[06:21] morning davecheney [06:22] fwereade: hello [06:22] thank you for your reviews [06:22] * davecheney is still pondering the theological implication of juju set KEY="" [06:22] davecheney, a pleasure, sorry they've been a little vague [06:22] davecheney, yeah, that is definitely a pain point... [06:23] davecheney, but, have you been reading the "incompatible charm upgrades" thread? [06:23] fwereade: not closely [06:24] reading the python juju set [06:24] it's a lot more complicated than KEY=VALUE [06:24] davecheney, I *think* what that should always mean, based on clint's point re fallback-to-defaults, is "clear out the setting, use the service default instead" [06:24] it's actually "key: value" _when_ the charm is a yaml charm [06:24] and other options when the charm is something else [06:24] davecheney, yeah, this is something I had not fully appreciated the intricacies of myself [06:25] davecheney, sorry, what do you mean by "yaml charm"? [06:25] charm_format = (yield charm.get_metadata()).format [06:25] formatter = get_charm_formatter(charm_format) [06:25] options = formatter.parse_keyvalue_pairs(service_options) [06:25] davecheney, oh, hell [06:25] (insert dramatic chipmonk) [06:25] haha [06:26] however get_charm_formatter does allow for something called "PythonFormat" [06:26] davecheney, I'm reluctant to claim possession of truth here, but please talk to niemeyer about this... I'm *sure* I saw him saying something to jimbaker yesterday that strongly implied that was not the correct approach on the client side [06:26] T [06:27] i'd suggest it's pretty crack for the user of a charm to have to specify the config data in the native charm format [06:27] given we don't make it obvious what that format is [06:29] davecheney, I *think* (again) that there was a recent thread which laid this out... or possibly just made reference to an earlier thread which did [06:29] ill search the archive [06:58] davecheney, fwereade: morning! [07:00] rog, heyhey [07:01] rog, PATH change is merged [07:01] rog, is it just the security groups now? [07:01] fwereade: i've just seen that. will pull and see what happens in my branch... [07:01] rog, (that we currently know of ;)) [07:01] fwereade: no, they're in [07:02] fwereade: it should just work now [07:02] rog, ah, cool, I didn't try because https://code.launchpad.net/juju-core/+activereviews still shows it aproved but unmerged [07:02] fwereade: although of course you won't be able to *tell* that it's working - the uniter-upgrade branch will do that [07:02] hello lads [07:02] rog, oh? what does that do [07:03] rog, well, obviously it upgrades the uniter [07:03] fwereade: good point - i'd moved it in to my "submitted" list, but it seems i never did [07:03] rog, but... can't we tell anyway? [07:03] fwereade: i suppose so, from the unit status [07:03] rog, via status [07:03] rog, yeah [07:04] fwereade: will submit the security groups stuff pronto [07:04] rog, cool [07:05] fwereade: ah, no, it *is* submitted, but i had to remake the merge proposal because the prereq changed [07:05] fwereade: i'll add a comment to the original [07:09] rog, ah cool [08:13] rog, dammit, it's the addresses again [08:13] fwereade: ah, sorry, that fix is in the unit-upgrader branch [08:14] rog, ahh that's why :) [08:14] fwereade: i should push it out as a trivial [08:14] rog, do it do it :) [08:14] fwereade: although it's not *entirely* trivial - there's the arguable case over "latest" vs some fixed version number [08:15] fwereade: which others might wish to weigh in on [08:15] davecheney: ping [08:15] fwereade: i have a feeling it was me that pushed for the fixed version in the first place. oh i am fickle. [08:16] fwereade: anyway, if you want it to work, just change "1.0" to "latest" [08:22] rog, I *kinda* think it's probably best to go with any specific version that has the metadata we need [08:22] rog, I do in general trust amazon not to screw it up [08:22] rog, but yu never know [08:22] fwereade: ok, i'll do that then. [08:22] rog, it's just one extra thing that *might* go wrong [08:22] fwereade: perhaps a relatively recent version, so that we have access to other stuff as well if we want [08:23] rog, yeah, the current target of latest would be fine with me [08:23] fwereade: BTW what state are the relation hooks in? [08:23] rog, basically, they're not [08:24] rog, because so many of the details depend on lifecycles not just existence [08:24] fwereade: ahhh [08:24] rog, I have a branch that I'm 90 sure will work with only minor tweaks once the substrate is available [08:24] rog, but it is 3 weeks old now [08:24] substrate is such a good word [08:24] rog, and I'm a bit worried about when I'll be able to make actual *progress* on it [08:25] rog, fingers crossed for lisbon [08:25] fwereade: yeah ikwym [08:25] fwereade: what i tend to do is keep things roughly up to date, so i don't have an almighty job when i come to merge it later [08:26] fwereade: but that can have its own problems of