[01:08] <mramm2> davecheney: sorry, I missed you at the meeting time, and then went to dinner
[01:19] <davecheney> mramm2: wanna talk now ?
[01:22] <mramm> davecheney: sure
[01:22] <mramm> I don't have much of an agenda, just touching base
[01:22] <mramm> since we talked sunday
[01:22] <davecheney> mramm: yeah, i don't have anything to add
[01:22] <mramm> cool
[01:22] <mramm> then we can skip it
[01:22] <davecheney> is wed 8:30 our regular time ?
[01:22] <mramm> and I'll let you keep up the good work ;)
[01:22] <mramm> well, not regular
[01:23] <davecheney> cool
[01:23] <mramm> I just scheduled it after my talk with Tim
[01:23] <davecheney> i'm sorta behind today agfter LP was out all yestedday
[01:23] <davecheney> no lp == no ppa == no boostrap :(
[01:23] <mramm> gotcha
[01:23] <mramm> no worries
[01:23] <davecheney> william has asked for one more load test run, with the presence code commented out
[01:23] <mramm> I sent an e-mail this evening about the blueprint stuff (to the whole team)
[01:24] <davecheney> given we have to give the load test resources back on the 3rd
[01:24] <mramm> davecheney: yea
[01:24] <davecheney> it's probably worthwhile fitting it in this week
[01:24] <davecheney> re: email, read and understood
[01:24] <mramm> davecheney: agreed
[01:25] <davecheney> jolly good, as you were
[01:25] <davecheney> mramm: one thing, re my email about tracking patches to backport
[01:25] <davecheney> i don't think that page captures it totally
[01:26] <davecheney> which means everyone is keeping their own imcomplete list
[01:26] <davecheney> which means we'll probably miss something for the next 1.10 point release
[01:26] <davecheney> which is bad
[01:26] <mramm> Yea
[01:26] <mramm> I think we should talk about this at the meeting
[01:27] <mramm> I'm not sure everything needs to be backported
[01:27] <davecheney> mramm: that is fine
[01:27] <mramm> 1.12 will probably be less than a month away
[01:27] <mramm> so it isn't a huge deal
[01:27] <mramm> but we should *know* what is being backported
[01:27] <mramm> and *know* what is not
[01:27] <davecheney> ^ exactly
[01:27] <mramm> and have *good reasons* for which is which
[01:28] <davecheney> mramm: maybe lp/bzr has good tools to do this automatically
[01:28] <davecheney> i don't konw
[01:28] <davecheney> a google doc is a lofi solution
[01:28] <mramm2> yea
[01:28] <mramm2> there is a target series thing in lp
[01:29] <mramm2> not that I know if it makes sense
[01:29] <mramm2> anyway, I think I'm going to bug out for the evening.... I'll look into the problem and see if I can find a solution to recommend before the meeting.
[01:34] <thumper> davecheney: I'm heading out to take the kids out for exercise.  I'm taking my laptop to work on blueprints/specs but won't necessarily be online
[01:35] <thumper> will have mobile
[01:36] <davecheney> thumper: no worries
[01:36] <davecheney> understood
[02:09] <davecheney> m_3: ping
[02:09] <davecheney> i'm doing another load test run
[02:09] <davecheney> fwereade: wanted to see what it looks like with the agent presence facility turned off
[03:59] <hazmat> is there a way to forcibly remove a unit, i hit another case of a wedged unit in  bug 1175031
[03:59] <_mup_> Bug #1175031: wedged unit state error but can't be resolved <juju-core:New> <https://launchpad.net/bugs/1175031>
[08:39] <rogpeppe> mornin' all
[08:52] <dimitern> guys, my network connection is rather flaky since yesterday evening - coming on and off frequently, and generally being unstable
[08:54] <rogpeppe> dimitern: welcome to the club!
[08:54] <dimitern> rogpeppe: thanks :| i hope it'll get better, but sometimes is almost unusable
[08:55] <rogpeppe> dimitern: i'm lucky that my mobile internet bandwidth is pretty good here, so i've got an alternative
[10:14] <dimitern> so, with my internet troubles and considering it's a public holiday today here, i'll be off
[10:14] <dimitern> see you all tomorrow
[10:16] <mgz> later dimiter
[10:48] <rogpeppe> fwereade: ping
[11:06] <rogpeppe> that annoying feeling when you know you've implemented something in the past, but can't remember where it is, or even what you called it.
[11:09] <rogpeppe> ha! found it
[11:31] <mgz> jam: standup?
[11:35] <jam> mgz: logging in now
[14:01] <rogpeppe> mramm, fwereade: kanban?
[14:02] <mramm> still in another meeting
[14:02] <mramm> be there in a couple of min
[18:03] <fwereade> rogpeppe, mramm: sorry, public holiday, I should have said
[18:03] <rogpeppe> fwereade: np
[18:04] <fwereade> rogpeppe, are you aware of a reason juju set doesn't use statecmd?
[18:04] <rogpeppe> fwereade: possibly. let me check.
