[11:25] rogpeppe: good morning. today I am encountering this failure in the ws auth process: {"Response":"","Error":"machine Login not found"} [11:25] frankban: hiya [11:25] frankban: what request message are you sending? [11:26] {'RequestId': 1, 'Type': 'Admin', 'Request': 'Login', 'Params': {'EntityName': 'user-admin', 'Password': 'iperuranio'} [11:26] rogpeppe: ^^^^ [11:27] frankban: hmm, weird [11:29] frankban: i'll just finish submitting the branch i'm doing, then i'll have a look [11:29] rogpeppe: great, thanks! [11:44] frankban: right, branches submitted. will have a look now. [11:53] frankban: how are you running the API server? [11:54] rogpeppe: I am connecting to the bootstrap node of a bootstrapped juju env [11:55] frankban: ah, ok. i wonder if you could try bootstrapping (with --upload-tools) from this branch: lp:~rogpeppe/juju-core/225-api-log-messages [11:55] rogpeppe: aha! is --upload-tools required? [11:55] frankban: if you don't use --upload-tools, it'll use the binary that's in the cloud [11:56] frankban: that is probably the problem then [11:56] rogpeppe: ack. so, can I bootstrap from trunk or your branch is still required? [11:56] frankban: no, try bootstrapping from trunk. that should work. [11:57] frankban: in theory :-) [11:57] rogpeppe: thanks, do you have a min for two other questions? [11:57] frankban: sure [11:58] rogpeppe: what's, in your opinion, the best way for a charm (to be deployed in a gojuju env) to: [11:58] 1) obtain the bootstrap node address (in particular, the API ws url): I've seen there are some yaml files around where that value is stored [11:59] 2) check the type of the juju env (go vs python) [11:59] rogpeppe: ^^^ [12:00] frankban: good questions both [12:00] frankban: the answers will probably change over time [12:01] frankban: but for the time being, the answer is probably to look in the config file read by the machine agent [12:01] rogpeppe: our world is moving ;-) [12:02] frankban: which can be found in .... let me check [12:03] rogpeppe: maybe /var/lib/juju/agents/machine-X/agent.conf? [12:03] frankban: /var/lib/juju/agents/machine-*/agent.conf [12:03] frankban: snap! [12:04] frankban: which is a yaml file [12:04] rogpeppe: yes, cool, so I only need to figure out the number of the current machine [12:04] frankban: if that file exists, it's go juju [12:04] rogpeppe: sweet ;-) [12:04] frankban: there should only be one machine-* directory [12:05] rogpeppe: makes sense, thank you [12:06] frankban: presumably you want to find out the address of the API server? [12:06] rogpeppe: yes, apiinfo/addrs[0] I guess [12:07] frankban: yes, that's the one [12:08] frankban: although in the future it will a) change and b) become more than one address, so maybe choose a random address from apiinfo/addrs, or try them in sequence until you succeed. [12:09] rogpeppe: ack, thanks again [12:09] frankban: np; any time. [13:09] frankban, could you please double check my understanding of the backlog above by reading/editing the last bullet point from https://docs.google.com/a/canonical.com/document/d/1OEOzDu9lh4ko8oSgl_tjQlk98x_rgtiiSSJYBRopic8/edit ? [13:09] and hi btw :-) [13:11] hi gary_poster: that's also my understanding [13:11] cool frankban thanks [13:12] gary_poster: welcome [13:12] * frankban lunches [13:19] bac: what is the news on the API front? (I'm trying to remember, did you work yesterday?) [13:20] benji: i did [13:20] benji: want to g+ in a few minutes? [13:20] sure [13:33] bac, benji: if you want some input from me, just say the word [13:33] rogpeppe: thanks. i'm sure we will [13:33] rogpeppe: cool, thanks; we may do that once we have synced up [13:33] bac: i'm just putting together the example state/cmd branch. [13:33] rogpeppe: cool [13:35] actually, i've decided it's better named state/statecmd [14:00] good morning [14:03] rmorning [14:13] frankban: how did you solve the "error: use of closed network connection" error when destroying an environment? [14:13] teknico, frankban how are plane flights coming? ok, or problems? [14:14] gary_poster: we confirmed the flights, waiting for the tickets [14:15] cool frankban [14:17] benji: not sure, killed ec2 instances manually. I was then able to bootstrap juju arbitrary changing the control bucket [14:17] frankban: thanks (unfortunately, I don't actually have any running, I wonder if the control bucket is messed up) [14:19] hatch, what are you working on? That grunt bug would be great to have addressed if it is not already, and should be super fast. You are blocked on cards atm, so you could also consider the YUI app panel work. Ben had some thoughts on that so I'd like for you two to have a conversation about that before you start, either with me participating (preferred) or with me getting a summary of the result (fine, if that's t [14:19] he most convenient) [14:19] frankban: that was it, I specified a new control bucket and it worked; I guess reusing a py-juju control bucket confuses go-juju [14:20] the grunt is stop the line though, so someone should address that [14:20] benji: agreed [14:20] gary_poster: isn't that error related to what Kapil, Matt and I talked about wrt upping the inotify watch limit? or is it something else? [14:21] cat /etc/sysctl.d/10-inotify.conf [14:21] # expand inotify limit [14:21] fs.inotify.max_user_watches=16384 [14:22] benji, could you write down what you just learned in https://docs.google.com/a/canonical.com/document/d/1OEOzDu9lh4ko8oSgl_tjQlk98x_rgtiiSSJYBRopic8/edit please ? [14:22] gary_poster: sure [14:22] thanks [14:22] gary_poster: re flights, we need to know the hotel address in order to update/compile the ESTA to enter the US. [14:22] ack frankban: hazmat, ^^^ [14:23] bcsaller_, https://bugs.launchpad.net/juju-gui/+bug/1129464 ? Did not make connection. If that's the solution could you or hatch update the bug? [14:23] <_mup_> Bug #1129464: 0.4.x grunt branch breaks all builds < https://launchpad.net/bugs/1129464 > [14:24] Need to run to school to take son from dr. back in 20 [14:24] gary_poster: I'll add a comment, I think its related [14:26] gary_poster: notes added [14:29] gary_poster: yep I can fix that - is everyone ok with me restricting the version of grunt? [14:31] bcsaller_: sorry which error did I post to irc? I was kind of talking to myself while I was tracking it down [14:31] heh [14:33] hatch: related to EMFILE error when using fs.watch on many items iirc [14:34] ohhh yeah ok :) [14:38] ok working on the grunt bug [14:38] thanks hatch. I will add a critical card for you on kanban board [14:39] done [14:39] bac benji dragging your card to prototype (?) [14:40] which command are you doing? [14:40] gary_poster: juju get [14:40] cool, will note, thanks bac [14:44] bcsaller_: just to confirm it was this error... [14:44] fs.js:837 [14:44] throw errnoException(errno, 'watch'); [14:44] ^ [14:44] Error: watch ENOSPC [14:45] frankban: you succesfully bootstrapped an EC2 node using go-juju, right? How long did it take to go from bootstrap to "juju status" working? [14:45] hatch: yeah, I think the listed workaround... works around that [14:46] yeah looks like I don't even have that file [14:46] I wonder if we should add this to HACKING [14:46] benji: right, between ~five and ten minutes before juju status returning something. [14:48] hatch, bcsaller_ does moving back to grunt 0.3.x address? IIRC, the discussion was that this was a result of a bug somewhere in the stack. If so, it would make more sense to me to pin at 0.3.x and wait for a fix later, rather than requiring more system changes, however innocuous, in our HACKING. All that said, I don't feel strongly about it. [14:48] gary_poster: two separate issues [14:48] grunt needs to be pinned back - at least for now [14:48] hatch, so also occurs in 0.3.x ? [14:48] gary_poster: yes, but it needs to be pinned to 0.3 [14:48] ENOSPC error is also happening with 03.x [14:49] ok hatch. then HACKING doc is fine. thanks hatch, bcsaller_ [14:49] although adding bcsaller_ 's fix doesn't appear to have solved the issue....investigting further [14:49] the change to 0.4 will be its own story and should be deferred until the community has changed to the new api, for example spritgen might not work on .4 yet [14:50] rumor has it the community is pissed that the api has changed so dramatically [14:50] the community is a fickle bunch [14:50] on a dot release, yeah [14:50] ;) [14:52] * gary_poster thinks perhaps 0.x releases are an attractive nuisance for devs. they seem to imply to people working