[00:38] <wallyworld> babbageclunk: for when you have a moment later, i've updated the controller facade PR. Besides a lot of cut and paste, it's a new reduction in LOC. https://github.com/juju/juju/pull/7286
[00:38] <babbageclunk> wallyworld: yay! looking soon
[01:15] <axw> wallyworld: another little one: https://github.com/juju/description/pull/14 (needed to land the MAAS storage branch)
[01:15] <wallyworld> ok
[01:17] <wallyworld> axw: don't you need to increment a version number somewhere?
[01:17] <axw> wallyworld: nah, it's an optional field
[01:17] <axw> wallyworld: old code doesn't care about it, new code doesn't care if it's missing
[01:18] <wallyworld> ok
[02:17] <axw> wallyworld: please see my comment on https://github.com/juju/juju/pull/7320
[02:22] <menn0> babbageclunk: how long would you expect the "transferring ownership" step of a migration to take?
[02:23] <babbageclunk> menn0: which cloud? Mostly quick - I think I was surprised how long Azure took to do it.
[02:23] <wallyworld> ok
[02:24] <babbageclunk> menn0: more often it would appear to take a long time because it was in a crash/retry loop
[02:24] <menn0> babbageclunk: not sure. heard about it from rogpeppe. he waited for 30mins and it didn't finish.
[02:24] <menn0> babbageclunk: i'll ask for more info.
[02:25] <babbageclunk> menn0: That's definitely a crash.
[02:25] <babbageclunk> *almost definitely
[02:25] <menn0> ok
[02:25] <wallyworld> axw: i was wondering whether the tests, which do exercise all of the Watch* methods, include enough data in the collections which would be excluded from the results, testing the newly added filtering.
[02:26] <babbageclunk> menn0: hmm, I guess they could be doing it on Azure and have lots of things to retag.
[02:26] <babbageclunk> menn0: maybe we should be doing that in parallel.
[02:30] <menn0> babbageclunk: the model had nothing deployed in it. could there be a bug with that case?
[02:33] <babbageclunk> menn0: maybe? I don't think it would be that specifically.
[02:33] <axw> wallyworld: like I said, it's 100% coverage: that includes the cases where the filter returns "false"
[02:34] <wallyworld> ok
[02:34] <axw> wallyworld: the tests might be a little non-obvious, because the "mixed" storage apps have two stores each, and there are default pools configured by the test helper
[02:34] <axw> wallyworld: by default one will be machine-scoped, one model-scoped
[02:35] <menn0> babbageclunk: i've copied you in on the email where i've asked for more details
[02:35] <wallyworld> rightio
[02:35] <babbageclunk> menn0: cool cool
[03:02] <axw> babbageclunk: weency review please? https://github.com/juju/juju/pull/7321
[03:29] <babbageclunk> axw: yup yup
[03:31] <babbageclunk> axw: approved
[03:33] <babbageclunk> wallyworld: you did an amend again. :( What's different between the version of the code I reviewed and this one? Do I need to scroll through it all again?
[03:34] <wallyworld> babbageclunk: ah sorry, our somewhat dumb policy because git sucks is to squash everything into one commit, but in this case I should have broken that rule
[03:35] <wallyworld> main thing is that grant/revoke cli now uses controller facade methods
[03:35] <babbageclunk> wallyworld: or do it after it's reviewed?
[03:35] <wallyworld> yeah, sorry
[03:35] <wallyworld> i owe you a beer or 6
[03:36] <babbageclunk> wallyworld: I haven't been squashing things - I guess I should be?
[03:36] <wallyworld> well, i don't agree that we should but that's meant to be the policy
[03:37] <wallyworld> bzr is much better than git at only showing what you want to se
[03:37] <wallyworld> otherwise it is discarding history
[03:37] <wallyworld> which i disagree with
[03:37] <wallyworld> so take an extra look at grant/revoke
[03:38] <babbageclunk> sure - it's unfortunate to lose the fine-grained commits.
[03:38] <wallyworld> and access_test.go in apiserver
[03:38] <wallyworld> it is
[03:38] <babbageclunk> ok, will do - thanks
[03:54] <babbageclunk> wallyworld: LGTM'd
[03:54] <wallyworld> you rock ty
[04:10] <jam> menn0: standup ?