[18:06] <rogpeppe> fwereade: possibly because it worked as well with methods on state.Service
[18:07] <fwereade> rogpeppe, hmm, ok, I might have to do something about that -- Service.SetConfigYAML doesn't feel like a sane thing at all to me
[18:09] <rogpeppe> fwereade: it SetConfigYAML may well have preceded statecmd, i'm not sure
[18:10] <rogpeppe> fwereade: hmm, no it didn't
[18:12] <rogpeppe> i wish i knew how to interpret the revnos printed by bzr blame
[18:13] <rogpeppe> fwereade: you didn't seem to think it too bad here: https://codereview.appspot.com/8626043/diff/5001/state/service.go#newcode739
[18:14] <rogpeppe> fwereade: but i'd be ok seeing it move if you do mind the churn
[18:14] <rogpeppe> s/do/don't/
[18:17] <rogpeppe> fwereade: BTW you might be interested in this. it's a JSON representation of the API: http://paste.ubuntu.com/5622552/
[18:18] <rogpeppe> fwereade: there are quite a few possible applications for the info
[18:29] <rogpeppe> fwereade: i have to jump on my bike and go back from town
[18:29] <rogpeppe> fwereade: might pop my head in later
[18:30] <rogpeppe> fwereade: if not, see ya tomoz
[21:12] <thumper> morning
[21:12] <thumper> hi fwereade (if you are still around)
[21:12] <fwereade> thumper, hey, more or less; sorry, long time no speak
[21:12] <thumper> fwereade: that happens
[21:12] <thumper> I was just thinking this morning that it had been a while since we had talked
[21:14] <fwereade> thumper, how's it going?
[21:14] <thumper> bit of a headache this morning, but not sure if that is lack of coffee or from being punched in the head
[21:15] <thumper> :)
[21:15] <thumper> it has been quite a slow week as I have had to deal with kids on school holidays
[21:15] <thumper> i'm trying to get some meat into some of the blueprints for next week
[21:17] <fwereade> thumper, ouch
[21:17] <fwereade> thumper, I am much exercised by service config setting
[21:17] <thumper> although it is the ribs that is hurting right now
[21:17] <thumper> what do you mean by service config setting?
[21:18] <fwereade> thumper, juju set
[21:18] <thumper> right, what's the problem?
[21:18] <fwereade> thumper, egregious incompatibility in --config flag
[21:18] <fwereade> thumper, but on looking at it, ow, my brain
[21:19] <fwereade> thumper, and, fair enough, the --config flag seems reasonably obvious
[21:20] <fwereade> thumper, *but* it appears to have been implemented without actually checking what python did
[21:20] <thumper> fwereade: I have some sekrit work to make jcastro happy... (I implemented switch)
[21:20] <fwereade> thumper, <3
[21:20] <thumper> fwereade: oh, and I guess that is a problem
[21:21] <fwereade> thumper, yeah -- and there's a SetConfigYAML method on Service which I reviewed with by-far-not-enough whining :/
[21:21] <fwereade> thumper, and is in hindsight profoundly wrong
[21:21] <thumper> I've been half thinking recently about things like pubsubhubbub
[21:21] <fwereade> thumper, but forgivable because the SetConfig implementation is *itself* profoundly icky, because it sets typed data as a map[string]string, and gaah
[21:22] <thumper> heh
[21:22] <fwereade> thumper, so with enough kicking that can be collapsed into a SetConfig(map[string]interface{}) that at least has the data of the right type, and validates it
[21:22] <thumper> my juju mind has been somewhat split between: logging, ha, containerisation, status announcements
[21:23] <fwereade> thumper, and with slight shame tacking the weird YAML behaviour (not at *all* obvious) onto charm.Config, and bleh
[21:23] <fwereade> thumper, I haven't really managed to settle cleanly on a high-level topic at all, I have to spend tomorrow on blueprints
[21:23] <thumper> fwereade: does it really need a "step back and reevaluate"?
[21:24] <thumper> and take a fresh look at roles and responsibilities?
[21:24] <fwereade> thumper, the config stuff?
[21:24] <thumper> yeah
[21:25] <fwereade> thumper, it was somehow implemented wrong and without tests, so I kinda needed to fine-tooth comb it to figure out what was going on
[21:25] <thumper> ouch
[21:25] <fwereade> thumper, it seems to be nonstandard in several ways
[21:26] <fwereade> thumper, each individually forgivably so, but end-to-end it's very confusing
[21:27] <thumper> it becomes a tough call
[21:27] <fwereade> thumper, that was the stuff I was trying to dump on you when you first joined, all I had then was a slight intimation of it being weird and needing a look
[21:28] <thumper> heh
[21:28] <thumper> yay, dodged a bullet there
[21:28] <fwereade> thumper, in this case I'm fairly comfortable choosing to fix rather than to layer on what feel like further hacks
[21:28] <thumper> yeah, good call
[21:30] <fwereade> m_3, ping
[21:33] <fwereade> thumper, anyway, sorry, you were saying much more interesting things
[21:33] <thumper> :)
[21:34] <thumper> I had a good chat with the BDFL of docker
[21:34] <fwereade> cool!