on the package that backward incompatible changes are ok, but for users that's never ok, or they expect something bigger than a dot release even of the series is 0.x. [14:52] s/even of/even if/ [14:53] bcsaller_: https://github.com/jashkenas/coffee-script/issues/2016#issuecomment-7499441 here is another possible fix, trying now [14:55] hatch: conceptually the same fix [14:55] hmm, was I supposed to do something after creating that file you mentioned? [14:56] you'd have to refresh sysctl or reboot, but yeah [14:56] oh well that's what I forgot [14:56] sorry coffee hasn't kicked in [14:56] heh [15:01] hmm [15:01] no luck [15:03] I'll try a reboot [15:06] anyone seen this one? "juju status" --> "panic: runtime error: invalid memory address or nil pointer dereference" plus lots and lots of other output [15:07] that looks fun :-/ [15:07] reboot solved it, I'll add to the hacking file with this fix [15:07] thanks hatch. of course, reboot may have also simply closed some open file handles... [15:07] but not worrying about it atm :-) [15:08] lol [15:10] benji, seen that one too [15:11] * benji waits with bated breath. [15:11] benji, no easy fix :-) [15:11] the traceback had lots of memory addresses, no identifiers [15:11] any hard fixes? ;) [15:11] I'm starting to have second thoughts :-/ [15:12] benji, tried again and it worked, no idea why [15:12] hmm [15:12] gary_poster: you were correct... [15:12] hatch :-( [15:18] at least rebooting vm's is really really fast [15:18] :) [15:21] well a couple reboots and it finally stuck [15:21] cat /proc/sys/fs/inotify/max_user_watches [15:21] 16384 [15:21] not sure why it didn't work previously [15:23] I like the new loading indicator [15:24] me too. Yay goodspud, greg and benji! [15:24] yay! [15:24] Yay! [15:24] it should say "Connecting to..." not "Trying to connect.." though.....think positive!!! [15:24] :P [15:25] heh [15:25] lol [15:25] Yay! (why are we cheering?) [15:25] obviously because you rock [15:25] because we like the new loading indicator [15:25] Yay! [15:26] And some people just like cheering :-) [15:26] Yay for cheering! [15:26] Hooray us! [15:26] lol [15:26] haha [15:27] jujugui call in 2 [15:30] bac hazmat starting without you [15:37] 2013/02/19 15:37:03 JUJU environs/ec2: opening environment "ec2" [15:37] 2013/02/19 15:37:04 JUJU environs/ec2: waiting for DNS name(s) of state server instances [i-92cb0ae1] [15:37] 2013/02/19 15:37:05 JUJU state: opening state; mongo addresses: ["ec2-50-19-23-159.compute-1.amazonaws.com:37017"]; entity "" [15:37] 2013/02/19 15:37:05 JUJU state: connecting to 50.19.23.159:37017 [15:37] oops, wrong chan [15:37] 2013/02/19 15:37:05 JUJU state: connection failed: dial tcp 50.19.23.159:37017: connection refused [15:41] benji: my go env: http://pastebin.ubuntu.com/1682925/ [15:41] hatch: could you link your branch to bug in LP? [15:44] doh [15:44] sorry one second [15:44] Makyo: "Delete cookies and other site and plug-in data" [15:47] bac did that work? I had to use --unchanged but I'm not sure if it actually pushed anything [15:47] hatch: you can also do it on the web site [15:47] ok I'll do that [15:48] hatch: not linked here: https://bugs.launchpad.net/juju-gui/+bug/1129464 [15:48] <_mup_> Bug #1129464: 0.4.x grunt branch breaks all builds < https://launchpad.net/bugs/1129464 > [15:48] * bac i *thought* lbox did it for you... [15:48] ok it's now linked [15:51] bcsaller_: did you want to chat? [15:52] hatch: yeah, let's see if the chat opened back up [15:53] it didn't, I can open another [15:53] sure, just invite me on G+ [15:53] new webcam hopefully arrives today so I don't need to switch computers all the time :) [15:56] benji: i'm going to grab lunch [15:57] bac: k [16:11] gary_poster: are you free to join this call on the panel? [16:12] yes bcsaller_ thanks. where? [16:12] gary_poster: should have an invite [16:19] bac: ok, I'm going to upgrade my development machine now. If you weren't following the discussion in juju-dev, you may need to do the same for the VM you are using [16:33] benji, you still there? :-) [16:33] teknico: yep (this iso download is kinda slow) [16:34] benji, here's a 32bit SSL MongoDb package: https://launchpad.net/~julian-edwards/+archive/mongodb/+packages [16:34] SSL-enabled [16:34] teknico: thanks. I don't think that helps though, I need the cloud-init script to know that it exists and download and install it [16:35] and even if that were the case, I strongly suspect that this sort of thing is going to come up again; I need to be on the same platform as the juju-core developers. [16:36] benji, yep, I'm going to install 64bit too at some point (and bac might want too) [16:36] benji, is there a way to upgrade from 32 to 64bit without reinstalling? [16:36] not that I know of [16:37] if the install CD offers me the choice I'll give it a try [16:37] benji, what's the "cloud-init script" btw? [16:38] (I think I'll need more than 4GB of RAM to have a workable 64 bit dev. env.) [16:38] It is the code that gets run when a fresh node is started. I don't know much about it. [16:38] I only have 4GB so I hope that won't be a problem. [16:47] benji, bac, gary_poster, teknico, anyone else: here's the first of two branches that will implement the juju set command in the API. [16:47] https://codereview.appspot.com/7326052 [16:47] rogpeppe: cool; thanks [16:47] ack will look soon thank you [16:47] this one just factors out the code [16:47] the next one will add the call to the API [16:48] unfortunately there were one or two things that needed fixing along the way, so the factorization is perhaps not quite as clear as it could be [16:48] but hopefully it shows the way nonetheless [16:51] ok guys, I'll be back in a while... hopefully [17:07] interesting parallels note - you need to have at least as much HD space free as you have specified for RAM for the VM [17:10] gary_poster: fyi - I left `sudo sysctl -p` out because it didn't appear to do anything on my system [17:10] is it possible there are any other flags required? [17:11] hatch AIUI it worked for bcsaller_ and is supposed to work generally. I suggest including it and then mentioning that it has not worked in some cases, and if problems are encountered a reboot should do the trivk [17:12] c [17:12] sounds good [17:24] teknico: any guesses as to which of those packages contains the i386 files i'm after? none of them look big enough (the binary size of the amd64 archive is ~40MB) [17:28] rogpeppe, I installed the four i386 packages, ~30MB overall [17:28] teknico: i can't use apt-get because i'm not on the right architecture [17:28] teknico: currently i'm unpacking the .deb files directly [17:29] teknico: but i'm not sure if that's the right thing to do [17:29] rogpeppe, there probably other actions performed by the install scripts in the packages [17:29] does `lbox submit` push as well? or do I need to propose -cr first? [17:30] hatch: yeah, it does push [17:30] thx [17:30] teknico: maybe. juju seems to make do ok with just using the binaries directly [17:31] rogpeppe, how is the SSL server certificate generated? [17:31] teknico: by juju bootstrap [17:31] rogpeppe, we got an interesting error in firefox after reboostrapping: [17:31] Your certificate contains the same serial number as another certificate issued by the certificate authority. Please get a new certificate containing a unique serial number. [17:32] teknico: ha ha. ok, i guess i should make the serial number more unique [17:32] teknico: please file a bug [17:32] rogpeppe, will do [17:35] rogpeppe, https://bugs.launchpad.net/juju-core/+bug/1130255 [17:35] <_mup_> Bug #1130255: Invalid SSL certificate after rebootstrapping < https://launchpad.net/bugs/1130255 > [17:37] teknico: thanks! [17:55] frankban, teknico, did you see the hotel email? [17:56] gary_poster: yes, cool [17:56] great [17:56] gary_poster, yep, thanks, adding the flight details to the wiki now [17:56] great thanks teknico [17:57] gary_poster: who do I email this link to? juju-gui-peeps? [17:58] hatch, whatchoo talkinbout? [17:58] the sub application development outline [17:58] hatch, oh! nah, juju-gui (public) [17:58] thanks [18:00] I don't think I have that email :) [18:00] it's probably a list i should be on though [18:02] hatch, https://lists.ubuntu.com/mailman/listinfo/juju-gui : juju-gui@lists.ubuntu.com [18:03] thanks [18:07] gary_poster: you might need to approve my sub request [18:07] I haven't received the confirmation email yet [18:07] hatch don't think so bu twill check [18:07] andddd