=== franciscosouza_ is now known as franciscosouza [05:42] good mornings === daker_ is now known as daker [15:04] <_mup_> Bug #812343 was filed: formula exec env should include DEBIAN_FRONTEND=noninteractive < https://launchpad.net/bugs/812343 > [16:03] m_3: pong [16:04] adam_g: was just responding with the rabbit stuff [16:04] adam_g: I attached it to the bug report [16:05] adam_g: was able to reproduce your problem, then got it 'started' [16:05] adam_g: misbehaving package dep [16:06] adam_g: lemme know if that fixes it for you pls [16:09] rabbit is a strange beast [16:09] causes all kinds of problems for LP devs because of its insistance that its hostname never change. [16:10] SpamapS: I think it was just a startup script in a dependent package not playing nicely [16:11] m_3: what package? [16:11] don't know which one... [16:11] apt-get install -y rabbitmq-server didn't work [16:12] adding noninteractive and --no-install-recommends fixed it for me [16:12] m_3: i actually was using the non-interactive frontend for a while, but took it out to see if it made any difference [16:12] ill give no-install-recommoends a shot, thanks [16:12] erlang-base | erlang-base-hipe, erlang-asn1, erlang-corba, erlang-crypto, erlang-docbuilder, erlang-edoc, erlang-eunit, erlang-ic, erlang-inets, erlang-inviso, erlang-mnesia, erlang-odbc, erlang-os-mon, erlang-parsetools, erlang-percept, erlang-public-key, erlang-runtime-tools, erlang-snmp, erlang-ssh, erlang-ssl, erlang-syntax-tools, erlang-tools, erlang-webtool, erlang-xmerl [16:13] I can narrow it down to a particular package if this gets you past the hurdle [16:13] lots of dependencies from erlang-nox [16:13] yup [16:24] m_3: still no luck with that install hook and those apt-get options. [16:26] m_3: were you able to deploy to a completely fresh system? [16:27] hey guys, didn't notice you come in :) [16:27] it's the end of my day now but I'll be back shortly [16:28] my mind in in a maze of twisty little tests, all alike, and I'm not going to be able to resisit another look :p [16:29] adam_g: yeah, from totally shutdown to rabbitmq 'started' [16:30] adam_g: are you specifying the ami or anything in your environment? [16:30] m_3: yeah, an oneiric daily from last week. which are you using? [16:31] not specifying ami... defaults to natty [16:31] ami-06ad526f [16:32] which ensemble rev? [16:32] sec [16:34] fwereade: hey man [16:36] adam_g: yeah, just tested again and it comes up fine... only bzr diff with your branch is http://pastebin.com/76H4dsBF [16:37] adam_g: my env (less keys) http://pastebin.com/fL0fSn3r [16:46] m_3: heyhey :) [16:49] oh, dammit [16:50] I just spent half an hour reexamining my apparently-flawed understanding of mocker and twisted [16:50] turned out to be a lower-case 'r' in some_mock.callRemote [16:50] fwereade, they are tricky to combine [16:51] jimbaker: yeah, I assumed I'd encountered some subtlety that would teach me something [16:51] all it's actually taught me is lrn2type [16:51] fwereade: \o/ [16:52] RoAkSoAx: hey, how's it going [16:52] we've all been humbled by twisted a bit. :) [16:52] fwereade: good good [16:52] you? [16:52] RoAkSoAx: yeah, recovering, not used to this jetlag lark [16:53] fwereade: hehehe yeah!! fun times !! :) [16:53] fwereade, in zed shaw's learning python the hard way, learning to type (and the resulting debugging) is a key piece of the learning strategy [16:53] jimbaker: still haven't looked into that... I usually enjoy zed's stuff [16:54] jimbaker: first encountered him ranting about statistics in the ZSFA days :) [16:54] fwereade, some interesting insight. i was perusing it for teaching my daughter python [16:54] jimbaker: cool :) how old is she? [16:54] fwereade, she's 10. she has already programmed in C (albeit through the use of a visual editor) and squeak [16:55] she learned squeak on her own. C was for a robotics camp [16:55] jimbaker: cool, I have that to look forward to, Laura's only 2 :) [16:55] adam_g: they're coming up, but rabbit's not fully installed... still trying to figure out where it's going [16:56] jimbaker: I have some blurred photos of her waving a cuddly snake, and that's as far as I've got [16:56] jimbaker: My 8 year old found Scratch this last weekend. He had quite a bit of fun with it. :) [16:56] fwereade, so zed shaw's premise is that those of who learned how to program by typing in code from a magazine/book is fundamentally different than the latter practice of doing cut & paste [16:56] Still its really hard to keep him focused on it.. its an unbelievable leap for him from that to what the xbox does. ;) [16:57] jimbaker: hmmm, yes [16:57] i don't know if he uses this analogy, but looking at code is like an expert chess player looking at a board. we can use pattern recognition to really grok what's going on - especially what's wrong [16:58] crap.. I suck at chess [16:58] _love_ scratch [16:58] jimbaker: nice [16:58] SpamapS, yeah, my