redir | good good | 00:02 |
---|---|---|
babbageclunk | Anyone around for a review? https://github.com/juju/juju/pull/6731 | 01:21 |
=== frankban|afk is now known as frankban | ||
=== deanman_ is now known as deanman | ||
=== deanman is now known as deanman_ | ||
perrito666 | Morning | 09:22 |
perrito666 | Babbageclunk still need that review? | 09:22 |
=== deanman_ is now known as deanman | ||
perrito666 | brb errands | 13:02 |
hoenir | could someone please review my patch ? https://github.com/juju/juju/pull/6523 | 14:21 |
rick_h | macgreagoir: can you peek at ^ and try to get natefinch or katco to second when they're around? | 14:30 |
macgreagoir | rick_h: ack | 14:30 |
aisrael | Does juju 2 support the use of lxc copy (cloning a base container)? | 15:20 |
=== deanman is now known as deanman_ | ||
natefinch | aisrael: I don't think so. Our philosophy was to let lxd maange everything itself | 16:25 |
aisrael | natefinch: Ok. Here's the scenario: I'm looking to see if the base ubuntu image used when launching machines can be customized, i.e., pre-installing packages that will be needed, setting up proxy detection, etc. The goal is to get a faster start-up time for local development. | 16:32 |
natefinch | aisrael: so, I believe you can fool juju/lxd by making an image called ubuntu-xenial (or whatever series) and then fix that up to be what you want | 16:34 |
aisrael | natefinch: Ok, awesome, let me give that a try. Thanks! | 16:34 |
natefinch | aisrael: also, for local development, manual provider is even faster and easier. There's no machine startup and you can prepopulate easily. | 16:35 |
aisrael | natefinch: manual instead of lxd provider? Hmm | 16:36 |
natefinch | aisrael: yeah. You can still use lxd to spin up the machine (if you want), but then just use it as a manual machine. | 16:36 |
aisrael | natefinch: That's an interesting approach. I'd be interested in benchmarking the two for comparison. | 16:40 |
natefinch | aisrael: I'd love to see benchmarks. | 16:53 |
=== deanman_ is now known as deanman | ||
=== frankban is now known as frankban|afk | ||
deanman | aisrael, If you manage to fool juju to use a local lxc image please give a shout. I did try the other day to copy an existing image and giving it alias like ubuntu-xenial but somehow i failed. | 17:54 |
aisrael | deanman: will do! | 18:57 |
frobware | natefinch: ping | 19:12 |
=== mup_ is now known as mup | ||
babbageclunk | perrito666: thanks for the review! it was a bit lonely here yesterday. | 20:22 |
babbageclunk | perrito666: Also, we didn't get to have that discussion about squashing commits at the sprint. We should! I think you're right, not being able to bisect is a problem. | 20:23 |
perrito666 | babbageclunk: sorry I left early | 20:28 |
perrito666 | babbageclunk: of course I am right, that usually is the case :p | 20:29 |
babbageclunk | perrito666: :) | 20:29 |
perrito666 | mm, I believe I found a rather ugly bug in our logic | 20:31 |
babbageclunk | perrito666: oh dear | 20:37 |
babbageclunk | perrito666: do you know anything about macaroons and authentication? I've been bumping my head on a bug but it might be something obvious to someone more familiar with it. | 20:41 |
=== deanman is now known as deanman_ | ||
petevg | Hi, everyone. I'm running into an interesting issue while testing matrix: the landscape bundle in the store deploys successfully when deployed w/ the command line client, but it fails when deployed via the gui, or by python-libjuju (the latter is used in the mtarix). | 21:12 |
petevg | Specifically, the config changed hook in haproxy fails, with a KeyError getting the "services" | 21:13 |
petevg | *getting "services" from the config. | 21:13 |
petevg | In the bundle.yaml, "services" is an empty string. Does anyone know if the command line client does anything special with falsey config values? | 21:13 |
petevg | Possibly something that the gui and python-libjuju should replicate? | 21:13 |
natefinch | babbageclunk: I know a tiny bit about macaroons | 21:18 |
perrito666 | babbageclunk: nope, but we can try to think together | 21:30 |
babbageclunk | perrito666, natefinch: yay thanks! I'm working on bug 1650451 | 21:33 |
mup | Bug #1650451: Migration silently fails when performed by a newley registered and granted super user <juju:New> <https://launchpad.net/bugs/1650451> | 21:33 |
babbageclunk | What seems to be happening is that there are normal API requests to the target controller that succeed (so must have the right macaroons)... | 21:34 |
babbageclunk | And then the websocket connection to the logtransfer endpoint fails with "cannot get discharge: interaction not possible" | 21:35 |
natefinch | do the previous actions need superuser creds? | 21:37 |
natefinch | I wonder if it's just a timing issue | 21:37 |
babbageclunk | natefinch: yeah, everything on the migrationtarget facade requires superuser | 21:40 |
babbageclunk | natefinch: maybe? It happens reliably though. | 21:40 |
babbageclunk | natefinch: I think thumper had to fix something in the connection to the target api for a similar scenario (maybe the same test). But I haven't been able to find his change, and he's on holiday. I might ping him on alternative channels to see if he can give me a quick pointer. | 21:45 |
natefinch | babbageclunk: so, the last login I see before migration starts starts here: 22016-12-16 01:44:21 TRACE juju.rpc.jsoncodec codec.go:120 <- {"request-id":1, | 21:53 |
natefinch | that's the response to the login, which returns some data, which seems relevant: "controller-access":"superuser","model-access":"" | 21:53 |
natefinch | I wonder if the lack of model access is a problem. It would be interesting to see if that access field is different in a case where the user is a superuser to start | 21:54 |
natefinch | I would expect that controller superuser access would just override needing model access, but I don't know for sure | 21:54 |
babbageclunk | natefinch: yeah, that makes sense - it should work in the case where the user doesn't have model access, but maybe the auth in logtransfer isn't getting that right. | 21:55 |
natefinch | I gotta run to make dinner, but I'll be back on later. | 22:00 |
babbageclunk | natefinch: ok, thanks! | 22:01 |
mup | Bug #1651260 opened: landscape bundle error when deployed via gui (KeyError in config changed hook in haproxy charm) <juju-core:New> <https://launchpad.net/bugs/1651260> | 22:43 |
babbageclunk | perrito666: ok, what about bug 1650251 | 23:16 |
mup | Bug #1650251: Model migrations fail if cloud names don't match <model-migration> <juju:New> <https://launchpad.net/bugs/1650251> | 23:16 |
perrito666 | babbageclunk: cannot open the bug for lack of bw | 23:21 |
perrito666 | bootstrapping a vmware machine is very bw intensive | 23:22 |
perrito666 | bbl when this is finished and I dont have 3 seconds of lag for each time I hit enter | 23:22 |
babbageclunk | perrito666: no worries - I was about to type my findings and current thinking to you, but then realised I should put them on the bug instead of in irc | 23:23 |
redir | perrito666: babbageclunk are we doing standup? | 23:45 |
babbageclunk | redir: I thought we weren't but I can! | 23:50 |
* redir shrugs it's on my calendar still. but that might just be mine | 23:50 | |
perrito666 | I also thought we werent | 23:55 |
perrito666 | also, I have been uploading a vmware image for the past 50 mins and it would seem it will take another 50 mins | 23:56 |
perrito666 | so no chance I can do a video call | 23:56 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!