course [08:26] rog, I gave up on that a long time ago, I'm just planning to recreate everything using it for inspiration [08:26] fwereade: lol [08:38] rog, ffs, git is not installed by default :) [08:38] fwereade: nbd. we have other dependencies too. [08:39] rog, indeed [08:39] rog, just another irritating step ;) [08:39] fwereade: and some time later, we will should provide images that don't take 5 minutes to get all their stuff before becoming usable [08:39] s/will // [08:40] rog, yeah, definitely [09:05] rog, can I get a trivial LGTM on https://codereview.appspot.com/6495129 before I pop out for coffee please? [09:06] fwereade: assuming it works, LGTM [09:06] rog, it appeared to :) [09:06] fwereade: cool [09:07] fwereade: did you see the charm actually running some commands then? [10:36] so, the *only* hack this is using (compared to trunk) is the latest metadata [10:37] http://ec2-174-129-106-129.compute-1.amazonaws.com:3000/ [10:37] rog, *surely* we can make that a trivial and get trunk actually *working* ..? [10:38] fwereade-on-juju: so... you're actually connecting from a charm? [10:39] fwereade-on-juju: YAYAYYAYAYAYAAYYA!!!! [10:39] fwereade-on-juju: ok, i'll frikkin' do it [10:40] rog, oh yes I am :) [10:40] oops, bugger, need cash for cleaner, brb [10:41] fwereade-on-juju: while you've got an instance up, could you do a quick curl http://169.254.169.254/ and paste me the results please? [10:41] fwereade-on-juju: that is - do that on an ssh session in your instance [10:42] fwereade-on-juju: actually, scratch that [10:46] fwereade-on-juju: https://codereview.appspot.com/6498128 [10:53] fwereade-on-juju: hmm, pity your server doesn't seem to be actually connecting to IRC for me [10:54] fwereade: oh yes, that CL also renames a directory, which the codereview page doesn't show. [10:56] rog, sorry, back; looking [10:57] rog, LGTM [10:57] rog, but... what's not working for you? [10:58] rog, you *should* just be able to do new connection to chat.freenode.net and pick a name [10:58] fwereade-on-juju: i tried, but it hung. i did use "irc.freenode.net" though [10:58] rog, tbh subway *does* apear to be a touch flaky, it misses messages occasionally [10:59] rog, but I *think* we can blame the service rather than juju for that ;) [10:59] fwereade: i tried it in two windows, different server names, etc [10:59] fwereade: but yeah, i see the page! [10:59] fwereade: i just wanted to join you as rog-on-juju :-) [10:59] rog, yeah, indeed [10:59] it's lonely up here ;p [10:59] fwereade-on-juju: ok, so trunk should now be working! [11:00] rog, ok, I'm shutting this down and trying it out :) [11:00] fwereade: "you were disconnected from the server" :-) [11:01] rog, that works, anyway ;p [11:01] fwereade: yeah, i could navigate other pages too [11:01] fwereade: just not get through to irc [11:37] rog, sorry, but I'm getting 404s from 2012-06-01... [11:37] fuck [11:37] fwereade: that's the version they mentioned on the web page [11:37] fwereade: i just assumed... [11:37] fwereade: bloody amz [11:38] fwereade: have you got an instance up now? [11:38] rog, bad luck [11:38] rog, yeah, I was just about to poke around and try to figure out what it should be [11:38] fwereade: ssh to it and do curl http://169.254.169.254/ [11:38] fwereade: that should give you the list of available versions i think [11:39] rog, huh, latest is 2012-01-12 [11:39] fwereade: ok [11:39] fwereade: i *so nearly* went for a 2011 version! [11:39] rog, haha [11:39] rog, wonder wtf the deal is with versions not existing :/ [11:40] fwereade: only amazon knows [11:40] fwereade: ok, will change it [11:40] rog, 2012-01-12 verified live btw ;) [11:41] fwereade: thanks. i didn't want to wait for another live test to complete :-) [11:42] fwereade: could you verify 2011-01-01 [11:42] ? [11:42] fwereade: i think i'll go for that, as who knows how consistent amz is across regions [11:42] rog, sorry, just killed the env -- but it was advertised [11:43] fwereade: that gives us everything we want, i think [11:43] rog, sgtm [11:51] fwereade: https://codereview.appspot.com/6494136 [11:52] rog, LGTM [11:53] fwereade: submitted [11:53] rog, trying it out [12:14] oh yes indeed [12:16] fwereade-on-juju: fanbloodytastic! [12:16] rog, ain't it :D [12:16] rog, not one lick of manual intervention [12:16] fwereade-on-juju: am very glad you got to be the first! [12:17] rog, bah, there's more than enough credit to go around :) [12:17] rog, http://ec2-23-20-195-110.compute-1.amazonaws.com:3000/ -- maybe subway will be happier this time [12:17] fwereade-on-juju: you don't have to log in or register first? [12:17] rog, I just picked a name and a server and that ws it [12:20] fwereade: nah [12:21] fwereade: tried from my phone too. page serves fine though [12:21] rog, weird [12:21] rog, ah well === rog is now known as Guest96454 [13:01] For those wondering about sprint plans, Marianna says: "I'm booking the same venue as for the July's sprint" [13:01] mramm, cool [13:02] something