[11:16] hey frankban, forgot about juju server for bug 1240708 [11:16] <_mup_> Bug #1240708: API server falls over repeatably during AllWatcher Next, killing GUI

[11:16] I still don't see why it would be on our side [11:16] but as long as it is fixed, I'm ok with it :-) [11:18] gary_poster: the builtin server is enabled by default in the new charm release. if you switch to the legacy server, the problems (errors in console, inspector frozen) disappear [11:19] frankban, ah ok [11:19] :-/ [11:20] gary_poster: I am investigating and it seems to be (drum roll and shame on me) a string encoding problem in a log message (when the charm info includes non-ascii bytes) [11:21] frankban, lol, yay string encoding [11:21] gary_poster: I stupidly assumed the ws messages were already encoded, but tornado (a good guy) gives them to me as unicode objects [11:23] frankban, completely understandable. ugh string encoding in Python. yay for easy fix, hopefully. [11:23] gary_poster: the fix is easy: add .encode('utf-8') to all the log calls which include a ws message [11:23] frankban, ideally we would have had some people using the server when we asked ;-) [11:24] cool frankban. have you verified that this fixes the symptoms? [11:26] gary_poster: https://ec2-54-234-223-41.compute-1.amazonaws.com/ (password in pvt): could you please double check? that's running the builtin server. [11:28] frankban, perfect [11:29] frankban, I still don't see exactly how this would have resulted in a stop in the core logs. Our server fell over and disconnected from core? [11:35] gary_poster: at least the builtin server was trying to get our attention: "Juju API unexpectedly disconnected". when a message arrives from the iostream, tornado calls the async callback previously registered. if an exception is raised by the callback, it aborts the connection. In out case the callback is WebSocketClientConnection.on_message, which calls WebSocketHandler.on_juju_message, which calls "logging.debug [11:35] (self._summary + 'juju -> client: {}'.format(message))". the error in the juju-core logs refers to a next() response not being able to be sent because the connection is dropped (I guess) [11:35] frankban, I did have a related thought: we ought to make our server disconnect webserver connections when we get disconnected from core [11:36] frankban, ah ok [12:02] frankban, do you want to postpone 1-on-1 or go ahead? completely up to you [12:02] gary_poster: joining [12:57] antdillon: did you see the feedback on your branch? [12:57] antdillon: make sense/in progress or need a hand with that? [13:00] gary_poster: thinking on the conversation about the error stuff with promises I added a high maint. card for a generic 'smart' promise error catcher with some ideas in the notes [13:00] gary_poster: let me know if that's not a good plan or there's a better one out there. [13:06] rick_h_, cool. I'll see if I can have an explicit time for this conversation in SF. thank you [13:24] gary_poster: could you please review https://codereview.appspot.com/14772044 ? [13:24] frankban, on it [13:24] gary_poster: if you need an environment ready to qa, you can use https://ec2-54-242-218-111.compute-1.amazonaws.com/ [13:25] perfect thanks frankban [13:29] frankban, had to look up {!r} :-) [13:30] :-) [13:34] frankban, shouldn't we have encoding protection for logging in server/guiserver/auth.py and server/guiserver/bundles/utils.py? Looking to see how _summary is generated also... [13:35] _summary looks fine as is [13:36] well, no... [13:36] request.uri can have unicode, right frankban ? [13:36] (server/guiserver/utils.py in request_summary function) [13:38] gary_poster: I'll check if request.uri is unicode. for the other places with logging, I'll add !r [13:38] cool thanks frankban. qa looks good and LGTM with those comments. recording on rv now [13:38] thanks [13:40] done [13:43] hey gary_poster [13:43] so I just deployed gui just now in AWS [13:43] rick_h_, antdillon did my comments on https://codereview.appspot.com/14700043/ make sense? antdillon, rick_h_ and I would love to get this landed [13:43] jcastro, listening [13:43] and now in the inspector the link is not only unclickable [13:43] I can't even highlight it to copy and paste [13:43] jcastro, what link? [13:43] in the inspector [13:43] oh [13:44] if