daughter was playing with scratch over at a friend's house. what kids do when you're monitoring them, it's dangerous ;) [16:58] you're *not* monitoring ... [16:59] jimbaker: it'll be like wargames all over again [16:59] fwereade, indeed. i think there will be a tv movie about it [17:02] RoAkSoAx: hey, cool, glad you're coming in august [17:06] fwereade: mee too.. this week I'll be working on getting everything Clint' [17:06] did onto the ensemble bootstrap [17:06] and I guess in the sprint we'll figure final details [17:06] RoAkSoAx: awesome [17:07] RoAkSoAx: I have the final(!?) state of that branch in review again, and I'm starting to integrate stuff from our shared branch on top of it [17:07] RoAkSoAx: it crossed my mind that from what Gustavo said I shouldn't be pulling from or pushing to that branch directly [17:07] RoAkSoAx: so I guess it doesn't need to be shared [17:08] RoAkSoAx: but no harm done :) [17:09] fwereade: right, but well I guess that as long as nothing changes significantely and breaks the stuff i'm working on, there wouldn't be much difference on how e handle that [17:10] fwereade: but still I'd like to work with a branch that's in sync with your work [17:10] so I can be sure nothing will break significantly [17:10] RoAkSoAx: indeed not -- I just thought you should know [17:11] fwereade: :) [17:11] ;) [17:12] RoAkSoAx: the major source of confusion I think is that I'm having to use the twisted xmlrpc client, and test it, and so the code is coming out the other end a ...little less recognisable [17:12] RoAkSoAx: but given I have an environment, I think I can be pretty confident in it if it has the same results as yours [17:13] RoAkSoAx: that said it might be sensible to get you to review my (your) changes once they've passed through my hands ;) [17:15] fwereade: hehe yeah I was planning to send my changes for your review, as really, the main concern I have is just use all the goodness of your refactoring [17:15] rather than me putting all the code in one single function [17:15] :) [17:15] RoAkSoAx: as long as it's moderately spread out I'm not bothered :p [17:16] RoAkSoAx: just shout if you find you're ending up duplicating code [17:16] RoAkSoAx: which would imply a deficiency in itc current shape [17:18] RoAkSoAx: actually I had a question [17:18] netboot_enabled [17:19] acquire_system will only grab systems with it enabled [17:19] start_system always tries to set it to True [17:19] I presume we only actually need one of those, and I was wondering which you thought would be better [17:26] fwereade: let me check [17:31] fwereade: no I believe they should be kept [17:32] fwereade: acquire_system will obtain an available system [17:32] fwereade: while start system will tell it to start by issueing the commands, like containing the system through IPMI or stuff like that [17:33] RoAkSoAx: not sure I follow [17:33] if a system is already netboot_enabled, why would we set that to True again on start? [17:33] fwereade: acquire_system browses for a system that is under foo-available management class, and grabs one [17:33] similarly, if we can set it to True on launch, why exclude systems that don't have it set? [17:34] RoAkSoAx: but acquire_system also skips those without netboot_enabled [17:34] RoAkSoAx: ...unless I'm looking at the wrong branch :/ [17:34] fwereade: then start_system, tells cobbler to contact any configured device to actually power on the machine. Setting it to netboot_enable when acquiring the system makes sure to acquire one that has netboot enabled [17:35] fwereade: but when starting the system, it is making sure that netboot enabled is correctly set, otherwise the system willfail to netboot [17:35] fwereade: it is just a way of making sure [17:36] fwereade: cause let's say for whatever reason we acquired a system, and suddently someone changes the system profile on cobbler, then when ensemble tells it to boot the machine up [17:36] fwereade: the machine wont be able to PXE [17:36] RoAkSoAx: if we also _set_ it when we acquire the system, I'm happy, but as it is we just skip the system if it isn't already set the way we want [17:36] RoAkSoAx: it makes sense to me to set it immediately before launch, even if it seemed to be set ok before [17:37] RoAkSoAx: but if we _can_ set it, why would we avoid acquiring a system _without_ netboot_enabled? [17:38] fwereade: maybe it would be a system that we would like to not use for a moment but still keep under foo-available [17:38] fwereade: I know it does not seem unnecessary, but for me I think it is a necessar evil :) [17:39] RoAkSoAx: I'd be inclined to make in-foo-available the single point of truth for that [17:39] <_mup_> ensemble/expose-provision-machines-reexpose r292 committed by jim.baker@canonical.com [17:39] <_mup_> Merged trunk [17:39] RoAkSoAx: but the critical thing is that it is actually intended [17:39] RoAkSoAx: and I don't need to worry that