[21:34] <thumper> he is CEO of dotcloud
[21:34] <thumper> unfortunately he is on holiday for two weeks starting Monday
[21:34] <thumper> but is going to make sure someone from their office goes to the gophers meetup
[21:34] <thumper> docker are very keen to work with us
[21:34] <thumper> to make sure we get what we need
[21:35] <thumper> anyway, I've added quite a bit to the containerisation spec
[21:35] <thumper> and I was going to start on HA spec today
[21:35] <fwereade> awesomesauce
[21:36] <thumper> however I have some ideas that don't fit into specs
[21:36] <thumper> in particular rationalizing logging
[21:36] <thumper> and having a way to publish status changes
[21:36] <thumper> a la pubsubhubbub type thing
[21:36] <thumper> just a publish/subscribe hub
[21:37] <thumper> if we had the api server and other commands publish on the bus,
[21:37] <thumper> we could potentially have something like 'juju observe'
[21:37] <thumper> very hand wavey right now
[21:38] <thumper> as I've spent exactly 5 minutes actively thinking about it
[21:38] <thumper> rest of the time it has been quietly perculating in the hind-brain
[22:37] <bigjools_> g'day
[23:01] <hazmat> thumper, that sounds nice.. api server publishing messages, instead of state txn tail
[23:01] <thumper> hazmat: it would be nice, but I've not put too much thought into it yet
[23:01] <thumper> hazmat: we could then have a log listener that takes the events and logs them
[23:02] <thumper> hazmat: but we could also hook in other listeners
[23:02] <hazmat> thumper, yes.. aka audit trail
[23:02]  * thumper nods
[23:02] <thumper> I may also hack up a different logging implementation based on my C++ work in the past
[23:02]  * thumper has written too many logging systems
[23:03] <thumper> there would be differences, but I think we could make on that is idiomatic go without too much trouble
[23:03] <hazmat> thumper, i was trying out a few web api server broadcasting messages for a thought experiment it works well
[23:03] <thumper> I don't get what you are saying there :)
[23:03] <thumper> what were you testing?
[23:03] <hazmat> i ended up on rabbitmq.. after trying hard to with zeromq..
[23:04] <hazmat> thumper, seeing what a juju api server for internal api would look like
[23:04] <thumper> ok
[23:07] <hazmat> thumper,  i ended up using the existing juju-core api for an embeded watch/observer in a new version of juju deployer.. the client side is useful already.. embedded in the server api is nice as stat/audit/rate limit.
[23:07] <hazmat> thumper, not sure what you mean by logging impl
[23:07] <thumper> hazmat: have logging categories
[23:07] <thumper> and be able to set verbosity on a category by category level
[23:08] <thumper> with sub categories
[23:08] <hazmat> log hierarchy
[23:08] <thumper> so set juju.api to INFO, juju.uniter.upgrade DEBUG etc
[23:08] <thumper> aye
[23:08] <thumper> hazmat: I had a good one I write for unity
[23:09] <thumper> hazmat: that is lovely in C++ but won't work in go
[23:09] <thumper> as no maros, and no object lifetime
[23:09] <hazmat> i gotta imagine there's something extant already
[23:09] <thumper> so slight changes needed
[23:09] <thumper> hazmat: you'd think that...
[23:09] <thumper> but most of the existing ones are horrible
[23:09] <thumper> log4cpp was a bad port of log4j
[23:09] <thumper> and boost logging was fugly
[23:09] <hazmat> thumper, then again i would have thought the same about flocks
[23:10] <hazmat> ;-)
[23:10] <thumper> :)
[23:55] <AeroNotix> hi guys
[23:56] <thumper> s'up?
[23:56]  * thumper waves at davecheney
[23:56] <AeroNotix> https://github.com/AeroNotix/hpcloud so I've been working on these
[23:57] <AeroNotix> I'm not super familiar with the launchpad site so I can't see how complete goose is
[23:57] <AeroNotix> or what features it provides
[23:57] <AeroNotix> but it seems that we have a very similar design
[23:57] <AeroNotix> i.e. you have a central Client type which you then embed in submodules
[23:57] <AeroNotix> from what I gather, at least
[23:57] <AeroNotix> I am thinking if you would be open to me trying to merge my code base with yours
[23:57] <AeroNotix> providing you want the features it has
[23:58] <AeroNotix> I assume you're looking to bind to a lowest common set of available endpoints which most openstack providers... provide
[23:59] <thumper> AeroNotix: a good person to poke about goose is wallyworld_
[23:59] <thumper> I don't know anything about it
[23:59] <AeroNotix> paging wallyworld_
[23:59] <wallyworld_> hello
[23:59] <AeroNotix> If you could read the scroll back it'd save me a lot of typing :P
[23:59]  * wallyworld_ reads