=== rogpeppe1 is now known as rogpeppe === jamie___ is now known as Guest95549 [12:48] hi frankban, thanks for the great review! i've reproposed and would like if you'd give a look, especially to the url encoding bits. https://codereview.appspot.com/26740043 === gary_poster|away is now known as gary_poster [13:09] rick_h_, hey. Dunno if hatch will be around today (sick). If it gets later with no reply from him on your branch, please ask around for someone else to follow up. [13:11] rick_h_, I also can give answer to hatch's last question, if desired [13:14] gary_poster: ok will do. It needs two reviews and he was generally cool with things and said he was debating a LGTM but wan't to check on using deltas [13:14] gary_poster: so I'll rope someone in for a second review/QA while I tinker with it per hatch's review [13:14] ok cool [13:15] gary_poster: if you know the 'why' for hatch appreciate the reply. (and I can see as well) [13:15] can be roped if desired, though I have interruptions for next hour or so [13:15] ok will reply [13:15] gary_poster: rgr, thanks [14:04] bac: looking [14:06] jujugui updated per hatch's review and looking for a second review/qa please https://codereview.appspot.com/26290043/ [14:06] morning [14:06] still sick, but not enough to miss :) [14:07] hatch: chicken noodle soup for your lunch today, rock on [14:07] hatch: we are two [14:09] yup [14:09] I had to dig in the freezer for some [14:09] you mean the back porch? :P [14:10] lol it's not QUITE cold enough every day for that [14:10] we had -7C the other day. Getting cold too early this year [14:10] woah! I had no idea it got that cold there [14:11] yea, we get down into some negative F range pre-wind chill [14:11] but that's usually more Jan weather than mid-nov [14:11] shoot, had our first snow before all the leaves fell [14:14] frankban: out of curiosity why can't guiserver receive a 'watch' rpc call from the gui and then send 'fake' deltas down the wire which we can react to? Why do we have to essentially long-poll for them instead? [14:15] rick_h_: I LGTM'd I'm still not sold on the util object vs class but I'll let it go ;) [14:16] hatch: rgr, did you see my point though that there will never be more than one 'instance' since that data is the same? I think that's what led me more utils vs object [14:16] yeah so I was thinking in that case it could be an extension on another clas....like say....app :) [14:16] hatch: well that and just refactoring the methods didn't have an object to start (well it was an extension, but on their own) [14:16] hatch: :P right but then you need a 'fake app' to test and such. Ugh [14:17] hatch: but I admit it's something that can go a few diff ways without being too crazy [14:17] no no, you can just var myext = new MyMyExtension() [14:17] yeah that's why I didn't argue over it :) [14:17] hatch: except this.env doens't exist in new MyMyExtension [14:17] implementers choice....for now! [14:17] wheee! I won the joy of this lovely branch with all the live ec2 QA/testing work yay me [14:17] myext.env = envStub; [14:18] :P [14:18] bah, invisible properties to the test reader then "what's this 'magic' attribute' [14:18] go get your soup early [14:18] hatch: no technical reasons, just a design decision. we decided to reuse a pattern already present in juju-cure (the megawatcher) and to not introduce a new one. Even if it feels more natural to just push changes from the server, the more I think about it the more I feel we made the right decision, especially considering that the deployer can eventually be implemented in core. [14:19] gotta love how things to first written and depoyed in python to get re-written in go down the road potentially [14:19] frankban: does the 'longpolling' introduce any potential performance/race conditions or is it simply a flag on the guiserver side that says 'when this comes up, notify the user here' [14:19] * rick_h_ wishes the config-changed hook to go faster...faster [14:20] hatch: think node-like :P [14:21] the latter, hatch [14:23] ahh cool cool [14:24] hatch: it's longpolling-like but in the same connection, so I don't see too much performance degradation. It just mimics a request/response over a wss connection. If races are discovered, that's a bug. In theory, when the client calls Next, it receives all the messages not yet received if there are, otherwise the next new event when it is generated [14:26] sounds good [14:26] I was mostly curious [14:26] rick_h_: is someone else doing the qa on your branch or would you like me to? [14:26] hatch: I was hoping to rope Makyo into the second review and qa, but if you've got bandwidth to qa would appreciate it [14:27] hatch: not sure what your live juju env is looking like these days [14:28] I have a juju core instance locally for lxc and setup for ec2 [14:28] hatch: ok cool, this is only for ec2-land so if you can qa would appreciate it [14:28] hatch: tried to put some cut/past instructions on there and note the various code paths [14:29] on it [14:29] still working on a success for all 4 paths myself with various bundles [14:29] keep finding new bugs or corner cases in things [14:34] hey frankban, I copied quickstart over to ppa:juju/stable yesterday [14:35] I think maybe "releasing" it can be copying it over from the beta ppa? [14:35] going forward as a process I mean [14:35] what do you think? [14:35] jcastro: used it for his blog post yesterday http://www.jorgecastro.org/2013/11/14/from-0-to-hero-in-a-few-minutes/ [14:36] yes, mwahaha, you can't remove it now [14:37] lol, trapped by a blog post [14:37] :-) cool [14:37] haha [14:38] jcastro: hey do you know anyone who knows nginx really well? I'd love it for the http interface for my ghost charm :) [14:38] we used to [14:38] gary_poster: cool thanks, sounds good. [14:38] he did the setup in the wordpress charm [14:38] great [14:39] hatch, is just straight up theft from wordpress not feasible? [14:39] not sure I haven't looked into it [14:39] I just figured we needed an nginx charm [14:39] there are some, but none are promoted [14:43] woot, it's working. Double bundle deploys, one is scheduled. party on [14:43] bac: thanks for the changes, just added a comment. do you have an example charmworld URL to use to wwatch deployments count? [14:43] :) [14:44] frankban: can't yet as juju gui and quickstart aren't sending them yet [14:44] frankban: i mean they are not yet sending the optional BundleID so it doesn't get triggered [14:45] frankban: i'll probably update the GUI today to do that [14:46] bac: ack [14:47] frankban: so i think the best QA is just deploying a bundle and seeing that nothing blows up. [14:47] bac: my branch plays around with that stuff so watch out [14:47] bac: I'll check the guiserver logs [14:47] bac: the good thing is that the deployBundle call takes a 'name' parameter that it looks like no one uses, just passes null [14:48] sweet, "bundle 1 completed" "bundle 2 starting" [14:48] rick_h_: the gui will use it once we have bundle disambiguation I guess [14:48] frankban: oh! that's what the name is for [14:49] frankban: makes sense. I was going to bring up that gary_poster should add that to his list of 'bundle features needed' as well [14:49] rick_h_, ack. added [14:50] ty [14:52] bac: is you branch merged with ~juju-gui trunk? [14:52] frankban: which branch? [14:53] bac: lp:~juju-gui/charms/precise/juju-gui/trunk/ [14:53] frankban: are you asking if i've pulled the latest changes from the charm trunk into my branch? no, i haven't [14:53] frankban: i'll do that and make sure there are no conflicts and tests pass. [14:54] bac: thanks, then I'll QA [14:58] how is everyone/everything today? All going good? [14:58] hatch: actually, nothing here should be ec2 specific. I was using quickstart to initially bring things up and that required ec2. [14:58] oh now you tell me [14:58] hatch: but that speed is offset by the other changes. Sorry to lead you to ec2 there [14:58] hatch: heh, yea I got thinking "why did I have to dothis on ec2?" [14:59] but hey, I got some good qa notes/bugs on ec2 yesterday/today :) [14:59] hey luca__ . all