[00:01] <CarlFK> marcoceppi:  hi!  ..  https://github.com/marcoceppi/charm-ubuntu   -  advice on how to set hostsnames ?
[00:01] <marcoceppi> CarlFK: could you elaborate why? Curious your use case and where you're deploying
[00:02] <marcoceppi> it'll help me get you the best answer
[00:03] <CarlFK> sure - there is an ansible playbook that provisions 12ish machines (hardware, not VM or containers)... it was started, but hasn't been run ever.  I want to test it
[00:03] <CarlFK> I don't care if the apps that get deployed actually run in the containers, I just want to A) sniff out the errors, and B) look at config files
[00:04] <CarlFK> https://anonscm.debian.org/git/debconf-video/ansible.git/tree/setup_ansible.sh
[00:05] <CarlFK> some machines were setup manually, then that playbook was written after the fact.  "No idea if it works, and  we haven't done the dynamic bits yet"
[00:06] <CarlFK> so I want to hack on it on my 8gig laptop.
[00:15] <CarlFK> marcoceppi:   here are the 12 hostnames:   https://anonscm.debian.org/git/debconf-video/ansible.git/tree/inventory/hosts
[00:16] <marcoceppi> CarlFK: confused on the reference to the Ubuntu charm
[00:16] <CarlFK> I don't mind cut/pasting those into some other file, but if there is something that will slurp them from that file, that would be handy
[00:17] <CarlFK> I want to bring up 12 things I can tell ansible to deploy into, then ssh in and look at what got setup
[00:17] <marcoceppi> CarlFK: there's two real ways for hostname setting on Ubuntu, `hostname` command and overwritting `/etc/hostname`. You'll need to do the latter for persistence between reboots
[00:18] <marcoceppi> CarlFK: that's not in the charm, but the best way to add it into the charm would be something like this:
[00:24] <marcoceppi> CarlFK: https://github.com/marcoceppi/charm-ubuntu/pull/1/files
[00:25] <CarlFK> marcoceppi:  thanks!
[00:27] <skay_> oh hey. for some reason I didn't know about config.changed.foo. that's nice
[00:28] <skay_> I eavesdropped on you PR
[00:32] <CarlFK> marcoceppi: I just started using juju a few hours ago.. just to do this.   I git cloned/checkout carlfk... how do I tell juju to use it?
[00:32] <marcoceppi> CarlFK: you
[00:32] <marcoceppi> CarlFK: you'll need to compile the charm, this involes installing charm-tools and then running charm build. I can just compile it for you and give you a URL to try it out with if it's easier
[00:38] <marcoceppi> CarlFK: I've uploaded a version with the configuration patch to the charm store in the edge channel
[00:38] <marcoceppi> CarlFK: `juju deploy ubuntu --channel edge` should get you the charm with the changes
[02:20] <CarlFK> marcoceppi: ERROR cannot resolve charm URL "cs:xenial/ubuntu": cannot get "/xenial/ubuntu/meta/any?include=id&include=supported-series&include=published": unauthorized: access denied for user "carlfk"
[02:33] <CarlFK> git... charm build ... seems to have worked.
[03:28] <siva> I have one application running in a machine. Is there a way to remove the application without removing the machine?
[03:28] <siva> In juju2.0 I find that when the application is deleted the machine is also removed
[04:09] <CarlFK> marcoceppi: +    with open('/etc/hostname', 'wt') as f:
[04:09] <CarlFK> t for text, cuz py 3
[04:17] <siva> In juju2.0 I find that when the application is deleted the machine is also removed
[04:17] <siva> Is there a way to remove the application without removing the machine?
[07:53] <kjackal> Good morning Juju world
[07:55] <kjackal> siva I do not think you can have direct controll over the resource provisioning. What you can do is juju deploy ubuntu, than juju deploy <your charm> --to <machine with ubuntu> and after that if you remove-application <your charm> the machine should remain because it holds the ubuntu charm
[09:11] <BlackDex> How do i upgrade the juju controller? I want to update/upgrade to 2.0.2 and i currently have 2.0.1
[09:18] <zeestrat> BlackDex: I think you have to switch to the controller model, run juju upgrade-juju, then do the same on your other models. But check locally just to make sure.
[09:20] <BlackDex> and how do i switch to the controller model?
[09:20] <BlackDex> the logic is lost for me atm :p
[09:20] <BlackDex> juju switch is telling me i'm already there
[09:22] <zeestrat> juju models should list which models you have and a star marked next to the one you're currently on.
[09:23] <BlackDex> Model       Cloud/Region  Status     Machines  Cores  Access  Last connection
[09:23] <BlackDex> controller  maas          available         1      2  admin   just now
[09:23] <BlackDex> maas*       maas          available        25    192  admin   2 minutes ago
[09:24] <zeestrat> Ok, so that says the maas model is active. "juju switch controller" should switch you to the controller.
[09:24] <BlackDex> so the controller which is used for maas is seperated then?