there it is [18:07] lol sorry [18:07] :-) np [18:10] ok link sent [18:11] hatch could you make editable pls? [18:12] hmm [18:13] * hatch googles how to make editable [18:13] hatch, share [18:13] top-right button [18:13] done [18:13] sorry I didn't see the 'see' not 'edit' part [18:14] cool thanks hatch. Do you want edits or comments? [18:14] Easier to see edits but then you have to integrate [18:14] I mean easier to see comments [18:14] but then you have to integrate [18:15] umm lets to comments, I think it's easier for people to see then [18:15] cool [18:15] do* [18:37] hmm, did I write any part of this doc correctly? lol!! [18:41] hatch, you did write big chunks of it correctly! I'm an editor, sorry. A lot of the stuff I added was correction. [18:41] * gary_poster needs some lunch [18:42] hatch I meant that a lot of the stuff I added was clarification [18:42] * gary_poster needs some lunch in order to clear his head :-P [18:44] bac, gary_poster: here's the second part of the "juju set" change, that actually exposes the functionality in the API [18:44] https://codereview.appspot.com/7305101 [18:44] hopefully that will work as a reasonable example. [18:45] rogpeppe: thanks [18:45] i'm off for the day now, have a good r.o.d. all [18:45] gary_poster: haha that's fine :) I'll just take lunch now and then I'll add the changes you commented [19:13] I'm alive! === Guest29269 is now known as benji [19:14] phew! [19:15] * benji pulls off the slimy goo that accompanies reincarnation via a celestial portal [19:16] I always envision reincarnation on demand to be more like those little pods in the Matrix [19:16] although it probably woudln't dump you into a sewer after [19:16] that would kind of defeat the purpose [19:16] ;) [19:17] heh [19:26] gary_poster: whenever you return - I have updated the document and left comment replies where more clarifiation is required [19:26] well, benji's reincarnation didn't last long. [19:26] looks like the sewer got him [19:29] so benji you all shiny 64 bit? [19:32] bac: yep; many little things to fix though. For example, no system beep so I can't hear when people talk to me on IRC [19:33] oy [19:33] yeah, i've spun up a new vm [19:33] oddly, alt-tab doesn't work either; that's a new one [19:33] for some reason i wrongly thought vmware only supported 32bit [19:34] have you been able to bootstrap an environment? [19:34] (I haven't even tried yet, still getting this thing into a usable state.) [19:35] benji: still getting stuff [19:35] * gary_poster not back yet [19:35] i'm surprised i was able to torrent an ISO as quickly as i did [19:36] * bac hearts having a subset of /home/bac in a bzr branch for easy replication... [20:07] bcsaller_, hatch I have to watch the tablet video and then I can have a call anytime :-) [20:07] tablet video? [20:08] hatch http://www.ubuntu.com/devices/tablet [20:08] * hatch watching video [20:14] wow [20:14] "take my money!!" === benji is now known as Guest47530 === benji___ is now known as benji [20:21] guihelp is there a cause for the gui just showing up blank ala http://i.imagebanana.com/img/bx7q184n/Selection_001.png [20:22] hazmat: that happened to me before when it coudln't connect the websockets [20:22] a quick restart of the servers and it hasn't happened since [20:22] hazmat: I don't have any particular knowlege of that happening, but it looks like the background that we use to hide the app while logging in/connecting isn't getting hidden [20:22] hazmat, yes, what hatch says. If we make a release, message on screen now says that [20:23] hazmat, suggest looking at network tab. I bet they will see ws requests failing to connect over and oevr again [20:24] hatch, bcsaller_ no rush but let's meet in juju gui when you are available [20:24] Ready when you are [20:26] joining now [20:55] benji: have you gotten your juju-core environment back up? [20:55] bac: very nearly, but I am about to knock off for the day (I didn't take lunch today because of all this) [20:56] benji: i'm trying to install juju-core and getting: [20:56] bac@quantal64:~$ go install -v launchpad.net/juju-core/... [20:56] work/src/launchpad.net/juju-core/state/api/apiserver.go:4:2: import "code.google.com/p/go.net/websocket": cannot find package [20:56] this is