=== wgrant_ is now known as wgrant | ||
axw | thumper: sorry, didn't realise our 1:1 and moved forward. now conflicts with standup.. can we move it? | 01:03 |
---|---|---|
axw | and=had | 01:03 |
blahdeblah | axw: Do you know where I can find more about the fix for https://bugs.launchpad.net/juju-ci-tools/+bug/1634556 ? I'm seeing the same error trying to bootstrap in GCE on 2.2.5. | 01:57 |
mup | Bug #1634556: autoload credentials fails on google <gap> <juju-ci-tools:Fix Released by mskalka> <https://launchpad.net/bugs/1634556> | 01:57 |
blahdeblah | ^ Or anyone else who happens to be around ... | 01:59 |
axw | blahdeblah: afraid not, sorry | 02:00 |
axw | blahdeblah: based on the people involved on the bug, I'm guessing there weren't any code changes... | 02:00 |
blahdeblah | Hmmm... | 02:00 |
blahdeblah | axw: Any suggestions on how to debug this then? https://pastebin.canonical.com/201303/ | 02:04 |
blahdeblah | Connectivity to www.googleapis.com does not appear to be the problem | 02:05 |
axw | blahdeblah: the source location suggests that the project ID is invalid. are you using a downloaded JSON file for creds? | 02:06 |
blahdeblah | yup | 02:06 |
blahdeblah | Maybe they're expired or something | 02:06 |
axw | blahdeblah: I would've thought you'd get an auth error, but perhaps. can you share the JSON file (with the key redacted)? | 02:08 |
blahdeblah | sure - 1 sec | 02:09 |
blahdeblah | axw: https://pastebin.canonical.com/201304/ | 02:11 |
axw | huh | 02:12 |
axw | blahdeblah: mine has a "project_id" field, yours doesn't. the code expects one | 02:12 |
blahdeblah | weird | 02:12 |
axw | ah this is from gcloud. I don't use the gcloud one | 02:13 |
blahdeblah | ? | 02:14 |
blahdeblah | I've pulled some more credentials and they have a project id, so I'll use that | 02:15 |
axw | blahdeblah: I created my creds file via the GCP console. I guess we need to handle those project_id-less creds differently somehow | 02:16 |
blahdeblah | The other JSON file I just downloaded looks very different: {"installed":{"client_id":"413983063871-35g0qo8jt7ea1ofovpj5bhajc9sijhhq.apps.googleusercontent.com","project_id":"ubuntu-os-mirrors-dev","auth_uri":"https://accounts.google.com/o/oauth2/auth","token_uri":"https://accounts.google.com/o/oauth2/token","auth_provider_x509_cert_url":"https://www.googleapis.com/oauth2/v1/certs"}} | 02:16 |
blahdeblah | blah | 02:16 |
blahdeblah | that doesn't seem like a secret | 02:16 |
blahdeblah | or is it? | 02:17 |
blahdeblah | If so, I'd better get it deleted... | 02:18 |
blahdeblah | Yeah - looks like that's the wrong JSON file, with no secrets | 02:24 |
blahdeblah | axw: Created a new service account, it has the project_id, along with the other expected fields, like private key. | 02:24 |
blahdeblah | And trying the bootstrap manually now works | 02:25 |
axw | blahdeblah: yeah not sure what that other one is for, doesn't look very secret at all | 02:25 |
blahdeblah | axw: This is very strange - now I'm getting the following: $ juju bootstrap google my-new-gce-controller | 03:39 |
blahdeblah | ERROR invalid character '/' looking for beginning of value | 03:39 |
blahdeblah | Any ideas? | 03:39 |
blahdeblah | Google searches seem to indicate it's coming from somewhere in the golang runtime | 03:40 |
blahdeblah | possibly the JSON parser: https://stackoverflow.com/a/37492750 | 03:41 |
blahdeblah | Weird thing is, if I run the same thing from the command line, it works. So something's rotten in the state of environment variables. | 03:49 |
blahdeblah | I might be doing something wrong here, but I don't think so, so I'm going to log a bug. | 04:06 |
blahdeblah | https://bugs.launchpad.net/juju/+bug/1726226 | 04:07 |
mup | Bug #1726226: bootstrap fails: ERROR invalid character '/' looking for beginning of value <juju:New> <https://launchpad.net/bugs/1726226> | 04:07 |
axw | blahdeblah: sorry missed your message. no ideas sorry, can you please include the credential file (with key redacted) | 04:14 |
blahdeblah | axw: thanks - will add it to the bug | 04:14 |
axw | blahdeblah: is project_id really "project-id", or did you substitute? | 04:32 |
blahdeblah | I substituted | 04:33 |
blahdeblah | sorry - should have made that more obvious | 04:33 |
axw | any / in the real value? | 04:33 |
axw | blahdeblah: ^ | 04:33 |
blahdeblah | nope | 04:34 |
blahdeblah | just alphas & dashes | 04:34 |
axw | ok, ta | 04:34 |
=== frankban|afk is now known as frankban | ||
axw | wallyworld: might need to talk about these estimates, I don't actually know what all of the line items mean | 06:51 |
wallyworld | axw: i added some commnts to yours, see if they make sense | 07:41 |
axw | wallyworld: so, your estimate for space selection for agent/controller traffic... is that *including* the controller-as-app work? | 07:46 |
wallyworld | axw: yeah, so i think it's too low perhaps | 07:47 |
axw | wallyworld: I've put in some numbers where I think we need more time. the ones I've left out are either OK, or too vague/broad to give an estimate on | 07:53 |
wallyworld | ok, ty, i'll take a look | 07:53 |
wallyworld | it's all a bit habdwavy | 07:53 |
Mmike | Hi, lads - is there a (simple-ish) way to recreate juju environment once I 'juju ssh' into a unit? I can't use 'juju run' as relations are in error state and I need to verify some relation variables across the environment. So before I resort to peeking into the database I wonder can I run 'relation-ids, relation-list, ...' from the unit itself somehow? | 08:06 |
=== akhavr1 is now known as akhavr | ||
=== akhavr1 is now known as akhavr | ||
=== akhavr1 is now known as akhavr | ||
=== frankban is now known as frankban|afk | ||
babbageclunk | wallyworld: ping? | 22:04 |
thumper | externalreality: https://github.com/juju/charm/blame/v6-unstable/config.go#L62 still shows the panic... | 22:05 |
thumper | externalreality: is that now what you fixed? | 22:05 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!