I click wordpress/0 under running units [13:44] gary_poster: yes, made sense to me. I've ping'd antdillon and if I don't hear back I'll take it over and get it done once I finish huw's here. [13:44] cool thanks rick_h_ [13:44] *tsk *tsk huw forgot to add tests [13:44] I noticed :-) [13:44] so adding to help it move forward so we'll have it on the tab [13:44] jcastro, hoping to make a release with the fix for the bug you reported today. unable to copy and paste is pretty odd [13:45] that is, hoping to make a release today of the bug you reported a week or so ago [13:45] yeah I think that might be a browser bug [13:45] gary_poster, is it within 3 hours? I am doing a demo during the ubuntu demo hangout [13:45] so I can either work around or wait, I can do either or [13:45] rick_h_, also, new presentation with the gui trick [13:45] since google hangouts let you share only one window [13:46] I added the google ssh for chrome extension, and then I ssh to localhost to do the juju cli stuff [13:46] jcastro, work around I'm afraid. we have one more blocker and person working on it doesn't start for a few minutes. I'll ping you if I have good news, but for now assume we won't make t in 3 [13:46] that lets me tab back and forth from the gui and the CLI in one window, badass [13:46] jcastro, actually [13:46] jcastro: hah! awesome [13:47] jcastro, if you want you could try using juju set juju-gui juju-gui-source=lp:juju-gui [13:47] jcastro if that works better for you, use it [13:47] rick_h_, Hey, sorry knee deep in release. I read the comments, will fix them up now and push them [13:47] it will take about 10 min to build, during which the original release will be served [13:47] gary_poster, I'll rehearse with the release for now and workaround [13:48] I learned my lesson from demoing trunk [13:48] not your trunk, I mean trunk in general [13:49] jcastro, :-) ack and understood. FWIW, we have an instance running 0.10.1 right now and copy/paste of url seems to be fine. :-/ [13:49] on ec2 [13:49] so dunno [13:49] gary_poster, ok that appears to be a browser issue on my side [13:49] ok cool [13:49] I killed chrome with a -9 and respawned it now it magically works [13:50] heh [13:50] ok [13:53] morning [13:53] hey hatch [13:53] and rick_h_ [13:53] a guy put haste in the store [13:53] it's a nodejs pastebin, same one that runs hastebin.com [13:53] it demos nice [13:53] oh very cool, I'll have to check it out [13:53] hatch, welcome back! [13:53] nice time? [13:54] thanks! glad to be back [13:54] jcastro: nice [13:54] it was nice to not be on a computer for a while :) [13:57] rick_h_, gary_poster Do you know how I should implement the routing code for onboarding? [13:57] antdillon: you don't need to. There's no routing really involved [13:57] antdillon, did you see the patch I provided? It's not perfect, but it's not bad. :-) [13:58] now onto the 400+ emails [13:58] antdillon: oh yea there's a pastebin link gary gave you [13:58] only 400+? bah :-) [13:58] haha I tried to stay on-top of them [13:58] failed I guess [13:58] lol [13:58] :-) [13:58] gary_poster, Ah sorry missed it, is it in the qa? [13:59] antdillon, yeah. lemme find... [13:59] antdillon, http://paste.ubuntu.com/6247161/ [14:00] gary_poster, Awesome thanks [14:00] welcome [14:12] gary_poster: request.uri is already a string, fixed the other occurrences [14:13] hey gary_poster. chat? [14:14] hi hatch [14:14] hey bac [14:17] frankban, awesome. bac, sorry, realized I was missing when I got some water. coming over now [14:48] emails done [14:49] gary_poster: any specific task you'd like me on right now? [14:52] hatch, wow, right you are here. :-) Makyo is working on the remaining critical bug, so f he could use help that would be great. Otherwise...you could qa new bundle stuff you haven't seen, you could try making a version of the wiki bundle that has x/y annotations and make sure that works properly, or... [14:53] hatch, fix firebug so that it doesn't close websockets if you set breakpoints. [14:54] hatch, bundle vis pan and zoom if you are up for challenge [14:54] hatch also could go for some clean up branches [14:54] hey luca__ I gave you a branch showing the behavior of hiding the inspector when you made a relation [14:55] added a bug today about the activeTab support missing from the bundle