yet [18:46] but good thing to note for the future [18:46] bdx: :/ yea that sucks. /me needs to figure out how we should do that better... [18:49] rick_h_: possibly, if an admin logs out of the controller we should add a message that alerts them that they will not be able to log back in unless they have their password saved somewhere outside of Juju [18:49] bdx: yea, and seems like a documented way for admins to get their password from the controller maybe [18:49] not sure if it's available over there, if you can ssh there can you get the password out. [18:50] rick_h_: I'm in [18:51] bdx: can deploy? [18:51] rick_h_: no, I was able to ssh into the controller node [18:51] bdx: ah ok [18:51] bdx: yea, I have no idea if the password is retrievable [18:51] perrito666: any ideas? ^ [18:52] ooho, I thought you were saying "if I could get in, there is a way" [18:52] bdx: sorry, I was saying "I wonder if we can document some path that allows you to get your password if ssh works" [18:52] rick_h_: aaah [18:53] rick_h_: its not a biggie, I can just tear down and redeploy [18:53] rick_h_: thx [18:53] bdx: k, sorry. Thanks for bringing up the issue [19:21] cory_fu, bcsaller: Work in progress PR for the glitch plan/actions refactor here: https://github.com/juju-solutions/matrix/pull/4 [19:22] (I merged the two dicts, but haven't actually attempted to execute the code just yet -- it is very much a WIP!) [19:22] https://github.com/juju-solutions/matrix/pull/4 [19:22] petevg: thanks :) [19:22] np [19:31] rick_h_: bdx no [19:33] ryebot Cynerva - bite sized one here https://github.com/juju-solutions/interface-sdn-plugin/pull/4 [19:33] lazyPower: looking [19:35] ryebot Cynerva less bite sized review here: https://github.com/juju-solutions/charm-flannel/pull/20 [19:35] lazyPower: +1 also looking [19:37] ryebot Cynerva - and finally - the consumer of those changes https://github.com/juju-solutions/layer-docker/pull/92 [19:38] lazyPower: cool looking [19:38] that should be it though, thanks for taking a look [19:49] lazyPower: commented! [19:50] ryebot good shout, seems odd but it is indeed possible [19:50] * lazyPower amends [20:00] rick_h_, perrito666: when I pass the debug flag to `juju bootstrap`, I see the subnets in my vpc/region listed as constraints, see http://paste.ubuntu.com/23350343/ [20:01] rick_h_, perrito666: does this mean I can somehow add a constraint to select one of the subnets on bootstrap? [20:02] to bootstrap to? [20:02] bdx: wouldn't know sorry :( [20:05] bdx: you may be able to if you create some spaces that encapsulate them, I'm not quite sure of the current state of that on ec2 [20:07] mgz: you have to have a controller first though, to be able to create spaces [20:07] hm, not in the maas model [20:07] but yeah, that's an issue with ec2 [20:08] from the context of juju though [20:08] so, from the context of juju on maas, you can create a maas space that includes a subnet, then bootstrap with a constraint that the machine must be in that sapce [20:09] mgz: totally [20:10] and the openstack provider accepts a 'network' config on bootstrap [20:10] that's mostly a hack to deal with multiple neutron networks existing [20:11] mgz: hack or no hack it provides a solution [20:11] so, I'm not actually sure what the finished solution should look like for ec2/openstack [20:12] mgz: is there a similar hack for ec2 that you know of? [20:12] :-) [20:12] I added the openstack one, I did not add an ec2 one at the same time, let me just check the provider [20:13] hm, no you can force a particular vpc [20:13] so, you can maybe create a seperate vpc with what you want and make juju use that? but pretty hard limits on a bunch of these network bits in ec2 [20:14] yeah, now thats a hack [20:14] sorry bdx [20:14] mgz: I feel it would be such an easy add [20:15] just filter the list of subnets returned on bootstrap and ensure the one selected exests and is in the region and vpc [20:15] is there something I'm missing here? [20:15] I think we probably don't want to do that without a good model that works across providers, is the holdup [20:16] but I'm still somewhat out of the loop on networking plans currently [20:17] meh [20:17] mgz: thanks for your input here [20:18] bdx: it may be worth a message to the list just to spell out your use case and find out what the plans are [20:18] mgz: entirely [21:05] mgz, rick_h_: so, a new environSchema field for 'subnet' here -> https://github.com/juju/juju/blob/staging/provider/ec2/config.go#L28 [21:06] and a 'getUserSubnet()' here -> https://github.com/juju/juju/blob/staging/provider/ec2/environ_vpc.go#L198 [21:08] or possibly 'getBootstrapVPC' [21:11] rick_h_: again, http://paste.ubuntu.com/23350619/ [21:13] rick_h_: `charm list -u createivedrive` shows my charm -> http://paste.ubuntu.com/23350624/ [21:13] but I can't seem to deploy it [21:14] I have a feeling this is highly related to the reason I can't see my charms for any other namespace other than my user namespace in the store [21:14] via juju-gue [21:14] *juju-gui [21:18] while logging into jujucharms.com, I am still able to view namespaces other than my personal namespace, see -> https://s18.postimg.org/skemexkk9/Screen_Shot_2016_10_19_at_2_16_15_PM.png [21:19] when logged into the charmstore via juju-gui, I don't have access to any of the charms in team namespaces, see -> https://s17.postimg.org/thm3ooapb/Screen_Shot_2016_10_19_at_2_18_15_PM.png [21:29] bdx: i've started work on a Juju cookbook, when you finish up your networking/subnetting and any other crazy stuff, i'll be picking your brains [21:29] content for the "crazy and advanced" section [21:47] magicaltrout: ha, thanks, hopefully I can have some good adds for you! ... (harsh segue) )do you have a launchpad team, or charmstore namespace other than your personal namespace? [21:49] i don't, i caused enough havoc in the charmstore by renaming my user in launchpad [21:49] that made enough shit break so i leave it well alone ;) [21:51] bahahah [21:52] your next mission, should you choose to accept it [21:54] create a launchpad team, push some charms up to different channels, add 50 users, have the users start pushing charms, rename the launchpad team, remove and re-add 1/2 of your users to the newly named team, change team name againn, submit bug [23:30] juju set/set-config, and get as well, was removed in 2.0.0? [23:38] MrDanDan: not removed. renamed - feel free to browse release notes for 2.0 for changes :D https://jujucharms.com/docs/2.0/reference-release-notes [23:40] thanks