[09:26] <BlackDex> ah
[09:26] <BlackDex> wait
[09:26] <zeestrat> The wording is a bit confusing. You're on the same controller as the switch command switches between models on that controller. The default model name for controller is "controller"
[09:26] <zeestrat> That caught me out the first time too :)
[09:27] <BlackDex> because the juju controller can handle multiple models it is "separeted"
[09:27] <zeestrat> Jupp. So if you setup multiple controllers with HA, they would show up in the list of machines in the "controller" model
[09:27] <BlackDex> and i can add multiple models to one controller, so i don't have to bootstrap for every single environment?
[09:28] <BlackDex> lets say, maas and openstack
[09:29] <zeestrat> Multiple models per controller = Yes. However different juju substrates/clouds (AWS, MAAS, OpenStack, etc.) need their own bootstrapped controller.
[09:29] <BlackDex> where are nice graphic reprecentations of this :p
[09:32] <zeestrat> I hear you! Took me a while to parse the docs. A graphic would help explain a lot. I'll add a bug to the docs.
[09:33] <BlackDex> But multiple times of reading the docs is making it a bit more clear
[09:34] <zeestrat> Note, if you're upgrading a prod environment I'd recommend going through the upgrade process on a dev environment such as AWS or just locally with LXD so you get the feel for it.
[09:35] <BlackDex> it's a prod/dev thingy, so no prob here :)
[09:40] <zeestrat> BlackDex: Added a bug to the docs for that graphic https://github.com/juju/docs/issues/1575
[09:41] <BlackDex> Thx :) I have subscribed to the bug! :) Thx
[10:53] <junaidali> hey guys, is there any HA bundle available for juju 2.0?
[11:13] <zeestrat> Hey junaidali, do you mean HA juju controllers?
[11:18] <zeestrat> If so, check out https://jujucharms.com/docs/2.0/controllers-ha. Otherwise check out https://jujucharms.com/docs/2.0/charms-ha and https://jujucharms.com/hacluster/
[12:55] <marcoceppi> CarlFK: hah, thanks for the catch. Not sure why edge wasn't shared, but it's now open to everyone
[13:20] <SimonKLB> it looks like there is a race between the relations and the reactive relation states - for example it takes a while before the influxdb.available state is removed from the time you have removed the actual relation
[13:21] <SimonKLB> that causes the relation data to be empty, even though you're expecting it not to when the available state is set
[13:22] <SimonKLB> i could do an extra check that the relation data is actually there, but it would be better if the state was better synced so that charms could rely on them more
[13:56] <jcastro> rick_h: invites out to you, I just realized that you should just be a coowner of the juju g+/youtube account
[13:57] <jcastro> https://www.youtube.com/live_dashboard is the url you want after you add Juju to your youtube dropdown thing
[13:57] <jcastro> from there you can schedule events, etc.
[13:57] <rick_h> jcastro: cool yea I was going to ask you about that at some point
[13:57] <rick_h> jcastro: do I get added to the youtube group some how?
[13:58] <jcastro> yeah
[13:58] <jcastro> youtube inherits the G+ permissions
[13:58] <rick_h> oic
[13:58] <rick_h> jcastro: ok, so I've followed Juju, do you then see me/grant access?
[13:59] <junaidali> zeestrat: deploying a multi-controller OpenStack
[13:59] <jcastro> yeah you need to add an account to the top right account google thing
[13:59] <jcastro> then sign in, and you should see juju as an option in the dropdown
[14:00] <jcastro> if that's a problem just lmk and we'll hop on a hangout
[14:04] <rick_h> jcastro: yea, that'd be great. Evidently I'm an addiot
[14:04] <rick_h> addiot? lol idiot
[14:04] <rick_h> cccccceefhftjrkgvjfinhtcfjhrbfjdhtvhnruucudi
[14:05] <jcastro> it's ok rick, it's just a youtube channel.
[14:05] <jcastro> :)
[14:05] <zeestrat> rick_h: Do I spot a yubikey?
[14:05] <rick_h> zeestrat: maybe :P
[14:05] <rick_h> nano is great but sometimes "bump"
[14:05] <rick_h> jcastro: https://hangouts.google.com/hangouts/_/canonical.com/rick?authuser=1
[14:05] <zeestrat> junaidali: There are some older docs at https://wiki.ubuntu.com/ServerTeam/OpenStackHA Otherwise I suggest asking in #openstack-charms or the official Canonical folks in here such as rick_h or jcastro
[14:15] <zeestrat> rick_h: yeah, I've had opsec go out the window a couple of times when focus follows mouse ;)
[15:39] <arosales> morning
[15:39] <arosales> rick_h: per your reply, ref = https://lists.ubuntu.com/archives/juju/2016-December/008341.html
[15:40] <arosales> are you saying local is not supported, or resources in bundles are not supported at all in bundles ?