details [14:55] you want that? [14:55] hatch ^^^ rick_h_ [14:55] trivial [14:55] gary_poster: ah yeah, I have not looked at that yet [14:55] sure [14:55] gary_poster: it might be better to review it next week quickly when together [14:55] luca__, :-P I make you a branch and everything! :-) ok cool [14:55] and Makyo sorry can't help there - does the native FF inspector work yet? :) [14:55] gary_poster: haha sorry :D [14:56] np :-) [14:56] hatch, heh, I haven't tried in so long, come to think of it. Should just use gary_poster's trick rather than messing with config-debug so I can just use chrome. [14:57] I guess that's better for QA than dev, though [15:06] gary_poster: Also, I think having a 0 unit state could actually really help us [15:07] rick_h_: I will review your onboarding branch [15:08] hatch: I'm actually reviewing/landing that for antdillon [15:08] gary_poster: 's done a second review [15:08] hatch: so we're good [15:08] * hatch rejects [15:08] [15:09] missing description and camelcase in app.js (so far) [15:17] * hatch really would have rejected that branch [15:17] * rick_h_ adds a description and fixes camelCase [15:17] hatch: due to the url hackery? [15:17] 4 event handlers for a hover change, url wakyness, plus some other minors [15:18] well as you can see it's not been without it's dicussions [15:18] and it's a first pass, it'll get updated. [15:18] behind a FF [15:19] I guess....... [15:19] I'd like to at least see cards for the follow-ups [15:20] just so that the 'bad' code doesn't stick around [15:21] sorry that came off a little...harsh - it wasn't meant to be like that :) [15:24] 4h of sleep is messing with my filter lol [15:30] love the skinnier header [15:30] yep [15:31] and I see the charm token has been used 'again'! lol [15:31] should put a big warning at the top of that file - "used EVERYWHERE" [15:31] haha [15:31] yep, jcsackett has claim for most reused bit of code in the gui anywhere [15:32] oh I thought the token was your baby? [15:32] or did you just adopt it when he left? [15:32] well I helped think hard about it :) [15:32] (woah that was kind of morbid reading it back) [15:32] lol [15:32] hatch: hover change: that's probably something you change in no time flat. slip it in? I forgot about the YUI hover helpers. url wackiness: not sure what you mean, once my patch is applied, other than the fact that the desired behavior is not clear to me until fullscreen mode is gone. I made a call with behavior I thought was ok in the patch. might be nicer if it were in browse code so it didn't have to repeat [15:32] the "sidebar" logic, but enh. Do you have more concrete concerns that we should make a bug for? other minors: slip 'em in? This was a code contribution, and as such I want us to help it go smoothly. [15:33] gary_poster: it's in and landed. Created a card to move to the browser.js module when we add support for storing the state of 'did the user see this, don't show again' logic [15:33] rick_h_, ok cool. [15:33] thank you [15:33] gary_poster: hover clean up can occur at that time as well, but we can get UX feedback as I've got some design concerns overall and prefer to get luca__ and company looking at it [15:33] cool [15:34] rick_h_: gary_poster oh no, rick has design concerns! run for the hills :P [15:34] :-) [15:34] haha [15:34] luca__: :P comingsoon is loading [15:34] expect a ping any minute now :P [15:34] rick_h_: rofl [15:34] gary_poster: the biggest issue I had was the url routing stuff should be handled via the routing code [15:34] the onboarding masking is a bit nuts [15:35] but then rick_h_ infrmed me there was discussion around that already [15:35] hatch: yea, and that'll go away entirely when it's in the browser [15:35] it'll be able top just check the viewState.viewmode [15:35] modals can die the most painful way possible [15:35] heh [15:35] the WORST is when people put videos in modals [15:36] 'yes, I totally want to look at a black window while this video is buffering instead of reading your page content' [15:37] tar --exclude affects the sources of symlink files. weird [15:37] man I just sound bitter this morning [15:38] so if you have build-prod/juju-ui/assets/source/api as a symlink...oh wait a sec... [15:39] is the 'Add to my canvas' text shifted to the left for everyone else too? I'm