benonsoftware | Hiya, I'm wondering if there's anyway to make juju launch spot instances? | 02:55 |
---|---|---|
thumper | o/ bigjools | 03:38 |
=== jtv1 is now known as jtv | ||
jam | morning dimitern | 06:59 |
dimitern | morning | 06:59 |
jam | dimitern: 1:1? | 07:30 |
dimitern | jam, sorry, coming | 07:34 |
jam | morning fwereade | 09:01 |
jam | rogpeppe: 1:1? | 09:01 |
fwereade | heya jam | 09:01 |
rogpeppe | jam: just going there | 09:02 |
wallyworld | fwereade: hi, i need to add a SetAddresses() api to machiner. most of the other machiner apis are plugged in via common structs since the api is applicable to machine or unit. But SetAddresses appears exclusive yo machine, even though unit does have SetPrivateAddress, SetPublicAddress. Are those latter 2 apis sticking around? Should they be refactored to SetAddresses? | 09:10 |
fwereade | wallyworld, those unit methods are due for retirement at some point, because unit addresses *themselves* are due for retirement -- a unit's address will be determined by its machine | 09:16 |
wallyworld | fwereade: that's what i was hoping :-) | 09:16 |
fwereade | wallyworld, (and likely what network it might be on, but that's a distinct issue) | 09:16 |
fwereade | wallyworld, one thing that needs to be considered | 09:17 |
fwereade | wallyworld, is how we handle identifying those networks | 09:17 |
fwereade | wallyworld, (if you look at state.Address, you will see a network field) | 09:17 |
wallyworld | fwereade: is is for the local provider which currently creates a container and we don't currently record the address | 09:18 |
fwereade | wallyworld, and it's not clear how the machine agent is to infer this | 09:18 |
fwereade | wallyworld, so in that case it's actually easy, it goes on the "private" network | 09:18 |
wallyworld | fwereade: so the use case is initially limited, but when this is fixed, juju ssh etc will work for local rpovider, since we will know the address | 09:19 |
fwereade | wallyworld, I understand why it's important | 09:19 |
fwereade | wallyworld, but I also want to keep all my machine agents acting the same way | 09:19 |
fwereade | wallyworld, and we need to figure out exactly how this overlaps with addressupdater | 09:20 |
wallyworld | sadly, we can't quite do that fully cause the address updater causes Addresses() on the provider and the local provider cannot get the container's address | 09:20 |
wallyworld | the only way i could see was for the container to write the address info to cloud storage and Addresses() would read it from there, but yuk | 09:22 |
fwereade | wallyworld, nah that's perfectly horrible, we shouldn't do that ;p | 09:22 |
fwereade | wallyworld, (hmm; lxc-attach?) | 09:23 |
wallyworld | maybe, but cannot be done for kvm outside container | 09:23 |
wallyworld | we really, really want to remove "not implemented" from Addresses() for local provder but i don't see how it is possible :-( | 09:24 |
wallyworld | sux to have to do it different for local provider | 09:24 |
wallyworld | happy to consider viable options we haven't thought of | 09:25 |
fwereade | wallyworld, hold on, given the total control we have over the situation, can we not literally carve out a range on our container network and assign the addresses ourselves | 09:25 |
wallyworld | for local provider do we really have the ability to do that? | 09:26 |
fwereade | wallyworld, I guess not if people are using their own bridge, bah | 09:27 |
wallyworld | and not if we want local provider containers to be exposed externally | 09:27 |
fwereade | wallyworld, we don't want *that* | 09:27 |
wallyworld | ok, so just access from local machine | 09:27 |
fwereade | wallyworld, but the forces there are very similar tothe forces involved in exposing non-local-provider containers | 09:28 |
fwereade | wallyworld, so I really think we need to come up with some model that really covers all the cases, at the moment it feels like there's a big fuzzy boundary somewhere around containers | 09:28 |
wallyworld | i think so yeah | 09:28 |
wallyworld | i'd be happy to defer any work to the folks working on container addressibility | 09:29 |
fwereade | wallyworld, haha :) | 09:30 |
fwereade | wallyworld, I'm really keen on you doing this stuff, but it would be good if you could sync up on it with mgz and/or dimitern | 09:30 |
wallyworld | if it ovrlaps greatly with non local provider containers, no poin me doing it | 09:31 |
wallyworld | andrew is actually doing it too | 09:31 |