[15:43] <rick_h> arosales: local not supported and on the roadmap this cycle. You can reference resources in bundles via the revision of the resource in the charm store
[15:45] <arosales> rick_h: so there is no new key in bundle.yaml. I just have to attach the resource to the charm when I upload to the store, and then be explicit about that version in the "charm:" key in the bundle.yaml, correct?
[15:46] <rick_h> arosales: no, second let me get the example
[15:46] <arosales> rick_h: ok thanks
[15:47] <kwmonroe> arosales: what you said is what i thought rick meant (charm rev dictates the resource rev).  thanks for clarifying!  i'm fixin to learn something.
[15:50] <rick_h> kwmonroe: arosales so I believe there is a key for the resources
[15:50] <rick_h> https://github.com/juju/charm/blob/v6-unstable/bundledata_test.go#L27kwmonroe: arosales see
[15:50] <rick_h> bah
[15:51] <rick_h> that didn't paste right
[15:51] <rick_h> https://github.com/juju/charm/blob/v6-unstable/bundledata_test.go#L27
[15:51]  * arosales looks
[15:51] <rick_h> arosales: kwmonroe so you can have the sub-key in the application for resources, and then for each resource "name" you can specify a revision number
[15:51] <rick_h> arosales: kwmonroe and the goal for this cycle is to allow that revision to also be able to be a local path like a local charm path would be
[15:53] <cory_fu> rick_h: If the bundle is referencing a local charm (and thus doesn't have a definite charm store entry), can you still indicate that it should be deployed with a resource from the store?
[15:54] <rick_h> cory_fu: honestly, never tried that and no idea how that'll work. I'm guessing it'll fail because the charm id in the bundle is ./ and not resolvable via the charmstore to see what resources it provides
[15:55] <cory_fu> rick_h: So, it would be nice for CI to be able to say, for a local charm in the bundle, use either this local URL for the resource, or this fully-qualified resource URL from the charm store.
[15:55] <rick_h> cory_fu: rgr
[15:55] <rick_h> cory_fu: I can take that as feedback that a local filepath, revision, or a fq cs url would be great for the bundle spec
[15:56] <cory_fu> Yep
[15:56] <cory_fu> Tahnks
[15:56] <cory_fu> Thanks
[15:56] <rick_h> thank you for the great feedback
[15:56] <arosales> rick_h: +1 to cory_fu's comment, and not the _or_ there
[15:57] <rick_h> arosales: rgr
[15:57] <arosales> rick_h: as a charm developer I want to iterate on a charm that has resoruces locally.
[15:57] <rick_h> arosales: understand
[15:57] <cory_fu> Also, as a charm developer, I want to test a local build of a charm that does already have a resource in the store
[15:58] <arosales> in a bundle testing context I need to specify both the resource url and/or the charm url. I could see a case where both bits are changing during development. A more common case though would be the charm is changing locally and the resource url hasn't changed.
[15:58] <rick_h> arosales: rgr
[15:58] <arosales> rick_h: thanks!
[16:00] <arosales> kwmonroe: kjackal: I guess for now in charm ci if we want to accommodate charms with resources the charm has to be uploaded to the charm store
[16:01] <arosales> kwmonroe: kjackal: for the 1st milestone we are working on I think we can assume that the charm doesn't have resources, but we should look at the work to accomodate a charm with resources until juju lands the above mentioned work
[16:11] <kwmonroe> +1 arosales
[16:18] <rick_h> jcastro: we need a hash tag for juju. #juju doesn't work well with the amount of Juju music stuff. ideas?
[16:22] <arosales> rick_h: #jujudevops or #jujucloud or #jujuops ?
[16:22] <arosales> just throwing some out there
[16:23] <jcastro> We've used #jujucharms in the past, but that might be too long
[16:23] <rick_h> arosales: yea, I want to start updating our 'join the mailing list, join irc..." with a hash tag setup as well I think
[16:23] <rick_h> #use-juju? hate it being the second word though
[16:25] <tvansteenburgh> #jujucharms imo
[16:27] <rick_h> tvansteenburgh: k, let's try that and see if we can get any update. /me goes to add a search for jujucharms
[16:36] <arosales> rick_h: pinged james donner and he has been using #jujucharms
[16:36] <arosales> that seems fine and has prior art
[16:36] <arosales> marcoceppi: ^
[16:36] <rick_h> arosales: cool ty
[16:49] <lazyPower> +1
[17:04] <arosales> petevg: did you continue to see the agent dying in your testing? if so did you have those bug links hand?
[17:05] <arosales> juju agent that is
[17:05] <petevg> arosales: The juju agent isn't dying. But the controller marks trusty machines as failed after a reboot.
[17:05] <petevg> arosales: bug here: https://bugs.launchpad.net/juju/+bug/1649637
[17:05] <mup> Bug #1649637: Juju agent in a "failed" state after machine reboot on some charms <juju:Triaged> <https://launchpad.net/bugs/1649637>
[17:06] <petevg> arosales: matrix will also deliberately kill an agent. It should come back up. If you're not seeing it do so, that might be a different bug.
