[00:49] wallyworld_: morning, got time to catch up with a hangout? [00:49] wallyworld_: it seems we have some critical work to do [00:49] ok, can i have a few minutes, just finishing something [00:50] ack [00:57] axw: morning [00:57] thumper: heya [00:57] have a nice weekend? [00:58] too short... [00:58] I forwarded you an email [00:58] and what I expected has come to pass [00:58] mmkay [00:58] so we need to catch up on a hangout with wallyworld_ [00:58] sounds ominous [00:58] :-) [00:58] not as bad as it sounds [00:58] but more organising work [00:59] ok [00:59] * wallyworld_ almost ready [00:59] can it wait 30 mins or so? [00:59] considered "critical, drop what you're doing" type work [00:59] aha [00:59] um... sure [00:59] I could catch up with wallyworld_ about some other work [00:59] a hangout we didn't get around to last week [00:59] but it would make sense for me not to repeat myself regards this other work [01:00] damn english [01:00] never mind, just give me a couple of minutes [01:01] ok [01:01] ok ready when you are [01:02] wallyworld_: ? [01:03] okaaay [01:04] axw: wallyworld_, https://plus.google.com/hangouts/_/76cpj05oqcceb2k1t0qdjatpr8?hl=en [03:10] thumper: the way I'm doing the plugin at the moment will have the machine started up with a juju-db. is that okay? IIANM, the restoration requires it to be there already [03:11] axw: that seems ok... [03:12] thumper: cool. just need to find a not completely horrible way of getting StartInstance to go half way [03:12] :) [04:30] * thumper waits for jam to wake up [04:30] * thumper looks at the world clock [04:30] hmm 8:30am local [04:59] fuck it. short term hack time [04:59] such a mess [05:07] thumper: /wave [05:07] I'm not officially started, but since you poked [05:07] jam: when do you normally officially start? [05:08] about now, but I have to go pick up my car from the repair shop [05:08] jam: ah, I'd like to have a hand-off call [05:08] for the critical work status [05:09] thumper: I've got time for a handoff call [05:09] ok [05:09] * thumper kicks a hangout [05:10] jam: https://plus.google.com/hangouts/_/72cpik63e8i2q1m4eu10q73qh4?hl=en [05:25] axw: wallyworld_: before you stop today, can you give updates on where you got to with the various backup plugins? In case we need to hand it off to someone [05:25] jam: certainly [05:25] yep, was going to :-) [05:57] wallyworld_: do you know a James Price from when you were at Caterpillar? [05:57] not offhand [05:57] what section was he in? [05:57] not sure actually [05:58] he's my cousin, used to work there [05:58] caterpillar has almost 100000 employees :-) [05:58] heh :) he's in Perth, just thought you may have known him since you did some owrk over here [06:07] wallyworld_: I'm around if you want to chat [06:07] jam: ok, for 10 minutes then i have to go [06:07] np [06:48] davecheney: so why do you need the 4.9 snapshot for gccgo? I guess the default gcc for T is 4.8.? Is there just not support for golang there, or its just broken, or ? [06:49] jam: 4.8 does not support the architctures we need [06:49] davecheney: isn't that going to be true for the platform as a whole ? [06:49] so if we want any Ubuntu on those platforms for T then we need the compiler [06:50] jam: to your first question: not necessarily [06:50] to your second: I have no idea [06:50] arm64, no, that works with 4.82 [06:50] for the other platform, i have no idea [06:51] but I would place a small wager that 4.8.2 as shipped my us in T will not suffice [06:52] davecheney: anyway, my point is that if 4.8.2 isn't going to compile for the platforms we want to support for golang, isn't that going to be true if we want to support C/C++ packages on that platform? In which case, there needs to be some sort of exception that needs to be made which isn't specific to us [06:54] jam: https://docs.google.com/a/canonical.com/document/d/1ip_WmLusPBRqtxZG2rjANbnaiig68hkZENKj1DEOmKg/edit [06:56] jam: hopefully this explains, at least in part, why gccgo != gcc [06:56] davecheney: yeah it does [06:56] davecheney: I'm not sure that the compiler you build a tool with needs to be in main, are you sure on that? [06:56] I know we have a bug where backports are a problem [06:57] jam: mathias and james page both tell me this is true [06:58] davecheney: https://wiki.ubuntu.com/UbuntuMainInclusionRequirements section 6 [06:58] > Could you clarify for me, if we want to use gccgo to build juju into [06:58] > main, does that gccgo compiler also have to be in main ? [06:58] yes. build dependencies of packages have to be in the same pocket. [06:58] "All build and binary dependencies must be satisfyable in main" [06:58] ^ from the man [07:01] davecheney: yeah, I found it on the official "What do you need for a Main Inclusion Request" page. [07:04] jam: this is quite a quandry [08:15] jam, morning [08:16] fwereade: morning. I'm in a hangout with Dimiter right now, but I know Tim really prefered the Juju process than the MaaS hackery. Partially because it works on any provider [08:17] jam, and it's fully achievable... today? [08:17] fwereade: well Juju-backup and juju-provision have been written [08:17] whether "they all work" yet [08:17] I'm not sure [08:19] fwereade: but we need them tested anyway [08:24] fwereade: I'm chatting with dimitern about having him pick up bug #1254577 [08:25] jam, I'm really worried that this is bad craziness and way too many moving parts, when what was asked for was "a procedure to back up and restore the bootstrap node", and we know they already have file-level backups of everything [08:26] fwereade: but are they file-level backups that are consistent mongo snapshots? [08:26] fwereade: hangout? [08:26] jam, yeah, sgtm [08:26] fwereade: https://plus.google.com/hangouts/_/calendar/am9obi5tZWluZWxAY2Fub25pY2FsLmNvbQ.6dasi8b7b79tgd1ij533qo0ma8 [08:28] jam: fwereade : let me know what you decide [08:29] wallyworld_: you can join the conversation if you want [08:29] ok [08:34] wallyworld_: you made some changes to stop the lxc provisioner from starting unnecessarily, right? [08:34] axw: yeah [08:34] only starts when first lxc machine is asked for [08:35] s/machine/container [08:35] cool. someone on #juju just reported a problem with manual provisioning, it was barfing because lxc-ls couldn't be found [08:36] right. so cloud init no longer apt-get installs lxc by default [08:37] axw: so I think you need "juju-local" vs just "juju" if you are doing things like manual work [08:38] jam: for the local provider, but not for manual [08:38] ah [08:38] well, until wallyworld_'s change I suppose it should have installed it, yeah [08:38] oh well, FIXED_UPSTREAM [08:39] axw: did i break something? [08:39] wallyworld_: no, you fixed it [08:39] \o/ [08:56] mornin' all [08:59] morning rogpeppe [08:59] axw: hiya [08:59] morning rogpeppe, currently on a hangout talking about how we're going to deal with the NEC stuff [09:00] https://plus.google.com/hangouts/_/calendar/am9obi5tZWluZWxAY2Fub25pY2FsLmNvbQ.6dasi8b7b79tgd1ij533qo0ma8 if you want to join [09:00] jam: joining [09:59] morning all. [10:00] morning mramm [10:00] morning mgz [10:00] morning [10:00] how are you jam? [10:01] a bit talked out, as we've been in "we need to fix the backup process" discussion overload [10:02] I saw Ian's branch [11:38] jam: hey - who's the best person to talk to about what juju needs in this stripped down mongodb package I'm about to hack out? [11:39] jamespage: we're a bit on critical response today [11:51] jam: natefinch: i've pushed the latest changes to the backup script. https://codereview.appspot.com/31960043 it still doesn't include the jenv info eg uuid. nate suggested mongo --eval but i don't know how far i'll sensibly get with that tonight. i *could* use juju get-env from the client side part of the backup script and include the resulting json output in the tarball. but if we can generate on the server side that would be better [11:52] /join #maas [12:07] wallyworld_: already looking too [12:18] rogpeppe, you're using 1.16 branch for the ec2 tests, right? [12:18] dimitern: yeah [12:18] dimitern: but the latest 1.16 branch, probably not the one that they're using [12:18] dimitern: i don't think it matters *too* much for these tests though [12:19] rogpeppe, ok, I'm pulling 1.16 now and starting [12:19] dimitern: i've written down the set of steps to take; now i'm deploying an environment, following the steps [12:19] dimitern: wanna join the hangout? [12:19] rogpeppe, i'm about to yeah [13:04] rogpeppe, hope you had a good weekend? would you be able to spare me 10-15 mins this afternoon (probably after 4pm) to talk about the api [13:05] mattyw_: had an excellent weekend thanks [13:05] mattyw_: hopefully, yes [13:05] mattyw_: ping me === gary_poster|away is now known as gary_poster [14:07] abentley, CI is ill I failed to fix it on Friday. I have not seen 1.17 pass in days and we need to do a release. I removed azure from testing, and that revealed the disk space. I suspect the space issue is why I saw 3 instances running for 19 hours when we expected no more than 30 minutes [14:07] abentley, I did update the assemble script to download less [14:07] sinzui: Okay, I'll get on it. [14:10] abentley, I hope to work with utlemming in a few hours to deploy the real set of tools and mirrors to streams.canonical.com. I think