I'm totally misunderstanding ;) [17:39] fwereade: yeha, but AFAIK I think we should not care that much right now about that, as it is super easy to override [17:40] fwereade: cause as I mention, i think we need better ways to ensure the interaction with the management classes [17:40] as I explained last week, we could set the management class from foo-available to foo-acquired and then the physicl machine might fail to start [17:40] while ensemble status still shows that machine as available [17:41] and doing the check, we can be assuming that "there's a reason by it has been set to disable. but let's just make sure we enable it before starting the machine" [17:41] s/by/why [17:41] RoAkSoAx: yep, makes sense [17:41] RoAkSoAx: thanks :) [17:43] anyway, I think that really is it for me now [17:43] gn all, see you soon [17:46] <_mup_> Bug #812441 was filed: ensemble ssh should passthrough args to ssh < https://launchpad.net/bugs/812441 > [17:51] <_mup_> ensemble/expose-provision-machines-reexpose r293 committed by jim.baker@canonical.com [17:51] <_mup_> Support reexposing a service [18:02] Hallo! [18:06] m_3: Nice blog post1 [18:06] ! === daker is now known as daker_ [18:26] m_3: SpamapS: *hand wave* who wants to go over the first cut of my ensemble slides? [18:27] jcastro: fire away [18:37] m_3: feedback on the NFS formula.. I don't think we should keep repeating the paradigm of using the remote service name for stuff. [18:38] m_3: I'd rather see us implement a mode where the consuming side suggests a name... that way two services can share the same NFS export. [18:38] m_3: we also need to put together some interface docs .. :) [19:05] m_3: nice blog post! [19:06] I wonder if it makes sense to do a debconf prompt when you install ensemble for the EC2 keys instead of punting the user to editing a dotfile right off the bat [19:20] niemeyer: thanks! [19:21] jcastro: are your latest on the U1 share? [19:23] SpamapS: yup... do you have any interfaces written up so far? [19:27] m_3: they are [19:27] seems like it's naturally part of the formula doc... we then have an 'interfaces' page on the wiki that scrapes formula docs from the repo [19:29] jcastro: used 22G on U1 already... gotta get it off the web [19:31] heh, nice! [19:34] jcastro: which talk you wanna go over first? [19:34] m_3: 5 minute is the same you saw, 10 minute is just the exanded version of that, it just needs a once over [19:35] Ensemble Long Presentation is the one I'm having a hard time with, it's like 30 slides but I am wondering if it needs to be so many slides vs. just doing a demo and going through it step by step [19:36] ok, I'll look [19:36] m_3: I just TODOed myself making a presentation of your blog post basically [19:36] it's nice to have something other than mediawiki to talk about, heh [19:37] m_3: no but I think we'll just use ensemble.ubuntu.com/Formulas/Interfaces/(interface-name-here) .. [19:37] m_3: I was thinking we can use a yaml format.. and leave it open to plugin test scripts that can be run at each point to verify an interface... but.. I wonder if we shouldn't just write them up in a way we can re-use for your cucumber idea. [19:39] jcastro: less slides unless you do them "Identity 2.0" style.. http://www.youtube.com/watch?v=RrpajcAgR1E [19:39] cucumber can consume yaml and adapt tests to each interface as nec [19:41] jcastro: I'll try to build out some other stacks and blog about them [19:42] jcastro: mediawiki full stack is still a very compelling xample [19:43] m_3: more compelling w/ config settings implemented. :) [19:43] jcastro: especially if we can connect other aux services we've developed [19:43] SpamapS: +1 [19:44] jcastro: ok, I personally don't like bullet points on slides... it automatically makes me want to sleep [19:44] m_3: watch identity 2.0 [19:44] best presentation I've ever seen [19:45] Its the top of the pyramid, but some elements fit where we will start.. in the middle. :) [19:45] jcastro: diagrams on 10min-page15 should maybe be flipped over (mysql on bottom)... totally minor though [19:45] m_3: someone from design will make us proper stuff so pretend the blocks and stuff look pretty (but noted) [19:46] wow I really like the animated parts of that... good story [19:47] hard to believe though... (everybody says what we're saying here) [19:47] Cloud 2.0.... [19:47] ;) [19:47] it's like... "no, really!" [19:49] jcastro: I'd expect we'll have different long-form talks depending on audience? [19:49] m_3: that's totally ripped off from Nick, but I have no shame [19:49] m_3: yeah so I was thinking of having fragments based on audience [19:49] and then swap out parts of the long one depending on audience [19:50] like, I don't think sysadmins will even want or care about the long presentation, I'd want to just see a terminal and someone typing. [19:51] jcastro:right, that's sort of the point of services though [19:51] they will care about the