went wrong in the notification system, so she got started late [13:02] but is finalizing the details this morning [13:02] so we should be all set to go [13:02] mramm, not sure if you saw that I just deployed a charm on trunk :) [13:02] rock and roll [13:02] great work everybody! [13:27] Hellos! [13:46] niemeyer, heyhey [13:47] niemeyer, with a pair of trivials merged by me and rog this morning, current trunk can deploy subway :) [14:02] fwereade: Wooooooo [14:02] :D [14:02] fwereade: Sooooo awesome.. we should have a beer on Sunday :-) [14:02] niemeyer, SGTM :) when are you getting in? [14:04] niemeyer: yo! [14:05] fwereade: Let me check [14:05] Guest96454: Morning! :-) === Guest96454 is now known as rogpeppe [14:05] how did that happen?! [14:06] guess i was on the last available name [14:38] fwereade: Just sent another reply on the upgrade thread [14:38] niemeyer, cool, looking [14:58] niemeyer, replied [14:58] fwereade: Cheers [15:16] * fwereade acknowledges a good point by niemeyer, and goes off to think for a bit [16:36] Lunch time! [17:10] * rogpeppe wishes there was an easy way of splitting a branch in two [17:21] gah! [17:28] rogpeppe: git is *slightly* better at that, via rebase [17:29] rogpeppe: Unfortunately we don't yet have something as complete as git's rebase yet [17:29] niemeyer: i'm just gonna copy the files into a branch off trunk and lose the revision history [17:29] niemeyer: sucks a bit, but i don't mind much [17:36] rogpeppe: Well, that's unavoidable unfortunately [17:36] rogpeppe: Splitting content in half always trashes or modifies history somehow [17:37] niemeyer: unless you use darcs apparently [17:37] niemeyer: well "alters" i'm sure [17:37] rogpeppe: I believe this is a theoretical constraint [17:45] we have a place to stay! [17:46] rogpeppe: Yeah, that was tight :-) [17:46] niemeyer: not quite as convenient as before [17:52] niemeyer: here's the upgrader logic we talked about: https://codereview.appspot.com/6492123/ [17:54] niemeyer: unfortunately i can't propose the uniter upgrade branch until we decide about UnitWatcher [17:55] rogpeppe: I don't get the comment about WaitForEnviron [17:55] niemeyer: WaitForEnviron absorbs the first event [17:55] niemeyer: but we need to see it [17:55] rogpeppe: Yeah, it absorbs to return the environ [17:55] rogpeppe: We don't "miss it" [17:55] rogpeppe: It's handling the event for us [17:55] niemeyer: yes, but we need that event in the upgrader loop [17:56] niemeyer: unless... [17:56] niemeyer: we could just start an independent environ watcher [17:56] rogpeppe: It just sounds a bit confusing.. the comment says that we can't use it because it does precisely what it should do [17:56] rogpeppe: The goal is precisely to "initialize the environ when we're first able" [17:57] niemeyer: ok. i'll try to think of another way of phrasing the comment [17:57] rogpeppe: So far I'm arguing mainly about the comment.. I don't understand what's the constraint yet [17:58] rogpeppe: It says something like "we initialize the environ when we're first able, so we can't use WaitForEnviron" but that's *exactly* what WaitForEnviron does.. not sure if you see what I mena? [17:59] rogpeppe: It feels like we could just use it, to be honest [17:59] rogpeppe: Why can't we? [18:00] niemeyer: if we use it and the environ config does not change, we won't see any more events on the watcher, so we'll never do an upgrade. [18:02] niemeyer: here's a possible updated version of the comment: [18:02] // We can't use worker.WaitForEnviron because then we don't [18:02] // see the first event from the watcher, which we need [18:02] // to see because it has version information in that we [18:02] // must see. [18:03] niemeyer: slightly better: [18:03] // We can't use worker.WaitForEnviron because then we don't [18:03] // see the first event from the watcher, which contains version information [18:03] // that we must see. === cmagina_ is now known as cmagina_away [18:05] niemeyer: i need to go, i'm afraid [18:05] niemeyer: looking forward to an awesome sprint! [18:05] niemeyer: hope your travel goes well [18:05] rogpeppe: I see, it's still not entirely right, though [18:05] niemeyer: oh? [18:05] rogpeppe: WaitForEnviron does more than simply waiting for first event [18:06] rogpeppe: We need it to handle the bootstrap behavior [18:06] niemeyer: i *think* i do that [18:06] rogpeppe: Doesn't look like so.. [18:06] rogpeppe: The logic within the switch says [18:06] 139 » » » » » // continue on, because the version number is stil [18:06] l significant. [18:06] rogpeppe: We can't move on with an invalid configuration [18:06] rogpeppe: We talk to the environment mid-way through [18:07] niemeyer: what about line 132? [18:07] rogpeppe: Ah, yeah, you're right [18:08] rogpeppe: Anyway, we can talk next week.. have fun there [18:08] rogpeppe: and safe travels [18:08] niemeyer: and you too. [18:08] rogpeppe: Thanks [18:09] niemeyer: you too! === cmagina_away is now known as cmagina_