[17:07] <arosales> petevg: cory_fu: is 1649637 the main bug or are there others we should make juju-core aware of others
[17:08] <petevg> arosales: that is the only bug that I am aware of. If there are others, they need to get filed.
[17:08] <arosales> petevg: thanks
[17:08] <petevg> np
[17:08] <cory_fu> arosales: Yeah, I don't think there are any others that are blocking us at the moment
[17:09] <arosales> cory_fu: it sounds like a bug didn't get filled for juju marking a machine as "destroyed" immediately in the api so further requests don't hit that same machine?
[17:25] <petevg> arosales: Nope. That was a red herring. Destroying the machine doesn't break stuff. The API does the right thing.
[17:25] <arosales> petevg: ah excellent to hear
[17:33] <skay> why would mojo not be able to find a charm in the store but calling juju deploy from the command line does?
[17:41] <skay> (I'm looking at the source code now, but if anyone knows, let me know)
[17:41] <stub> skay: You are using a cs: URL in your manifest?
[17:41] <cory_fu> tvansteenburgh: Updated https://github.com/juju/python-libjuju/pull/31
[17:41] <skay> stub: yes, cs:~codersquid/submission-service
[17:42] <stub> skay: That is codetree rather than mojo. Might make it easier to track down the issue.
[17:42] <skay> stub: thanks
[17:43] <skay> I'm looking at this stacktrace https://paste.ubuntu.com/23629882/
[17:46] <stub> https://jujucharms.com/u/codersquid/submission-service gives me a 404
[17:46] <skay> does this work? https://jujucharms.com/u/codersquid/submission-service/0
[17:46] <stub> Nope
[17:47] <skay> (I'm assuming it will give a 404 to you but might as well check)
[17:47] <skay> arg this is confusing. I'm not logged in
[17:48] <skay> I probably pushed it incorrectly, so I'll review the docs for making a release
[17:48] <cory_fu> tvansteenburgh: Ugh.  Thanks for the catch.
[17:48] <tvansteenburgh> cory_fu: :) np
[17:48] <stub> skay: You might have missed the step granting read access to everyone?
[17:49] <stub> (I don't know if codetree/mojo supports private charms on cs: URLs)
[17:57] <skay> stub: you are right, I missed the grant step
[17:58] <skay> stub: I added a resource (which should be optional) so that users of the charm can either use the snap store or attach a snap resource
[17:58] <skay> when I try to do a release, I get an error, resources are missing from publish request
[17:59] <stub> yes, that policy in the charm store is causing lots of people grief.
[18:00] <stub> The work around is to attach 0 byte files as resources, but I haven't had a chance to see if that works with the snap layer yet
[18:00] <stub> (it will be a minor fix if it doesn't)
[18:00] <skay> I could try it out and see what happens. I tried attaching a snap as a resource for the charm, but I really don't want to do that
[18:02] <stub> skay: If it fails, you can work around by having mojo deploy from a git branch rather than the charm store. I can sort the snap layer tomorrow (unless you beat me to it)
[18:13] <anrah> Question about resources.. can I set them on bundle-file?
[18:14] <anrah> i tried to use resources: and then list needed resources one by one but that did not work
[18:17] <skay> anrah: I opened an issue for it https://bugs.launchpad.net/juju-deployer/+bug/1643692
[18:17] <mup> Bug #1643692: add feature to support juju2 resources <juju-deployer:New> <https://launchpad.net/bugs/1643692>
[18:17] <skay> I want to be able to use options too
[18:18] <anrah> ok, thanks for that :)
[18:18] <anrah> i'll guess that I have to use block when resource is not available
[18:37] <skay> stub: it does cause a hiccup. is checking for a zero length file a good approach for it? https://paste.ubuntu.com/23630058/
[18:49] <jcastro> rick_h: wanna fire this up early? the hangout I mean
[18:49] <rick_h> jcastro: already there
[18:50] <rick_h> jcastro: https://goo.gl/mgACZH for the HO
[18:50] <rick_h> jcastro: http://youtu.be/FwLEMa7XE64 for the youtube watcher link
[18:52] <skay> stub: https://paste.ubuntu.com/23630104/ I'm going to try this, but I don't know how to actually test changes I make for a layer
[18:54] <rick_h> alexisb: cory_fu kwmonroe links above ^ use the HO link
[18:56] <cory_fu> rick_h: Access denied
[18:56] <rick_h> cory_fu: boooo, /me looks
[18:56] <tvansteenburgh> cory_fu: i got that at first to
[18:56] <tvansteenburgh> but refreshed and it worked
[18:56] <tvansteenburgh> (too)
[18:56] <rick_h> cory_fu: please try again, it let everyone else in
[18:56] <rick_h> or did you offend the google gods?