good. releases have gone well to good reaction. I have some jaas thoughts I want to share with you and ale before Tuesday on what we think we ought to start with next for jaas (hint: we think our previous plans are still necessary for jaas) [15:00] gary_poster: cool, we should schedule something for Monday [15:00] agree [15:01] rick_h_: updated status for deployment: 0 [15:01] that is kind of...odd [15:01] like what's the 0 for? [15:01] frankban: merged, unittests pass, pushed [15:01] hatch: yea, I'm hoping with the naming /id stuff bac is doing we can make that nicer [15:01] hatch: it's the deployment id, we don't have anything else to call it [15:01] hatch: but you can do parallel imports so we have to be able to tell notifications apart [15:02] luca, you and ale only have 4:30-5:00 Monday. Will schedule. [15:02] how about "Updated status for deployment id: 0" [15:02] :) [15:02] luca__, I should say :-P [15:02] hatch: I like small edits :) [15:02] haha, just wait, it's still running, there may be more [15:03] gary_poster: ok [15:03] hatch: all good, thanks for trying it [15:03] gary_poster: add Peter too if you think he should be there [15:03] luca__, he won't be at Tuesday meeting will he? [15:03] prep for that is what this is about in my book [15:03] aligning among ourselves [15:04] I wonder if I should give you a preview today so you can warn others if you think it is necessary :-) [15:04] gary_poster: he won't be at the tuesday meeting [15:04] ok [15:04] luca__, you have 30 min today sometime? [15:05] rick_h_: so both discource and postgres dropped onto the canvas almost right away and I only saw a single notification [15:05] gary_poster: sure, I'm in a cloud installer call from 4-5 but I can jump on after that [15:05] bug? [15:05] luca__, isn't that your EoD? [15:06] gary_poster: sort of, it's ok though [15:06] hatch: right, but they're still loading right? not green [15:06] bac: please fix a lint error introduced by your last change [15:06] postgres turned green without a notification [15:06] still waiting on discourse [15:06] ...ok luca__ I'll schedule and try to make it fast :-) [15:06] gah [15:06] hatch: right, it'll wait until it's done and relations are done to say it's completed [15:07] hatch: there's not a ton of updates unless it errors then you get a notification the instant things error [15:07] otherwise it's "Started" ....wait a bit ... "Completed" [15:07] gary_poster: :) [15:07] ohh ok, I was under the impression that it had one on each machine being spun up [15:07] hatch: if you deploy another one right now you'll get notified it's "scheduled" and then when the first bundle is done it'll to go "Started" and then "completed" [15:07] but that wouldn't make any sense [15:07] because you can see it in the canvas [15:08] hatch: yea, it's more of a 'how is my bundle doing' state and the big thing was that errors would fall silently so this is providing a "Did it work or fail" summary notification [15:08] hatch: yeah it would be like having subtitles [15:08] lol, "Juju Gui, now with french sub-titles" [15:08] :-) [15:08] hey rick_h_, I forgot to ask yesterday, the ability to deploy --to in a bundle? Is that v2? or v3? [15:08] I want to have an all in one discourse instance, including gui/bootstrap [15:09] jcastro: so that should owrk from the deployer now. Not sure what's up with that on the quickstart front. frankban on the radar? [15:09] discourse sure takes a while to spin up [15:09] yeah I know it works with proper deployer [15:10] jcastro, shoudl work from deployer now, yes. won't be visible or changeable in GUI though. When that changes is currently up for debate. [15:10] hatch: yea, it's got to pull the world of ruby from gems, github, etc [15:10] but now that I have quickstart I want to point people towards that instead of "raw" deployer [15:10] frankban: done [15:10] oh jeebus I forgot it was ruby [15:10] lol [15:10] ruby.....because python wasn't....we've got nothing [15:10] lol [15:11] gary_poster: yea, but quickstart could support it even though the gui doesn't