long presentation at oscon [19:51] lots of sysadmins and devs there [19:51] you used to have different perspectives on this [19:52] SpamapS: I look forward to stealing your long form one as well. [19:52] SpamapS: actually if you have it done and want to jet me a copy that would be swell [19:52] infrastructure meant very different things to different people [19:52] jcastro: I really like the idea around having 40-50 total slides, and just building presentations from those based on a few clues. [19:52] hahaha done hahahahaha [19:53] high levefolks thought of services, sysadmins thought of servers [19:53] we're bringning th two together [19:56] jcastro: lightning's look great, lemme see what sort of notes I can come up with for longform. [19:56] SpamapS: yeah, right now that's exactly what the 10 minute and long form one are, it's a shuffle of what I had around + more added in [19:56] m_3: you can plop your notes right into the slide notes if you want [19:56] definitely steal from SpamapS... [19:57] SpamapS: how long is your talk? [19:57] (going to be) [19:57] m_3: 40 min [19:58] I have approximately 2 slides done.. in outline form only. ;) [19:58] jcastro: there's the answer :) [19:58] nice [19:58] I actually plan to have only about 10, with a couple repeating, if I do it right. [19:58] that what mornings are for [19:59] SpamapS: let us know if we can help... (draw a diagram or something) [20:00] Heh.. this was my deck for the last Ubuntu Cloud Days.. http://spamaps.org/files/Ensemble%20Presentation.pdf [20:02] SpamapS: I just stole it, got any others around? [20:02] <_mup_> ensemble/expose-provision-machines-reexpose r294 committed by jim.baker@canonical.com [20:02] <_mup_> Removed debugging [20:02] SpamapS: great last slide [20:03] m_3: I think there are a few too many arrows and lines.. it was the first ensemble picture I painted. ;) [20:04] we'll just get someone to do a proper awesome diagram [20:04] once we get the content the way we want it [20:04] so hopefully it won't look so complicated, heh [20:05] robbiew: I'd like to go over these slides today with you too if that's ok [20:05] * jcastro is looking for any excuse to do a G+ hangout. :p [20:29] jcastro: heh...cool...got a 1:1 with jimbaker now...in about 30min? [20:29] whenever is fine for me! [20:31] <_mup_> ensemble/expose-provision-machines-reexpose r295 committed by jim.baker@canonical.com [20:31] <_mup_> Removed sleeps in ensemble.state.tests.test_base and added test for client being disconnected in a watch [20:32] jimbaker: ping :) [20:33] robbiew, hi [21:01] jcastro: yo [21:01] howdy [21:01] ready to go, when you are [21:03] robbiew: sure, wanna hang out? [21:03] er, "hangout" [21:04] sure [21:04] started [21:07] jcastro: so this is my big problem with G+ [21:07] I can never figure out how to hangout [21:07] beyond hanging with a whole damn circle [21:08] oh, are we hanging out? brt [21:08] j/k [22:07] jcastro: hrm, why doesn't u1's little messaging indicator tell me when you have shared something with me? In this case, it only told me that I had accepted a share.. :-P [22:07] seems rather silly [22:11] jcastro: "Ensemble Manges Services, not Machines" .. how baout "Ensemble Manages Services, not Servers" [22:12] ooh, I like that [22:12] SpamapS: I am heading to dinner, if you can just PM me your feedback that would be great and I can check it out when I get back [22:20] jcastro: Or in the morning.. no rush. :) [22:24] that reminds me... [22:25] im going to give a demo of ensemble at a local AWS users group. are there slides anywhere that might be of use? [22:27] I've got some [22:27] when is your talk? [22:30] jcastro: thanks! [22:30] tomorrow night :) [22:42] SpamapS: regarding your comment on https://bugs.launchpad.net/ensemble/+bug/810808 , is there a way to run debug-hooks so that it can catch exec of the install hook of a yet-to-be-deployed unit in time? [22:42] <_mup_> Bug #810808: Formula install hook never completes < https://launchpad.net/bugs/810808 > [22:44] adam_g: I think if you get debug-hooks going before the install hook it will work.. but that may be tricky as you probably can't run debug-hooks until ssh is working [23:03] SpamapS: ssh isn't the tricky part, but deploying the service and getting debug-hooks to run before the install hook fires [23:16] adam_g: looks like we need a --debug-wait option to deploy/add-unit [23:19] SpamapS: might be useful. i was actually able to catch the hook by looping while waiting on ec2 [23:20] and it turns out, if i run the install hook manually it installs rabbit, exits and continues onto start after [23:20] * adam_g puts on his strace goggles [23:21] Haha [23:21] * SpamapS has a vision of a gunner in a Mad Max / Waterworld tank preparing to mow down the unfortunate syscalls.. [23:23] hehe. [23:25] which process actually execs the hooks? the machine agent or the unit agent? [23:26] unit [23:26] machine execs unit agent..