[00:17] SpamapS, no re test broken, it was an lxc provider test that broke, fix queued for the next branch to merge [00:23] hazmat: Ok cool I think i mis-copied/pasted that test that failed.. the lxc provider tests are the ones failing for me. [01:45] <_mup_> juju/local-unit-deploy r390 committed by kapil.thangavelu@canonical.com [01:45] <_mup_> update machine agent to use global settings to determine provider type [01:57] <_mup_> juju/trunk r360 committed by kapil.thangavelu@canonical.com [01:57] <_mup_> merge local-unit-deployment [r=niemeyer][f=855146] [01:57] <_mup_> Deploys service units on the local machine with lxc. [02:02] niemeyer, interesting zk competitor, experimental http://www.osrg.net/accord/ [03:37] hazmat: Pretty interesting indeed [03:57] hazmat: Wow, and it has transaction support [04:00] On the other hand a few other areas are on the weak side [08:06] hi all [08:52] hi all [12:23] Morning all [14:05] We have an empty review queue! [14:50] <_mup_> juju/go r7 committed by gustavo@niemeyer.net [14:50] <_mup_> Merged the go-formula-config branch [r=hazmat] [14:50] <_mup_> [14:50] <_mup_> This is the initial support for parsing of config.yaml in the Go port. [14:51] <_mup_> juju/go r8 committed by gustavo@niemeyer.net [14:51] <_mup_> Merged go-formula-config-validation branch [r=hazmat] [14:51] <_mup_> This complements the initial config.yaml handling with validation support. [15:02] <_mup_> juju/go-formula-dir r22 committed by gustavo@niemeyer.net [15:02] <_mup_> Bundle filepath.Rel as filepath_Rel while the submission upstream [15:02] <_mup_> doesn't go in (http://codereview.appspot.com/4981049). [15:05] <_mup_> juju/go r9 committed by gustavo@niemeyer.net [15:05] <_mup_> Merged go-formula-dir branch [r=hazmat] [15:05] <_mup_> This introduces support for handling formula directories in the Go port. [15:05] <_mup_> This includes the bundling of them, needed for the store. [15:08] <_mup_> Bug #858267 was filed: Bundling and unbundling must support perm bits in the Go port < https://launchpad.net/bugs/858267 > [15:10] <_mup_> juju/go r10 committed by gustavo@niemeyer.net [15:10] <_mup_> Merged go-formula-bundle branch [r=hazmat] [15:10] <_mup_> This introduces support for formula bundle files in the Go port. [15:10] <_mup_> There's an unhandled review point in this branch regarding formula bits [15:10] <_mup_> that will be addressed in a follow up branch coming soon. The problem [15:11] <_mup_> was described in bug #858267. [15:22] <_mup_> juju/go r11 committed by gustavo@niemeyer.net [15:22] <_mup_> Applied the juju/charm renaming to the Go code base. [15:22] <_mup_> Dropped need for the silly "header" field in metadata.yaml. [15:28] <_mup_> Bug #858282 was filed: support for wireless USB modems < https://launchpad.net/bugs/858282 > [17:12] niemeyer, the lxc-library-clone stuff is ready for another look [17:12] niemeyer, awesome re review queue [17:12] hazmat: Awesome, thanks! [17:12] niemeyer, also i wanted to talk about the placement stuff [17:12] hazmat: Sure [17:12] * hazmat is preparing to be in airports for 28hrs starting in a few [17:12] ugh [17:12] hazmat: Ugh indeed [17:15] hazmat: Do you want to talk now about it? [17:15] niemeyer, sure [17:16] niemeyer, so the get_placement_policy on the provider is to ensure that the provider gets final choice on the placement policy.. the user currently has two options for selecting the policy, environment, command line, and then failing that most providers default to unassigned except local which always picks local [17:17] the 'preference' value is basically the user's cli preference if specified, but the provider is responsible for returning the actual value, which is why it delegates there and for the base provider implementation does if preference: return preference [17:18] else the base provider consults the environment, and failing that returns the default [17:18] ah ic [17:19] so the reason to not push the if preference is none check to the calling site, is that some providers (local) only support a single policy [17:20] hazmat: Still, the way we figure the actual preference feels a bit convoluted [17:20] hazmat: The provider shouldn [17:20] 't be responsible for analyzing the user preference and failing out [17:22] hazmat: The current implementation also makes this more obvious since all it's doing is returning the provided value without any analysis [17:23] hazmat: I suggest having a get_placement_policies() in the provider interface instead, ordered by the provider's preference [17:23] niemeyer, but then the user can't do a cli selection [17:23] hazmat: Why not? [17:24] niemeyer, un moment on phone with bcsaller [17:24] hazmat: Ok.. will get some coffee meanwhile.. [17:25] niemeyer, so it should be doing some analysis at least to validate the user option [17:25] and it does when it goes to place the unit [17:26] niemeyer, i don't really understand your suggestion, doing an intersection against the provider policies to the user preference at the call site? [17:27] ie.. get_supported_placement_policies on the provider, intersect to the user preference and environment section at the call site? [17:27] hazmat: The user selection shouldn't be a "preference" [17:28] hazmat: Either it is supported by the provider, in which case it should be honored [17:28] hazmat: or it's not supported, in which case the request should error out [17:28] hazmat: DOing this is the responsibility of the command line management [17:28] hazmat: Not the provider [17:28] hazmat: The provider simply has a list of supported placement policies [17:29] hazmat: Which the user request must be compared against [17:29] niemeyer, sounds good [17:29] hazmat: We don't even need to default to unassigned [17:29] hazmat: If the user request is None, pick the first entry from the provider's list [17:29] which will generally be unassigned [17:29] i also think the placement stuff shouldn't be in state, but just in machine or provider package, but that's a different topic [17:30] niemeyer, okay.. i'll restructure that a bit then [17:30] hazmat: That's tricky.. either we support a user selection when the provider has multiple options, or we don't [17:30] hazmat: If we do, we need to store the option somewhere [17:31] <_mup_> juju/unittests r8 committed by jim.baker@canonical.com [17:31] <_mup_> Use proper path for build between butler and churn [17:31] hazmat: I'm on the unix permission stuff, btw [17:31] Fixing Go's archive/zip [17:31] Well.. improving [17:31] There's no bug.. just doesn't support it yet [17:32] niemeyer, so current local-dev plan is bcsaller's going to do some work merging the two endpoints (lxc-library-clone w/ local-provider-config), and add support for cloning containers into local unit deploy. i'm going to use that as a base for doing the public/private address stuff in the unit [17:32] hazmat: Sounds awesome [17:32] cool [17:33] yeah, applying updates relative to branch evolution since the sprint [17:34] bcsaller: Heyo [17:34] :) === medberry is now known as med_out [17:40] my son's first web game.. http://src.objectrealms.net/kaleb/web-asteroid-defense/ [17:40] cute [17:43] <_mup_> juju/unittests r9 committed by jim.baker@canonical.com [17:43] <_mup_> Clean pyc before running unittests [17:50] hazmat: Indeed, quite neat [17:51] hazmat: Good trigonometry lessons I bet :) [18:09] bcsaller, can you push, and let me know the name of the branch that merges [18:12] hazmat: I haven't tested it all yet, but I can push it [18:13] bcsaller, sounds good, i just want to have the origin of my merge point to have your branch an ancestor [18:13] i'm going to be heading out in about 30m, and will have spotty connectivity till i get to the sfo lounge [18:13] in about 6hrs [19:50] kim0: how to set juju using micro instance .. and changing the region [19:53] hazmat: still around? [20:08] ejat: in .juju/environments.yaml default-instance-type: t1.micro [20:08] ejat: it needs to be indented at the same level as 'type: ec2' [20:09] ejat: to change the region, do 'region: xxxxx' [20:13] default-instance-type: t1.micro type: ec2 [20:13] mean like that ? [20:14] before or after the type ? [20:14] region in which line ? [20:19] Doesn't matter where, just that its indented the same [20:19] owh okie [20:19] environments: [20:19] sample: [20:19] type: ec2 [20:19] environments: [20:19] sample: [20:19] type: ec2 [20:19] oops [20:19] ? [20:20] ejat: so if that were the first few lines of your environments.yaml .. then anywhere after sample: you just add [20:20] okie [20:20] default-instance-type: t1.micro [20:20] Note that once you deploy something, the service will always have that instance type even if you change it and do add-unit [20:24] owh okie .. this is for the testing purpose .. [20:25] but if without it .. it will default using small instance right ? [20:25] yeah [20:34] WARNING: Charm is using obsolete 'str' type in config.yaml. Rename it to 'string'. [20:34] is it from the charm ? [20:34] mediawiki charm [20:43] SpamapS: i cant view my mediawiki after i lunch it .. is it because of the security group policy ? [21:07] ejat: yes [21:07] ejat: you need to expose it [21:07] ejat: juju expose name-of-mediawiki-service [21:07] owh .. okie [21:07] ejat: if any charms in your repo have 'str' instead of 'string' you will get that warning. Its one of my TODO's to fix them all.. feel free to submit merge proposals :) [21:07] i just destroy the environment .. means that .. we need to expose the services ? [21:08] ejat: sorry, what? [21:08] ill try later ... sleepy .... :) [21:08] SpamapS: will ya be at UDS ? [21:09] u need to fix all the charm ? [21:09] ejat: yes I'll be at UDS.. quite a few of us will be. Are you coming? [21:10] ejat: yeah we need to fix all the charms that use 'str' to use 'string' [21:10] but warnings < failures ... so failures first.. :) [21:10] hopefully .. .. need to renew my passport n make the visa .. [21:11] ejat: its in just over 1 month.. do it *soon* [21:11] SpamapS: yeah .. thanks for the reminder .. a lots of todo starting next week need to be done .. [21:13] coz . its my 1st time joining the uds .. [21:13] * ejat a lot of thing need to be learn n catch up .. [21:43] bcsaller, i'm here [22:16] <_mup_> juju/provider-determines-placement r392 committed by kapil.thangavelu@canonical.com [22:16] <_mup_> add an error for invalid placement policies [23:04] http://paste.ubuntu.com/696384/ [23:04] anyone can help me?