[04:11] <jam> thumper?
[04:13] <menn0> jam: sorry! coming
[09:29] <rogpeppe> first stage in getting juju controllers to know their own public hostnames, review appreciated: https://github.com/juju/juju/pull/7323
[09:29] <rogpeppe> axw, menn0, babbageclunk: you might wanna take a look ^
[09:30] <axw> looking
[09:36] <axw> rogpeppe: stupid question: what's the point if you have to have the CA cert in the first place?
[09:36] <rogpeppe> axw: you don't need a CA cert
[09:37] <axw> rogpeppe: to login the first time, to get back the public IP
[09:37] <rogpeppe> axw: you can connect without the CA cert
[09:38] <axw> rogpeppe: how do you know what you're talkign to then?
[09:39] <axw> rogpeppe: is this for register?
[09:39] <rogpeppe> axw: so... currently things are quite fragile
[09:39] <rogpeppe> axw: the client is the only thing that knows the public DNS name
[09:40] <rogpeppe> axw: the controller doesn't know the name, so (for example) it can't pass it to agents
[09:40] <rogpeppe> axw: we had a problem after migrating to a public controller because the public dns name is lost
[09:41] <rogpeppe> axw: if the DNS name changes, there's no way of notifying clients (not that it *can* change right now)
[09:43] <axw> rogpeppe: the controller doesn't know the name? it's part of teh controller config (autocert-dns-name)?
[09:43] <rogpeppe> axw: after "juju register", you won't know the public host name without this.
[09:43] <axw> (fair enough re updating the client)
[09:44] <rogpeppe> axw: yeah, the controller knows that name, but doesn't let the clients know (and actually, as mhilton points out, the public name might not actually be held in autocert-dns-name, something we need to consider)
[09:45] <axw> okey dokey
[09:58] <axw> rogpeppe: LGTM
[09:58] <rogpeppe> axw: tyvm
[10:22] <rogpeppe> anyone know what the failure is here? http://juju-ci.vapour.ws:8080/job/github-merge-juju/10859/artifact/artifacts/grant.log
[10:22] <rogpeppe> axw: ^
[11:36] <rogpeppe> make clients know about public hostnames explicitly: https://github.com/juju/juju/pull/7324
[11:47] <rogpeppe> jam: looking for a review of the above if you have some time...
[14:08] <rogpeppe> some juju code garbage collection: https://github.com/juju/juju/pull/7326
[16:02] <rogpeppe> provide CACert in the juju client API: https://github.com/juju/juju/pull/7327, if anyone's arround
=== akhavr1 is now known as akhavr
[16:40] <rogpeppe> here's another juju PR that just deletes unused stuff: https://github.com/juju/juju/pull/7328
=== ejat_ is now known as ejat
=== frankban is now known as frankban|afk
=== akhavr1 is now known as akhavr
[20:54] <thumper> morning
[21:02] <stokachu> wallyworld: we've had to work to make the latest jaas stuff work https://github.com/conjure-up/conjure-up/pull/879
[21:03] <stokachu> wallyworld: i think we're still good for tomorrow, just need to make a push for libjuju and conjure-up then run some tests
[22:18] <balloons> wallyworld, thumper, axw, hml, the joyent infastructure issues are fixed. Sorry for the noise. We'll get the jobs run again and off the failures
[22:26] <thumper> balloons: awesome, what was the underlying issues BTW?
[22:36] <balloons> thumper, keys keys keys. It worked for a short time, then another update killed it. Many of our clouds use more than one key (don't ask me why). Joyent fell into the trap of requiring a fingerprint for one key, but the actual public key of the other (again, don't ask me why)
[22:37] <thumper> heh
[22:37] <thumper> ok
[23:27] <menn0> wallyworld: easy review pls? https://github.com/juju/juju/pull/7329
[23:28] <menn0> following on from one of yesterday's topics
[23:28] <wallyworld> menn0: so just removing the -S
[23:28] <menn0> yep
[23:29] <wallyworld> lgtm
[23:30] <menn0> wallyworld: cheers
[23:31]  * wallyworld disappearing for swap day, off to take kid to hospital