fwereade | wallyworld, well, that's what I want to discover, possibly we can chat about it at standup time? | 09:31 |
dimitern | wallyworld, we haven't started actual work on the containers yet, just preliminary prereqs, like changes to goose (goamz to follow) | 09:31 |
wallyworld | ok, if i'm still awake, i'm in nz :-) | 09:32 |
fwereade | wallyworld, my suspicion is that there is a chunk of work that could be done by either group and that it's important to figure out what that chunk is and who's best placed to do it | 09:32 |
wallyworld | yeah | 09:32 |
fwereade | wallyworld, it might alternatively just be a matter of defining exactly how juju builds up a picture of the network by paying attention to *both* agent- and provider-supplied addresses | 09:33 |
fwereade | wallyworld, that might be best of all, even | 09:33 |
wallyworld | i think andrew was experimenting with doing something with jujud | 09:34 |
fwereade | wallyworld, oh yes? | 09:34 |
fwereade | wallyworld, (btw, I hope the sprint is going nicely) | 09:34 |
wallyworld | jujud runs inside container, gets addresses, calls machiner.SetAddresses() | 09:34 |
wallyworld | yeah it is, lots of talking today, plus hacking | 09:35 |
wallyworld | fwereade: if you wanted to discuss, perhaps we can have a chat before standup with whoever, so i can get some sleep? | 09:36 |
fwereade | wallyworld, dimitern: that sgtm -- I'd like mgz involved though, so please sleep whenever you need: I think I can represent your interests adequately ;) | 09:38 |
wallyworld | ok :-) | 09:39 |
wallyworld | fwereade: if i'm not around, can you pop me a quick note so we can see what the outcome is? | 09:39 |
fwereade | wallyworld, will do | 09:39 |
wallyworld | i'll try and stay | 09:40 |
mgz | morning | 10:03 |
jam | mgz: mumble? | 10:06 |
mgz | it's not happy for some reason.... | 10:07 |
jam | mgz: g+ works fine for me, so whatever works for you | 10:07 |
mgz | need new mumble version it seems, updating | 10:07 |
mgz | jam: okay, I'm on | 10:09 |
eagles0513875_ | hey fwereade :D dimitern | 10:26 |
dimitern | eagles0513875_, hey man! :) | 10:27 |
eagles0513875_ | dimitern: missed you at the meeting on saturday :) | 10:27 |
dimitern | eagles0513875_, well, it would've been a bit difficult to attend - i'm back home in sofia | 10:27 |
eagles0513875_ | dimitern: well we have it recorded. :) | 10:28 |
eagles0513875_ | dimitern: hows the weather there | 10:28 |
dimitern | eagles0513875_, getting better despite the forecast - 12-14ish and sunshine | 10:28 |
eagles0513875_ | cooler then here for sure | 10:29 |
eagles0513875_ | sunny and warm here already | 10:29 |
dimitern | i know :) don't tease me hehe | 10:29 |
eagles0513875_ | dimitern: not trying to if its this warm now I'm predicting a super hot summer | 10:32 |
dimitern | eagles0513875_, and diving! | 10:32 |
eagles0513875_ | i dont dive but this is off topic for this channel :) | 10:33 |
eagles0513875_ | i would like to start contributing to juju and I would like to work to potentially and eventually try my luck at becoming part of the juju dev team | 10:33 |
dimitern | eagles0513875_, yep | 10:33 |
eagles0513875_ | what would be a good bug for me to start off with | 10:33 |
dimitern | eagles0513875_, great! | 10:33 |
dimitern | eagles0513875_, well, there are some docs to read, in the source /doc | 10:34 |
dimitern | eagles0513875_, you can pull the source from lp:juju-core with bzr and try setting up so you can run the unit tests locally | 10:34 |
eagles0513875_ | already have core installed but don't have my linux laptop atm | 10:35 |
dimitern | eagles0513875_, and before all, you'll need to get golang packages and setup your GOPATH | 10:35 |
eagles0513875_ | ok | 10:35 |
* eagles0513875_ taps fingers patiently on the desk waiting for my mom to come back with the car and my laptop lol | 10:35 | |
dimitern | eagles0513875_, you can use virtualbox or something similar and run ubuntu in a vm | 10:35 |
eagles0513875_ | dimitern: in the case of my laptop i would prefer taking advantage of the full performance it can give me | 10:36 |
eagles0513875_ | dimitern: im also thinking of working with my vps provider to allow ubuntu users to easily use juju on their vps's | 10:37 |
dimitern | eagles0513875_, that's a great idea, please do | 10:38 |
eagles0513875_ | :) | 10:39 |
jam | rogpeppe: fwereade, mgz: standup ? | 10:46 |
mgz | I joined just now, and there was no one there... | 10:47 |