[18:56] <cory_fu> rick_h: I fixed it by removing the explicit authuser=0
[18:56] <rick_h> ah ok
[18:57] <rick_h> HEY EVERYONE! Juju Show #2 about to start if you're interseted
[18:58] <rick_h> chat here: https://goo.gl/mgACZH and follow here: http://youtu.be/FwLEMa7XE64
[19:02] <jrwren> jcastro: that is not funny. ;)
[19:07] <kwmonroe> i need optional resources!  ibm-was-base, for example, takes an installer resource and a fixpack resource.  fixpack should be optional!  https://api.jujucharms.com/charmstore/v5/~ibmcharmers/ibm-was-base/archive/metadata.yaml
[19:17] <cory_fu> http://svg.juju.solutions/
[19:23] <kwmonroe> https://lists.ubuntu.com/archives/juju/2016-December/008312.html
[19:27] <zeestrat> The controller/model upgrade procedure sure could use some better docs both in CLI and online. See https://bugs.launchpad.net/juju/+bug/1638714
[19:27] <mup> Bug #1638714: Upgrading a juju model before the controller gives unhelpful error message <usability> <juju:Triaged> <https://launchpad.net/bugs/1638714>
[19:27] <cory_fu> https://github.com/juju-solutions/matrix
[19:27] <arosales> kwmonroe: is there a bundle in the charm store for cwr-bundle-ci?
[19:29] <arosales> kwmonroe: or should folks just use https://github.com/juju-solutions/bundle-cwr-ci
[19:29]  * arosales guesses the latter
[19:30] <kwmonroe> arosales: yeah, there is a bundle, but it's in -edge.  you can juju deploy -c edge ~juju-solutions/cwr-ci
[19:30] <kwmonroe> let me check our edge perms to see if i can give everyone read access
[19:31] <arosales> kwmonroe: thanks
[19:35] <lazyPower> rick_h <3 https://github.com/juju/docs/issues/1576
[19:43] <kwmonroe> arosales, et al: ci bundle:  https://jujucharms.com/u/juju-solutions/cwr-ci/2;  same thing with revq in the mix:  https://jujucharms.com/u/juju-solutions/cwr-rq/3
[19:43] <arosales> kwmonroe: thanks
[19:44] <jrwren> rick_h: you would not like my machine name :)
[19:44] <rick_h> jrwren: :P
[19:44] <rick_h> one word or bust!
[19:44] <jrwren> rick_h: oh, its one word... gogogogogogogo
[19:44] <rick_h> lol
[19:44] <jrwren> and yes, I get the number correct every time.
[19:44] <rick_h> tab complete ftw
[19:45] <jrwren> no tab complete. I just know the song to sing.  Also, its just "gogo" in my PS1. Does that help at all?
[19:45] <sparkiegeek> cory_fu: https://github.com/juju-solutions/matrix/issues/46 FYI
[19:47] <cory_fu> sparkiegeek: There is already a PR to address that: https://github.com/juju-solutions/matrix/pull/44  :)
[19:47] <sparkiegeek> cory_fu: damn it, did you steal Guido's time machine again? (thanks)
[19:47] <cory_fu> :)
[19:48] <cory_fu> sparkiegeek: Feel free to review that PR for me.  ;)
[19:48] <cory_fu> (Rendered view: https://github.com/juju-solutions/matrix/blob/ef7d2171635243fc0d8ed3e7adfaebd2026bdbe2/README.md)
[19:52] <Merlijn_S> Happy holidays!
[19:52] <jrwren> thanks rick_h
[19:53] <jrwren> great demo Alexis
[19:53] <marcoceppi> sad I couldn't join in, looking forward to watching soon
[19:54] <rick_h> marcoceppi: we missed you!
[19:54] <marcoceppi> NEXT TIME
[19:54] <rick_h> Jan 4th
[19:54] <rick_h> be there!
[19:54] <sparkiegeek> cory_fu: looks like requirements.txt is asking for ubuntui >= 0.1.1 but that doesn't exist (on pypi)
[19:54] <marcoceppi> bcsaller, petevg: Juju Show might run over making me late for the Matrix sync
[19:57] <cory_fu> sparkiegeek: All the deps for matrix are bundled in the wheelhouse/ folder, but I will ping stokachu to get that version out to pypi. :)
[19:58] <cory_fu> sparkiegeek: Hrm.  I see, the instructions say to install it from GH, which won't use the wheelhouse.
[19:58] <skay> My charm uses a layer, and I have a fix for hte layer to disregard empty resources. I don't know how to test changes to a layer. helphelp
[19:58] <cory_fu> stokachu: Can we get a 0.1.1 release of ubuntui?