having some caching issues so just want to make sure before I file a bug [15:40] hatch, charm? looks ok on comingsoon to me [15:40] hmm ok I'll screenshot it [15:40] bundle looks fine too [15:41] gary_poster: https://www.evernote.com/shard/s219/sh/75bfc4a3-83ae-4357-8803-903590073a59/f27d7423845e6b325848d3880255f296/res/f32bf30f-e9a9-47db-acd4-dfa4ea52e33a/skitch.png [15:42] that's where I see it hatch. wrong? [15:42] shouldn't the text be centered in the box? [15:42] oh! [15:42] hatch, that tells you how fabulous my eye is [15:42] yeah hatch [15:43] I bet the issue has something to do with the similar bundle button [15:43] which is centered [15:43] * bac wonders if that is the WORST. #firstworldproblems [15:43] :-) [15:43] haha ok np, chrome has just been crashing for me a lot this morning and refuses to clear the cache so I wasn't sure if it was just me :) [15:43] bac: no, modal videos are clearly worse :P [15:44] hatch: ok, yeah i see your point. [15:45] darn I missed http://www.kickstarter.com/projects/556341540/pressy-the-almighty-android-button while I was away [15:45] such a great idea! [15:47] rick_h_: ^ figured you'd like that [15:47] hatch: yea, almost got into that [15:49] * Makyo pointedly goes to make coffee. [15:49] rick_h_: I'm assuming that this tabview hash code is handled separately somehow? It should be handled by the tabview code by default but that doesn't appear to be the case on the bundle tabview even though it's using identical code [15:50] lol [15:50] jujugui call in 10 [15:50] * hatch waited until Makyo had left [15:50] nicely done, all [15:51] * Makyo claps for hatch "Well done! You beat me to it!" [15:51] "pressy comes with fourteen click combinations" -- that does sound intuitive! i can rarely get double-click-to-skip song to work on my teeny tiny headphone button [15:53] ugh, charmworld login tests fail if you lose network connectivity. [15:53] hatch: the view has to handle an activeTab attr and get that from the browser.js state controller [15:54] hatch: I didn't look into why it didn't work since I'd think it'd be shared. I wonder if the bundle stuff just isn't setting the activeTab in browser.js like for charm.js [15:54] hatch: and then there's a line that takes the activeTab ATTR and forces the tabview widget to go to that tab [15:54] rick_h_: ahhh I see that now [15:54] I missed it because of the name difference [15:54] cool thanks [15:55] hatch: coolio [15:55] yup trivial fix, I'll clean it up so it's shared [15:55] hatch: awesome, thanks [15:55] do we know why it's not working by default? [15:55] like why the tabview code isn't reacting to the hash in the url? [15:56] probably knew at one point [15:56] hatch: yea, trying to recall. The issue was that we ignore parts of the tabview because if it routes it causes issues [15:56] changing the url, cauing a dispatch, wants to redraw, something [15:56] I'll accept that [15:57] lets blame double dispatch [15:57] lol [15:57] I feel like that 'Blame Candada!' song from south park [15:58] "blame double dispatch! blame double dispatch!" [15:58] lol [15:58] jujugui call in 2 [16:01] rick_h_: technically isn't the charm model the most used code anywhere? [16:01] s/used/reused/ [16:01] jcsackett: it' [16:01] maybe I guess [16:01] jcsackett: I don't know what you just did, but your text changed :) [16:01] used changed to reused [16:02] hatch: i sent a substitute command, a la vim. some clients render it, and anyone using a client that doesn't probably knows what it means. [16:02] oh cool [16:02] /s/used/reused ? [16:03] * jcsackett nods [16:03] less the leading slash. [16:03] coolio [16:03] s/coolio/foobar [16:03] doesn't look like I can do it [16:03] haha [16:03] ...that's odd. [16:03] if it worked for what i sent, it should work for what you send. [16:03] it's not part of IRC--it's just a thing some clients do. [16:04] oh, ending slash. [16:04] like, s/coolio/foobar/ [16:04] hmm [16:04] s/hmm/fooo/ [16:04] oh nice [16:04] :) [16:04] thanks [16:04] haha [16:05] rick_h_: mind you, i'm happy to take the reward for most reused code. :-) [16:15] gary_poster: in Ubuntu I get the same results as you - so there is something not quite right there [16:15] hatch, :-/ ok cool that means good stuff to fix :-) [16:16] hatch: added two cards for the two things brought up in call [16:16] right - so was there discussion on a release date? I'm assuming that it was pushed? [16:16] hatch: the bundle button and mystery charm token event helper [16:16] thanks [16:17] Housemate wandered upstairs in his boxers with headphones in, I figured I'd spare you all. :P [16:17] :-) [16:18] haha - oh that would have been funny [16:19] hatch, we are making non-bundle release today. we will be able to demo bundle story 1 (display) & story 2 (deploy); we will have most of story 3 (stability) and maybe all of it. We will arrange with Jorge and Antonio to pre-populate approved bundles in store before we release. [16:20] coolio [16:20] hatch, for quickstart, we will have story 1 (equivalent to juju bootstrap --with-gui) in a PPA for testing, and maybe story 2 ready (adds ability to deploy bundles). story 3 (we handhold through init) we can maybe sneak in next week [16:20] that's 'bout it [16:21] is the quickstart stuff a separate ppa? [16:21] hatch yes will be for now [16:21] I noticed that it was python not go [16:21] ahh ok gotcha [16:22] frankban I will call project juju-quickstart? [16:22] on Launchpad [16:22] gary_poster: sounds good [16:26] I had lunch with a guy in Vanc who's using Juju in their startup [16:26] using it to deploy their app of course but also using it in CI like we do [16:27] her marcoceppi. we are making a juju-quickstart plugin. should we put it in https://launchpad.net/juju-plugins or make a new project? [16:27] cool, hatch! [16:28] yeah it was pretty cool talking to someone who's actually using it in production [16:28] http://www.picatic.com/ is the company [16:28] gary_poster: you're welcome to put it in juju-plugins or start your own project. I haven't gotten around to making packaging for juju-plugins, but that could be the fire under me rear to do so [16:30] gary_poster: the idea was to collect a lot of small plugins and put them there, to simplify packaging, etc. deployer (arguably a juju plugin), charm-tools, etc all have their own project. There's not forced convetion one way or the other [16:30] cool thanks marcoceppi, makes sense. wdyt frankban ? we have our own makefile stuff, so I'm thinking we should have our own project because that will be easier, at least for now. agree? [16:31] Also, I've come to talk about "Providers" on the charm info page https://jujucharms.com/precise/wordpress I've gotten some reports of confusion from users assuming that those providers listed are the ONLY supported providers for the charm [16:31] gary_poster: agreed [16:32] gary_poster: for now a separate project is better imho [16:37] cool thanks frankban . https://launchpad.net/juju-quickstart [16:38] thanks gary_poster [16:38] marcoceppi, I just finally parsed what you were saying, sorry. ok... [16:38] You mean AWS/EC2 and LXc [16:39] gary_poster: yes! Sorry, always feel free to smack me for clarification :) [16:40] gary_poster: http://irclogs.ubuntu.com/2013/10/17/%23juju.html#t14:09 from earlier today, and jamespage had similar experience during a talk [16:40] marcoceppi, :-) . OK, I can understand that confusion. Would you mind filing a bug? I will if you need me to. If you have a suggestion for alternate wording, we'd really appreciate it. [16:40] gary_poster: pushed trunk to the new project [16:40] gary_poster: I'll gladly open a bug, just wanted to make sure it wouldnt' be in vain if you guys had something else planned [16:41] frankban, could you push it to ~juju-gui? [16:41] marcoceppi, no, this would be great to hear [16:41] thank you [16:42] gary_poster: lp:~juju-gui/juju-quickstart? [16:43] frankban, lp:~juju-gui/juju-quickstart/trunk , yeah [16:44] gary_poster: done [16:44] gary_poster: so lp:juju-quickstart will be lp:~juju-gui/juju-quickstart/trunk, correct? [16:45] gary_poster: the real question is, what shoudl I put into .lbox? [16:45] frankban, yes, lp:juju-quickstart is that [16:45] frankban, heh. make check? [16:46] gary_poster: make check is already in .lbox.check [16:46] oh [16:46] right, .lbox [16:47] .lbox now contains "propose -cr -for lp:juju-quickstart" [16:47] frankban, I think that is right [16:47] gary_poster: hum... perhaps we have to change the trunk series in https://code.launchpad.net/juju-quickstart [16:48] Do you guys have the concept of tooltips in the "store" portion of the