frankban | core-devs: morning all. I saw "juju switch" in trusty just returns the default environment name. The output no longer is "Current environment: x". Should we consider this the new behavior? Should we expect this change in saucy too? | 11:08 |
TheMue | frankban: it is the change based on https://bugs.launchpad.net/juju-core/+bug/1193244 | 11:25 |
_mup_ | Bug #1193244: juju env could be friendlier to scripts <improvement> <juju-core:Fix Released by themue> <https://launchpad.net/bugs/1193244> | 11:25 |
TheMue | frankban: is in with 1.17 | 11:26 |
frankban | TheMue: yes I know. I saw that juju-core is in trusty/universe. the juju stable ppa for previous series is still on 1.16.5. So I guess this change will also land there, right? | 11:29 |
TheMue | frankban: hmm, here I don't know the planning, sorry | 11:34 |
frankban | TheMue: ok thanks | 11:35 |
dimitern | jam, reviewed https://codereview.appspot.com/54270043/ | 11:42 |
* dimitern is afk for a while | 11:44 | |
eagles0513875 | fwereade: ping | 16:12 |
eagles0513875 | hey fwereade | 16:44 |
fwereade | eagles0513875, linked here for the convenience of any other potential lurking contributors: http://bazaar.launchpad.net/~go-bot/juju-core/trunk/view/head:/CONTRIBUTING | 16:46 |
eagles0513875 | :D | 16:46 |
dimitern | i have proposed a long chain of mostly simple CLs which will lead to the the firewaller server-side API soon, if anyone can take a look I'd appreciate it: (in order) https://codereview.appspot.com/54620043/, https://codereview.appspot.com/52840044/, https://codereview.appspot.com/54630043/,https://codereview.appspot.com/54640043/, andhttps://codereview.appspot.com/54650043/ | 17:53 |
eagles0513875 | dimitern: i have a crazy idea I would like to run by you if you dont mind | 18:06 |
dimitern | eagles0513875, sure | 18:07 |
eagles0513875 | dimitern: im thinking of potentially making linux desktop users that work in cloud environments a GUI interface that interfaces with juju to make deployments easier instead of needing command line | 18:07 |
eagles0513875 | what do you think | 18:07 |
dimitern | :) | 18:08 |
dimitern | eagles0513875, have you seen Juju GUI ? | 18:08 |
dimitern | eagles0513875, https://jujucharms.com/ for a demo | 18:08 |
eagles0513875 | dimitern: yes Its good but lacking i think | 18:10 |
eagles0513875 | dimitern: what i was thinking was more of an application one installs on their laptops and can manage deployments that way | 18:10 |
dimitern | eagles0513875, well, it's open source, you can fork it and change it as you need | 18:10 |
eagles0513875 | good point | 18:11 |
eagles0513875 | dimitern: i want to do work that would be beneficial to juju as a whole as well as potentially allow me to get hired and join the team :) | 18:11 |
dimitern | eagles0513875, and what you're talking about might be achieved with a "web application" that runs the gui in a browser frame, but has an icon and a setup | 18:11 |
eagles0513875 | dimitern: there is one problem lets say your internet to the outside is down | 18:11 |
eagles0513875 | then what | 18:11 |
dimitern | eagles0513875, sure, i understand | 18:12 |
eagles0513875 | very hypothetical situation in a datacenter environment but y ou are kinda screwed when it comes to deplooyments | 18:12 |
dimitern | eagles0513875, you'll need internet to interact with a juju environment anyway | 18:12 |
eagles0513875 | good point | 18:12 |
dimitern | eagles0513875, even if it's a private cloud, unaccessible from outside public | 18:12 |
eagles0513875 | im having fun with the local provider on my laptop | 18:13 |
dimitern | eagles0513875, you still need a connection to it, and can deploy the gui there | 18:13 |
eagles0513875 | is there a way though to shutdown a machine and any instances | 18:13 |
dimitern | eagles0513875, sweet! does it work? :) | 18:13 |
eagles0513875 | yes :) | 18:13 |
eagles0513875 | i found a corner case regarding ssh though | 18:13 |
eagles0513875 | it seems though the local provider and im not sure about any other cloud providers that you have to have the .pub ssh key in the .ssh folder | 18:13 |
eagles0513875 | with out it it gives an error | 18:14 |
dimitern | well, if you use juju destroy-machine XXX in a local environment, that will shutdown the container | 18:14 |
eagles0513875 | then i would need to destroy the environment then | 18:14 |
dimitern | you'll need A ssh key somewhere, if it's not in ~/.ssh/, you can specify it in the environments.yaml | 18:14 |
dimitern | no, you can destroy individual machines, without taking down the whole env | 18:15 |