[19:58] <skay> change is here https://code.launchpad.net/~codersquid/layer-snap/+git/layer-snap/+ref/handle-empty-resources
[19:58] <sparkiegeek> cory_fu: yeah I didn't quite follow those anyway - I just ran "pip3 install ." from a clone
[19:58] <skay> I think stub will be around tomorrow and could take a look, but I would like to test it
[19:58] <cory_fu> sparkiegeek: Fair enough.  If you add `-f wheelhouse` to that, it will work
[19:58] <stokachu> cory_fu: yea
[19:59] <cory_fu> stokachu: Thanks!
[20:00] <stokachu> why is it so difficult to upload a resource to the charm store
[20:01] <stokachu> http://paste.ubuntu.com/23630347/
[20:01] <stokachu> what is the problem here
[20:02] <sparkiegeek> cory_fu: oh, looks like bcsaller already merged it :/ still looks bogus to me - not your branch but contents of requirements.txt
[20:03] <sparkiegeek> cory_fu: the next headache I got was "juju>=0.0.1,<1.0.0" pulling in pyjuju from PyPi and trying to install zookeeper (!)
[20:03] <sparkiegeek> y'all should get some CI on your CI framework :P
[20:05] <cory_fu> tvansteenburgh2: ^ is us hitting up against the already claimed "juju" project on pypi.  Can we prioritize reclaiming that or working around it with a different package name?
[20:05] <tvansteenburgh2> that's why you need to install from the wheelhouse sparkiegeek
[20:06] <tvansteenburgh2> the main dep isn't even released yet
[20:06] <tvansteenburgh2> cory_fu: yeah, i'll look into that this afternoon
[20:06] <cory_fu> tvansteenburgh2: Thanks.  Other option is for me to  update the instructions in the README, but I dislike not being able to install it the normal way
[20:07] <cory_fu> tvansteenburgh: But if it's going to take some time, then I'll update the README
[20:07] <tvansteenburgh> cory_fu: yeah i need to get this sorted asap anyway since i wanted to do a libjuju release next week
[20:09] <sparkiegeek> tvansteenburgh: I'm trying. "pip3 install . -f wheelhouse/" lead to: http://paste.ubuntu.com/23630386/
[20:11] <stokachu> cory_fu: https://pypi.python.org/pypi/ubuntui/0.1.1
[20:12] <cory_fu> stokachu: Great, thanks!
[20:12] <stokachu> np
[20:12] <kwmonroe> stokachu: did you get the charm attach issue sorted?
[20:13] <stokachu> kwmonroe: not yet
[20:13] <kwmonroe> stokachu: have you run the exact same command again?  i know it sounds dumb, just run it again.
[20:13] <stokachu> kwmonroe: http://paste.ubuntu.com/23630362/ my latest failed attempt
[20:13] <stokachu> k running again
[20:14] <petevg> sparkiegeek: does "tox -r" succeed? (We might have a dependency listed in a place that tox can spot, but pip can't.)
[20:14] <stokachu> kwmonroe: still fails
[20:14] <stokachu> i was just told new packages are being pushed to charm store
[20:14] <stokachu> so ill try again later
[20:14] <sparkiegeek> petevg: yes it does
[20:15] <sparkiegeek> cory_fu: petevg: bcsaller: I'm going to reopen #46, since although the specific failure case is no longer happening after Cory's branch landed, it does still exhibit errors
[20:18] <kwmonroe> ok stokachu -- i have seen that before, but running attach twice worked.  maybe it's just bad store timing.
[20:21] <tvansteenburgh1> hazmat: can i take over the 'juju' package name on pypi? we have a new python client that we want to put there
[20:25] <petevg> cory_fu: I don't get sparkiegeek's permission denied error when building zkpython, but I do get an error about a missing header file.
[20:26] <sparkiegeek> petevg: yeah the permission thing is my local system being a bit screwy - http://paste.ubuntu.com/23630440/ was in a clean LXD
[20:27] <sparkiegeek> in the (recently changed) README.md there's no use of the wheelhouse
[20:28] <cory_fu> sparkiegeek: Also, your pastebin with the permission denied didn't include "sudo"
[20:29] <sparkiegeek> cory_fu: indeed, deliberately so
[20:29] <petevg> cory_fu: Yeah. But there's a build issue even if you're not sparkiegeek, and trying to build stuff in your system Python :-)
[20:30] <sparkiegeek> a snap for it would be nice (and avoid this mess)
[20:30] <petevg> True.
[20:32] <sparkiegeek> sudo pip3 install git+https://github.com/juju-solutions/python-libjuju.git#egg=juju
[20:32] <sparkiegeek> sudo pip3 install 'git+https://github.com/juju-solutions/matrix.git'
[20:32] <sparkiegeek> in that order seems to be the right incantation
[20:32] <sparkiegeek> (but it still ignores the wheelhouse)
[20:32] <cory_fu> sparkiegeek, petevg, bcsaller: https://github.com/juju-solutions/matrix/pull/47
[20:41] <cory_fu> sparkiegeek, bcsaller, petevg: I was missing the --no-index which was causing it to try the old juju lib from pypi
[20:41]  * sparkiegeek spins up another clean LXD
[20:41] <petevg> cory_fu: aha. That works for me in my virtualenv :-)
[20:53] <sparkiegeek> cory_fu: :( so the install "worked" but actually running it fails miserably
[20:54] <sparkiegeek> cory_fu: http://paste.ubuntu.com/23630574/
[20:54] <cory_fu> sparkiegeek: Ah.  Missing the file from the manifest.