have a representation [15:11] haha [15:11] hatch, man you want a good time, you think that's bad ... deploy the cloudfoundry bundle and debug-log that [15:11] rick_h_, yeah, was agreeing with you on that. [15:11] rick_h_, I was thinking more from a "just ignore that section and pass it on to deployer" [15:11] haha [15:12] that way we could do "all in one" bundles for people to deploy and just say "they don't show up in the gui properly but we'll fix that later." [15:12] yeah, think that works now [15:12] jcastro: quickstart deploys bundles using the guiserver, which then uses the last deployer version. so it should work [15:12] also, I had this crazy dream trying to figure out a nickname for bundles that are all in ones, instead of "one shots" as I've been calling them [15:12] frankban: oh, whenI looked at the source the --to was a special flag picked up by code the guiserver path wasn't using [15:12] I'm going to start calling them happy meals. [15:13] [15:13] self contained, all in one boxes. [15:13] I knew you'd get a kick out of that [15:13] :-) [15:13] where's my toy! [15:13] frankban, ok I'll give it a shot now [15:14] how is --to documented in the bundle file? [15:14] jcastro: cool, let me know how it goes. rick_h_: I am referring to http://pythonhosted.org/juju-deployer/config.html#placement [15:14] so to: 0 [15:15] should work? [15:15] jcastro: give it a try [15:15] (never used myself) [15:16] hey bac [15:17] did you guys ever finish off that bundle doc and hand it off to nick? [15:17] we should probably land that soon [15:17] I can finish it off today if you'd like [15:17] bac: deploying [15:17] jcastro: i spoke to nick about your comments and he said he'd take care of it [15:17] jcastro: probably last week [15:18] ya know I just read a blog post yesterday about why distributing ruby apps suck....now I see why first hand [15:18] lol [15:18] ok I'll chase him down, thanks [15:19] rick_h_: install hook failed on discourse and I didn't get a notification [15:20] frankban: ah, you're right. I was confused back when I missed the juju-client has machine_spec [15:20] hatch: hmmm, no notification that the bundle errored? [15:21] nope [15:21] and clicking retry in the gui doesn't appear to do anything.... [15:21] frankban, it didn't error out [15:21] but it also didn't to: 0, fired up individual instances [15:21] also, whoever make "juju quickstart ." work, thank you! [15:21] hatch: hmm, yea I just noticed that resolve issue on aonther bundle where I did get the error [15:22] hatch: can you try another bundle, the wordpress-simple or the mediawiki one? [15:23] yup [15:23] also found a sidebar bug haha [15:23] hatch: thanks, I'll try out the discourse onein particular. [15:23] sidebar bug? [15:24] when I opened the unit details it was under the sidebar [15:24] so I closed it [15:24] yea, that's intentional [15:24] now I opened it and it no longer has the 'jorge' search results [15:24] oh hmm, lovely [15:24] oh well....once fullscreen is gone we can hopefully rely on the yui routing [15:25] hatch, rick_h_ : the deployer can take some time before returning a hook error [15:26] frankban: k [15:26] frankban, should I file a bug? [15:26] frankban: yea, I noticed a small lag, but it's purely a wait/see [15:27] jcastro: investigating [15:27] hatch: filed the bug on the resolved button [15:28] rick_h_: haha I filed one on retry [15:28] doh [15:28] rick_h_: ok I deployed the wordpress bundle it had the 'started' message [15:28] then say....5-10s later it had the 'completed' message [15:28] but both are still yellow [15:29] hatch: what ID was that for? [15:29] 1 [15:29] both were for 1 [15:29] * rick_h_ wonders if that was bundle 0 [15:29] hatch: it said completed? not scheduled and started? [15:29] it said started, then completed [15:29] no scheduled [15:30] so it must think that the discourse one is done? but didn't notify [15:30] hatch: yea, if it started the second then it's killed off the first one [15:30] sounds like a bug [15:30] not sure from which end though [15:30] this is tough to