eagles0513875 | dimitern: this is on my laptop what happens if i need to shut it off | 18:15 |
eagles0513875 | will my system take down the machine as well and bring it back up on boot up | 18:15 |
dimitern | just so in a local env, any "machine" you can see is actually an lxc container running on your machine (machine 0 is the only "real" one) | 18:15 |
eagles0513875 | dimitern: O_o what cropped up for me is this. i dont have my .pub in the .ssh folder on my local machine but on a remote server in its .ssh which tbh i dont see why you really need your .pub key if you are deploying to other things besides the local provider | 18:16 |
eagles0513875 | ahh ok :) | 18:16 |
dimitern | hmm.. that's a fair question - I think there was some work done to restart any containers at boot time | 18:16 |
dimitern | you can certainly try rebooting without destroying you local env and see if it comes back up after | 18:17 |
dimitern | i think it should (we're using local upstart jobs for the agents) | 18:17 |
eagles0513875 | ok :) | 18:18 |
eagles0513875 | im guessing fwereade has gone to eat so ill have to continue on the ssh issue i managed to unearth tomorrow with him | 18:18 |
eagles0513875 | dimitern: btw mlug has our own irc channel lug-mt | 18:19 |
dimitern | eagles0513875, ah, 10x - i'll try lurking there from time to time :) | 18:20 |
eagles0513875 | no problem | 18:20 |
eagles0513875 | dimitern: how can i deploy to my local provider using the juju charms site | 18:26 |
eagles0513875 | or do i need to deploy juju-gui in a container on my laptop | 18:26 |
dimitern | yeah, you'll need to deploy the juju-gui charm in your local env, and then use it to deploy stuff | 18:27 |
dimitern | i think you need to expose it as well, so you can access it through a browser | 18:27 |
dimitern | check the juju-gui docs if in doubt | 18:27 |
eagles0513875 | dimitern: ok :) i think the example will gave it was exposed to the internet :) | 18:28 |
eagles0513875 | dimitern: i have a question though the install of the os in regards to 12.04 is it a full install or a minimal install of ubuntu installing only features that are needed | 18:28 |
dimitern | which os install? | 18:29 |
dimitern | in a container? | 18:29 |
eagles0513875 | yes | 18:29 |
eagles0513875 | of precise | 18:29 |
eagles0513875 | i think over all deployment speeds can be improved by using a minimal install no? | 18:30 |
dimitern | it just acts like precise, but it uses the same kernel as your host (which appears as precise) | 18:30 |
dimitern | it's a minimal system | 18:30 |
eagles0513875 | ahh ok | 18:30 |
eagles0513875 | seems a bit slow to setup the containers etc | 18:30 |
eagles0513875 | even when william was deploying to hpc or ec2 | 18:30 |
dimitern | initially yes, but I think we cache some stuff | 18:30 |
eagles0513875 | oh ok :) | 18:31 |
dimitern | but don't take my word for it - try it out (I think there were plans for speeding up local env deployments, but i'm not 100% sure we did it yet) | 18:33 |
dimitern | i'm signing off for today | 18:35 |
dimitern | g'night all | 18:35 |
fwereade | eagles0513875, fwiw, we need the public key to authorize on the instances we deploy, so you can ssh to them if you want to | 18:44 |
wallyworld_ | fwereade: hi, you around? | 20:03 |
thumper | o/ fwereade | 20:12 |
thumper | fwereade: we need coffee | 20:12 |
thumper | wallyworld_, axw and thumper afk to get it | 20:12 |
thumper | back | 20:20 |
fwereade | wallyworld_, thumper, hey guys | 20:42 |
thumper | fwereade: hey | 20:43 |
wallyworld_ | yello | 20:43 |
fwereade | thumper, wallyworld_, how's it going? | 20:43 |
wallyworld_ | did you want to talk about azure? | 20:43 |
wallyworld_ | peachy | 20:43 |
wallyworld_ | now that we have had coffee | 20:43 |
fwereade | wallyworld_, yeah, we probably should | 20:43 |
fwereade | I'm slowing down pretty badly myself | 20:43 |
wallyworld_ | should != want :-) | 20:43 |
wallyworld_ | we can do it tomorrow | 20:43 |
wallyworld_ | we have enough work to do today | 20:43 |
thumper | yeah, getting the proxy stuff done today hopefully | 20:44 |
fwereade | you know what, I think I'll take that as an answer and go to bed | 20:44 |
wallyworld_ | ok :-) | 20:44 |
fwereade | you rock guys :) | 20:44 |
wallyworld_ | i might see you for the standup, if not tomorrow :-) | 20:44 |
wallyworld_ | nighty night | 20:45 |
thumper | fwereade: ack, sleep well | 20:45 |
thumper | axw, wallyworld_: review plz https://codereview.appspot.com/54720043/ | 22:28 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!