[20:56] <sparkiegeek> cory_fu: I'll open a new issue
[20:58] <sparkiegeek> https://github.com/juju-solutions/matrix/issues/49
[21:05] <sparkiegeek> petevg: I guess you'll also be seeing https://github.com/juju-solutions/matrix/issues/49 ?
[21:06] <petevg> sparkiegeek: ayup.
[21:06] <petevg> Agree w/ cory_fu that it should just be a matter of adding the default .yaml to the manifest, though.
[21:07] <cory_fu> petevg, sparkiegeek: https://github.com/juju-solutions/matrix/pull/50
[21:10] <petevg> cory_fu, sparkiegeek: that branch seems to work for me.
[21:10] <petevg> Matrix is running, at least ...
[21:10] <petevg> Went ahead and merged it ...
[21:11] <petevg> cory_fu: ... though it looks like maybe you didn't rebase from master, as I see a datetime related error.
[21:12] <cory_fu> petevg: It definitely was rebased from master
[21:12] <petevg> cory_fu: darn it. Looking at the trace ...
[21:13] <petevg> Huh. Maybe it's grabbing the wrong version of libjuju ...
[21:15] <petevg> cory_fu: yeah. Wrong version of python-libjuju. The file sizes are different. Maybe a merge clobbered my checkin.
[21:17] <cory_fu> petevg: https://github.com/juju-solutions/matrix/commits/master/wheelhouse/juju-0.0.1-py3-none-any.whl looks fine to me
[21:19] <petevg> cory_fu: I must have messed up the commit, then. Copied the wrong thing. :-/
[21:19] <cory_fu> petevg: I'm running a test in a fresh venv.  I'll let you know if I see the same thing
[21:20] <cory_fu> petevg: At what point did you see the error?
[21:20] <sparkiegeek> I don't see timestamp issues per se, but I get a libjuju traceback trying to decode JSON from constraints
[21:20] <petevg> cory_fu, sparkiegeek: I just pushed an updated python-libjuju build to matrix master. If you have trouble, pulling the new build should fix things.
[21:20] <petevg> sparkiegeek: interesting. That's new. What bundle are you testing?
[21:22] <sparkiegeek> petevg: https://jujucharms.com/u/landscape/landscape-dense/
[21:22] <petevg> sparkiegeek: cool. I'll download it and see if it does the same thing on my machine.
[21:23] <cory_fu> petevg: I didn't get any error wrt timestamps on the dep before you just pushed, but a new one won't hurt anything either.
[21:23] <sparkiegeek> petevg: FWIW http://paste.ubuntu.com/23630680/ is relevant snippet from matrix.log
[21:24] <petevg> cory_fu: Cool. I figured the new build wouldn't hurt anything (it's actually possible that I hadn't pull from master before I tested).
[21:25] <sparkiegeek> petevg: naively reading the traceback, it seems libjuju might be expecting constraints to be JSON?
[21:25] <sparkiegeek> *shrug*
[21:27] <petevg> sparkiegeek: That's the root of the problem, but there's a trip to the plan builder API in there, and a bunch of annoying complexities. I think that this may be related to a bug that I squashed, but maybe didn't squash all the way ...
[21:29] <petevg> cory_fu, sparkiegeek: looks like we need to do more of the thing that I did in https://github.com/juju/python-libjuju/blob/master/juju/constraints.py, but for resources. Will file a ticket.
[21:36] <petevg> cory_fu, sparkiegeek: https://github.com/juju/python-libjuju/issues/35
[21:37] <petevg> Not a fast fix, unfortunately :-/
[21:38] <sparkiegeek> petevg: resources? there are no resources here?
[21:38] <sparkiegeek> petevg: am I missing something?
[21:44] <petevg> sparkiegeek: I thought that the thing under "options" in landscape-server was a resource. In any case, the planner says that we need to do something with "https://github.com/juju/python-libjuju/issues/35", and that's getting turned into a Value object in python-libjuju, and that's failing, because we need to parse it first.
[21:44] <petevg> ... which is going to change in future versions of juju, because that's too much parsing on the client's part. But for now, it's what we have to do :-)
[21:46] <petevg> sparkiegeek: actually. nevermind. I'm totally wrong. Misread the traceback. It's the constraints. Which are supposed to be fixed.
[21:46] <petevg> yarr.
[21:51] <petevg> sparkiegeek, cory_fu: https://github.com/juju/python-libjuju/issues/36
[21:52] <petevg> That's easier to fix. I have to head out to an event in a few minutes, though, so I will have to fix it tomorrow (unless someone else grabs it).