GUI? [16:49] marcoceppi, no [16:49] frankban, looks right already to me [16:50] gary_poster: yes indeed, thank you [16:50] welcome :-) [16:54] gary_poster: https://bugs.launchpad.net/juju-gui/+bug/1241075 [16:54] <_mup_> Bug #1241075: Wording of "Providers" confusing on charm information page [16:56] thanks again! [16:57] marcoceppi, I like your solution. Note that we explicitly call OpenStack "HP Cloud" because of concerns from Antonio and others about a lack of portability across OpenStack providers. I don't think that affects your solution though. I'll ask Luca to look at this. [16:59] guihelp: quickstart branch ready for 2 review 1 qa. could anyone please take a look? thanks! [16:59] frankban, on it. [16:59] Makyo: ty [17:00] I'll take the other [17:00] gary_poster: this is fine, I think including that in the caveat would be fine too, "OpenStack is tested against HP Cloud's public OpenStack service" or whatever [17:00] cool marcoceppi [17:03] jujugui looking for a quick review/qa https://codereview.appspot.com/14702045/ [17:03] hatch, on it. [17:03] thanks [17:03] *lboxing-light* [17:05] what is a '360 feedback' ? [17:05] hatch: welcoem to the party! [17:05] rick_h_: lol I'm just doing the survey right now [17:06] I'm going to guess I don't do 360 feedback since I have no idea what it is? [17:06] hatch, code LGTM, qa in a sec. [17:07] hatch, it is our yearly review [17:08] hatch it is "360" because you review manager, peers, and reports (if you have any) [17:08] ohh ok [17:08] jujugui review+qa please: https://codereview.appspot.com/14771044 Please QA in a real environment as well as sandbox, incl. bundles. [17:09] Makyo: on it [17:09] it'll be a bit though - I need to fire up the juju-core box :) [17:14] gary_poster: reordered the quickstart cards: hight priority for minimum story one (normal for story one nice to have) [17:14] frankban, looks just right [17:15] gary_poster: cool, thanks and EOcrazyD. have a great evening! [17:15] frankban, :-) you too [17:15] thank you [17:21] Makyo: such a small diff for such a large issue :D [17:22] hatch, I spent alittle too long trying to figure out why the env wouldn't work if I had a breakpoint in there, sorry. [17:22] hatch, side effect from the recent refactor, though, which got the rest of the issues. Overall win,just this initial centering thing missing. [17:23] oh no that's fine - it just seems that every time we run into a canvas issue it's a lot of work to find the issue then a small fix :) I wonder if there is a way we can simplify it? [17:23] I really have no idea - you're the expert in this area :) [17:24] Makyo: so how should I qa this on a real env? [17:24] it's just pulling down your branch now [17:25] Deploy a service, drag it somewhere, and refresh, it should be centered. [17:25] cool [17:25] The issue was that if you refreshed, it'd be off the edge of the viewport. [17:25] ohh [17:25] yeah that's not ideal [17:30] jujugui, smallest review ever: https://codereview.appspot.com/14771045 . To qa, go to https://ec2-54-234-204-211.compute-1.amazonaws.com/ and verify in the inspector that sourcemaps are working for you. [17:30] on it [17:31] thank you [17:32] dunzo! [17:35] :-) [17:37] Makyo: sorry I'm having an issue deploying the GUI on core [17:39] it appears it can't install the python depts [17:39] anyone familiar with this issue or possibly a one off? [17:40] Not seen that before [17:40] ok will kick it off again [17:41] when I run `juju resolved -r juju-gui/0` the unit is still in error state [17:42] is this a juju bug? [17:42] er [17:42] known juju bug [17:42] the -r is retry [17:44] hatch, is this LXC on saucy? [17:45] LXC on..... [17:45] precise [17:45] hatch, LXC is broken right now with that exact symptom. The apt-get is failing, right? [17:45] right [17:45] yeah, hatch, LXC is hosed. panic in the streets, etc. :-/ [17:46] aww darn [17:46] hatch, ec2 or etc. to the rescue [17:46] ok Makyo looks like I'll need to ec2 this, sorry for the delay [17:46] gary_poster: thanks for confirming [17:46] hatch, nbd [17:46] welcome [17:46] * gary_poster goes for lunch [17:46] goes downstairs, that is [17:50] * hatch also takes lunch [17:59] * Makyo lunches too. Peer pressure. [18:15] just took the dog to get his hair cut.....so he decided to throw up on the truck seat.....aweeeeesommmeee [18:18]