with a fresh juju-core trunk [20:57] hmm, no idea [20:58] bcsaller_, hatch, everything resolved and the world is a beautiful place? [20:58] rainbows and butterflys [20:59] I'm just reading the D3 module doc/code to see how we can reuse it like bcsaller_ mentioned [21:00] ok cool hatch thanks [21:05] bac or benji, I want to contact Roger about the details about the pyJuju/Go Juju interaction issues. Could either of you supply them for me? I think you said that you had to change your bucket name? [21:05] gary_poster: right, I just tacked "-go" on to the end of the previous bucket name; Brad did something similar [21:06] benji, control-bucket in environments.yaml? [21:06] gary_poster: yep [21:06] gary_poster: yes, in environment.yaml you have to make the bucket name different from the one used by pyjuju [21:06] benji, bac, cool thank you both [21:06] gary_poster: yep [21:06] yep [21:07] benji bac, sorry one more thing. what's the error symptom? [21:08] gary_poster: I put a copy of the error in that google doc you pointed me at [21:08] oh cool! thanks benji [21:11] benji: i'd forgotten to install hg and git. 'go get' silently fails and that's why the packages were not found. [21:11] heh [21:11] bac: yay! I have a bootstrapped environment! [21:12] it's one thing to let you shoot yourself in the foot but another to plant land mines. [21:12] benji: yay. i will before the eod [21:13] heh, yeah [21:13] that said, i'm breaking to walk the dog [21:14] I'm going to stop on that positive note. See you all tomorrow. [21:17] See you [21:21] Makyo, bcsaller_, hatch, two of us should review Nicola and Francesco's branch https://codereview.appspot.com/7301105/ . I'll take one, but then we need another. (If we get two volunteers I'll happily give up my spot ;-) ). Could at least one of you review their branch before EoD? [21:22] gary_poster: I'll look at it now [21:22] I'll be responsible for QA, unless Makyo has a GoJuju environment ready [21:22] ok thank you bcsaller [21:22] No gojuju yet, sorry :/ [21:22] s'ok [21:22] I don't think I have Go running [21:22] Go is only a deb away [21:22] I can put some time into it though if you need [21:23] but GoJuju is more involved. Naah, s'ok hatch. hatch, let's talk tomorrow morning about timeline for what you are tackling with the app, and whether we can divide it up. [21:23] but meanwhile, explore :-) [21:24] sure thing - I've read through the d3 component docs and code and now I'm reading through the Y.App.Base code [21:24] I think Y.App.Base provides us with what we need already [21:24] cool [21:32] re: the sub-application aspect? [21:33] yes, Makyo [21:33] gary_poster, Ah, rock on. [21:33] :-) [21:36] I suspected but did not have proof that 'lgtm' as 'not lgtm' are actually official rietveld signals [21:36] but now I see http://code.google.com/p/rietveld/source/detail?r=756dc683ac1060be54e23e6f913b258ffd1582b8 [21:37] I'll suggest we change to these codes [21:37] gary_poster: bcsaller_: can you check the 'coding' section of the google doc - I feel that these are the steps that we need to take - everything else is handled automatically by the core YUI code [21:43] hatch only one other comment, but +1 generally [21:45] GREAT [21:45] oops [21:45] :) [21:46] :-) [21:49] hatch: so does that imply you'd rather not base it on the component framework? [21:50] yep - and the only reason is that Y.App.Base has almost everything we need with the exception of the couple additions I mentioned - but maintains a consistant api for working and interacting with views like you normally would [21:50] things like 'showView', app container event bubbling [21:50] 'createView' [21:51] 'activeView' [21:51] these are things that will be benneficial as the complexity of these sub apps will (no doubt) grow into [21:51] activeView is a little sketchy in the namespaced world anyway [21:51] it will be the active view of the sub app [21:53] if we extend Y.App.Base then we can pass it to anyone and point them to the YUI api (with a few additions) and they can get up to speed really quick [21:53] the only reason we have to disable router and pjax in the first place is because Y.App was never built with sub views in mind -> I'll probably work on a core patch at some point to split that out