[22:17] <rick_h> cory_fu: kwmonroe ping, either of you free to help me out with something please?
[22:17] <rick_h> lazyPower: as well ^
[22:18] <cory_fu> rick_h: Sure, what's up?
[22:19] <rick_h> cory_fu: can you join https://hangouts.google.com/hangouts/_/canonical.com/bs-squad?authuser=1 please? (adjust authuser)
[22:36] <marcoceppi> I like the name bs-squad
[22:36] <lazyPower> it clearly stands for "Bear suit"
[23:05] <jrwren> is there a way to juju run on every machine? e.g. I don't know what contraints were used to deploy and so I want to cat /proc/cpuinfo and run free on every machine and collect the results.
[23:05] <jrwren> nevermind. NOW I see juju run --all
[23:31] <lazyPower> jrwren hey there's juju run --all
[23:31] <lazyPower> (extremely latent)
[23:32] <kwmonroe> next time someone brings up the elastic beats stack or prometheus or nagios, i'm gonna be like "juju run --all 'cat /proc/cpuinfo && free".  thanks jrwren!
[23:32] <lazyPower> kwmonroe i think i lost context....
[23:32] <lazyPower> [15:05:12] jrwren:	is there a way to juju run on every machine?  -- is the last of the log i have.
[23:33] <kwmonroe> lazyPower: read the "e.g." on that same line
[23:33] <lazyPower> hah, ok
[23:33] <lazyPower> but unit profiling is different than unit instrumentation.
[23:34] <kwmonroe> ugh.  can't you just give me a sensible chuckle and move on?  it was joke.
[23:34] <lazyPower> hehe, that was *totally* clever kwmonroe. You should quit your day job ;)
[23:34] <kwmonroe> :)
[23:39] <teranet> do we have any juju expert on here
[23:39] <lazyPower> teranet debateable ;)  What can we help you with?
[23:39] <teranet> I try to figure out why certan charms and machine's are down
[23:40] <teranet> I have a MAAS /  JUJU Enviroment which will alter has also Openstack
[23:40] <teranet> but rihgt now I try to deploy the JUJU charms and it looks to me it's somewhere hanging
[23:41] <teranet> for example is there a way to see what the spec for the a charm is like CPU . RAM DISK and so on ? Or isn't there ?
[23:42] <lazyPower> teranet - Spec's are defined as 'constraints'
[23:42] <kwmonroe> teranet: what version of juju are you running?  (output of 'juju version')
[23:42] <lazyPower> constraints aren't really modeled in teh charm, they are instead modeled in the bundle you're using to deploy the solution.
[23:43] <teranet> running juju 2.0.2 xenial
[23:44] <teranet> http://picpaste.com/5830848bfd3851694ceb8e615e676610.JPG
[23:44] <teranet> this is my status when I run juju status
[23:45] <kwmonroe> teranet: how long have those machines been in the 'down' state?
[23:45] <teranet> juju deploy cs:bundle/openstack-base-48   is what I had run
[23:46] <teranet> and for over 12h now if not even longer
[23:47] <kwmonroe> ahh.. ok.  i was going to say "give it a minute or 2".  but *hours* means something else is happening.
[23:47] <marcoceppi> teranet: they've never seemed to have been allocated
[23:47] <marcoceppi> teranet: how mnay maas nodes do you have?
[23:47] <teranet> currently 6 all same hardware specs
[23:48] <marcoceppi> teranet: how many have been allocated in the MAAS ui?
[23:48] <teranet> all
[23:49] <teranet> all 6 are deployed with Ubuntu 16.04 LTS   24 core  32GB Ram and 479GB Disk space
[23:49] <marcoceppi> teranet: can you run `juju status --format yaml 23 25`
[23:49] <teranet> my concerns is somewhere it's hitting limits but don't know exaclty where and what
[23:50] <marcoceppi> teranet: well, machines are deployed that are no enlisted in juju
[23:50] <teranet> ok let me upload the output hold on
[23:51] <teranet> http://paste.ubuntu.com/23631220/
[23:51] <teranet> here is the output of it
[23:57] <marcoceppi> teranet: so the first block is interesting, because the machine is pending
[23:57] <marcoceppi> teranet: meaning Juju really doesn't have a machine assigned
[23:57] <marcoceppi> 25 is a failure to deploy from MAAS
[23:57] <marcoceppi> teranet: what I would do, is destroy this controller
[23:58] <marcoceppi> make sure maas returns all the machines to ready
[23:58] <marcoceppi> teranet: then deploy again
[23:58] <teranet> ouch that's some bigger work than
[23:59] <teranet> ok so how do I destroy it ?
[23:59] <marcoceppi> teranet: you've got all 6 machines as deployed in maas, but this install only sees two machines assigned to it, and one machine failed in juju's mind
[23:59] <marcoceppi> teranet: `juju destroy-controller --destroy-all-models <controller-name`