debug [15:30] :) [15:30] yea, it's been a tough branch to do [15:31] no console errors or anything [15:31] and now you're kiling my joy in testing it ok this morning [15:31] well I'm going to start on my 'fail testing charm' and bundle [15:31] no console in the charm (and I tried to turn that on) [15:32] gah! /me starts up another ec2 env [15:32] rick_h_: here is the retrying bug https://bugs.launchpad.net/juju-gui/+bug/1251664 so you can relate it to your bug [15:32] <_mup_> Bug #1251664: Retrying doesn't appear to do anything [15:32] hatch: yea, done. I marked mine a dupe of that and figure they'll go together [15:32] rick_h_: so I'm going to tear this ec2 instance down, cool? or do you need anything from it? [15:33] frankban: so did your deploy and QA complete? [15:33] hatch: yea, kill it. I'll run again here. You did hte discource bundle first and then wordpress right? [15:33] yeah [15:33] and the discourse charm failed [15:33] hatch: I swear it worked here, but I did different bundles in a different order [15:33] I had a different charm fail and such [15:33] hatch: rgr [15:33] ohh ok [15:33] well I'm going to write a charm which fails so we can test these things [15:34] hatch: rgr [15:34] destroy-environment should have like 3 levels of 'are you sure?' [15:34] which gets more dramatic each time [15:34] "are you sure you want to destroy your environment" [15:34] "all of your data will be lost forever" [15:35] "lost into the deep dark abys that is unallocated memory never to be seen or heard from again, can you really do that to such innocent data" [15:36] jcastro: maybe I've found the problem. hazmat: ping [15:37] jcastro: to double check, could you please try to deploy with "to: '0'" (quoted) [15:38] got it [15:44] frankban, that worked [15:47] frankban, jcastro ack, in meeting, noted re bug [15:47] * jcastro nods [15:47] jcastro: ok, thank you, we will need another release of the deployer and of the charm to fix that. in the meanwhile, I think it's fine to use quotes there [15:47] ok so now with to: and constraints: I am good to go [15:48] jcastro: cool [15:48] frankban, yeah for now I'll leave the current discourse one in the store and create a different all in one [15:48] discourse-happymeal <--- just kidding [15:50] jujugui call in 10 [15:54] bac: AttributeError: 'Deployer' object has no attribute 'charmworldurl'. w eboth missed that, used without the underscore in _import_callback [15:54] frankban: thanks [15:55] bac: and so I guess that "if" is the only part of that branch code that's not covered [15:58] jujugui call in 2 [16:29] frankban: before you take off for the day I have a couple questions about Go for my talk tomorrow [16:29] hatch: now is ok [16:30] https://plus.google.com/hangouts/_/calendar/Z2FyeS5wb3N0ZXJAY2Fub25pY2FsLmNvbQ.t3m5giuddiv9epub48d9skdaso?authuser=1 [16:30] it'll be really quick :) [16:31] lies! don't be fooled frankban [16:38] lol === TheFreeMue is now known as TheMue [17:15] I have a charmworld branch up with some refactorings needed for my current work. I added some comments for the reviewer to make it easier to see what is going on: https://codereview.appspot.com/22190049/ [17:37] unfortunately reviewer comments don't migrate across patch sets [17:37] jujugui: charmworld branch up for review: https://codereview.appspot.com/22190049/ I added some comments for the reviewer to make it easier to see what is going on. [17:38] gary_poster: nope they don't :( [17:38] benji will look later if no o ne else takes you up on it :-) [17:38] thanks [17:42] * rick_h_ curses npm!!!!!!!!!!!!! [17:46] it's down again eh? [17:46] yea, so juju set juju-gui-source is full of fail for QA [17:47] * rick_h_ takes this branch out to the shed out back [17:48] I've always wondered where that saying comes from [17:48] it would make a mess of the shed [17:54] https://twitter.com/npmjs/status/401406454088732672 guess it's coming back wheee [18:02] jujugui is manage.jujucharms.com not responding for anyone else? [18:02] wfm [18:02] bac: ok, thanks [18:04] heh, it completed after 1.7M [18:15] hmm, looks like it's something in the inter-tubes http://www.downforeveryoneorjustme.com/manage.jujucharms.com shows it as down as well [18:16] hatch: aha! you didn't get the error because there was no error message in the websocket [18:16] * rick_h_ goes to check branch and see what to do about a completed with error without a message [18:51] benji reviewing [18:53] jujuguipals: charmworld review please https://codereview.appspot.com/23330046 [18:53] benji LGTM [18:54] bac reviewing [18:54] will not do qa though [18:54] gary_poster: you are a reviewing machine [18:54] :-) [18:54] well, ok then [18:54] gary_poster: thanks [18:55] welcome [18:58] bac, I'm wondering about edge cases for that new search behavior. [18:58] bac, for instance, what if I type ~gary demo? [18:58] I'd be looking for Gary's demo [18:59] Is there some autocomplete rule I should be aware of? [18:59] Since the autocomplete is now GUI's only search tool... [18:59] it would ne nice if it had the features that m.j.c search had [18:59] maybe (?) [18:59] well the old autocomplete rule was you'd get a 'name: text' search [19:00] I have another call [19:00] bac, yeah, we can separate that out into another card/branch for The Future [19:00] but [19:00] at the least [19:00] brb [19:00] ok [19:05] hatch: got a sec? [19:07] rick_h_: yup [19:09] hatch: https://plus.google.com/hangouts/_/7acpi4rj1mataa0gd8t6esan2g?hl=en [19:19] bac ping me when you have a sec for a call. comments will go faster there and you can beat me over the head more readily [19:20] gary_poster: now? [19:20] cool [19:21] gary_poster: ringing [19:24] hatch: ok, they both errored here. With crappy error boxes because they've got no error message. However, that's a fix for another branch in another project of code [19:24] hatch: let me know what you get and then Makyo can just review if he gets a chance and get this landed [19:25] sounds like a plan - the gui is almost up [19:28] Will do [19:29] well I THOUGHT it was almost up [19:30] lol [19:30] lxc's apparently take a long time to start [19:35] should only the first time you launch your first one [19:35] after that it's fast! [19:38] rick_h_: when you say 'fast' like <1m? [19:38] hatch: my lxc units pull up in < 2min [19:39] bootstrap is < 1min and juju deploy juju-gui is < 2 or 3 [19:39] hmm yeah mine is definitely longer [19:39] now setting the branch takes a bit, 10 maybe? [19:39] ~15m total for the gui [19:40] http://www.indiegogo.com/projects/linux-voice [19:40] looks interesting [19:41] unfortunately the price in GBP is a little craycray [19:41] will be over $100/yr for me :( [19:46] yeah I think these lxc's are running at 10m [19:46] maybe it doesn't do multiple lxc creation at once very well [19:49] rick_h_: well it has staged up the bundle properly so yay [19:50] gary_poster: i was so proud of my tilde-matches-owner test with bonnie and clyde. next time. [19:51] bac, :-) sorry [19:56] oh c'mon discourse fail already so I can qa ok this! [19:58] yay [19:59] error notification this time? [19:59] uh oh [19:59] no, no uh oh allowed [19:59] there's no message, that's the known issue part. [19:59] https://plus.google.com/hangouts/_/7acpjifot1t4vf39qmvgc5if18?hl=en [20:00] ^ rick_h_ [20:01] yayayayyaay. Makyo so hatch has QA and we're good. [20:02] Coolcool, [20:21] \o/ [20:45] * bac off to festivities. have a good weekend. [20:45] you too [20:45] enjoy [21:06] bye bac [21:06] benji, are you outta here also? [21:06] or do you have another hour? [21:08] gary_poster: I'm pretty much outta here :) Did you have something in particular? [21:08] benji, cool. wanted to run a document past you. it'll wait. have a great weekend! [21:08] gary_poster: something to look forward to [21:08] ;-) [22:05] gary_poster: email lgtm :) [22:05] hatch, :-) thanks [22:05] feel better hatch [22:06] have a great weekend everyone [22:06] * gary_poster waves [22:06] thanks, you too cya === gary_poster is now known as gary_poster|away [23:08] hatch you still here all by yourself? [23:08] heh, hatch can't leave! Moar branches!