it is realistic to say by our mid-afternoon, we will see production and code ready for the release. [14:18] sinzui: tests running. [14:18] with my new script? [14:21] jam, hazmat, I have an agreement/plan to fix Lp's listing of releases on +downloads. I will propose a fix later this week. [14:21] sinzui, cool [14:26] sinzui: I didn't update the script. [14:33] sinzui, is manual provisioning on the automated tests? [14:33] no, net yet hazmat [14:33] sinzui, there was a regression on trunk for it recently. [14:33] hazmat, bug that I can track? [14:33] sinzui, https://bugs.launchpad.net/juju-core/+bug/1254642 [14:33] <_mup_> Bug #1254642: manual provider configures API Info with state server addresses [15:05] I'm going to be variably around for the next while, working on ug 1254579 [15:10] hey all, can at least one of you join the new P1 conf call I am creating [15:10] sending out the details to everybody in email. [15:19] mgz: ping [15:23] rebooting [15:23] rogpeppe, mgz, fwereade can one of you give an opinion about Bug #1253576? As I do see relation errors all the time, I think the reporter didn't wait long enough to see the problem reported [15:23] <_mup_> Bug #1253576: Juju does not show relation status

[15:39] sinzui: if the hook is hung, then there is no problem, as we currently conceive things anyway [15:39] sinzui: hooks are allowed to take as long as they like [15:40] rogpeppe, thank you. In this case then, the charm needs to verify its hooks respond properly [15:41] sinzui: you mean that it's a charm testing issue? [15:41] yes [15:41] rogpeppe, I don't think this is a juju issue [15:41] sinzui: agreed [15:41] * sinzui moves bug [15:42] * rogpeppe goes to grab a bite to eat [15:42] sinzui, sorry, just commented, should have replied here, go distracted looking for another bug [16:02] yay sinzui [16:05] mgz: how goes? === adam_g_afk is now known as adam_g [16:39] rogpeppe, would you be free in 15 minutes? [16:39] mattyw: i could probably spare a little time, sure [16:39] rogpeppe, I'll try to not take up too much of your time :) [18:01] rogpeppe: how goes with landing your bits? [18:01] mgz: i might have been duplicating your work, i'm afraid [18:01] mgz: i couldn't contact you [18:02] mgz: so i went ahead and did the stuff to edit the config files etc [18:02] in a plugin command, right? [18:02] mgz: no, just as a standalone, but it could be a plugin [18:03] oh, okay, so not completely the same then [18:03] mgz: if i'd thought to do it as a plugin, i probably would have [18:04] I was assuming you'd be working on getting the machine listing working after putting a state server back online, so just have a lookup then ssh/sed replace/hup thing [18:04] mgz: this is what i've got currently: http://paste.ubuntu.com/6475020/ [18:05] mgz: there's not an enormous amount more in terms of easily scripted stuff [18:05] okay, that's too bad [18:06] mgz: sorry, i tried to ping you, 'cos i thought we could pair on it [18:06] yeah, I had to leave internet-world unfortunately [18:06] mgz: i should get your mobile number :-) [18:07] the other thing I wondered if it should be part of this command, or assumed to be done already, is the updating of provider-state [18:08] mgz: i'm not sure [18:09] mgz: in my mind that part of things isn't quite so fixed as to how we're going to do things [18:09] sinzui: ping [18:10] rogpeppe: have you got anything up for actually correcting the address of the state server in state yet? [18:10] mgz: the addresser should do that automatically, shouldn't it [18:11] hm, yes, though leaving it to that I find a bit scary, because presumably everything will start up with the wrong details till the update happens... may not matter [18:13] mgz: it seemed to work when i tried it [18:16] jam, fwereade: I have garage maas all set up, not sure if there's anything else I should be doing to help out? [18:16] natefinch: I think rogpeppe has set up some instructions, we should probably have you try them out [18:17] jam: sure thing [18:17] jam: my instructions are somewhat different to the ones that dimiter was putting together [18:17] jam: but are less ec2-specific so maybe useful anyway [18:18] jam, rogpeppe: let's start this way - what version of juju should I be installing? [18:18] natefinch: i used the latest 1.16 [18:18] natefinch: and a version with a patch to worker/provisioner [18:18] rogpeppe, natefinch: certainly a 1.16 version. The question from ehw was about using 1.16.1 [18:19] jam: think I figured it out; 1.16.0.1 shows up in the logs; not sure which jujud returns that version [18:19] natefinch: i bootstrapped with the current 1.16... or whatever bootstrap found, anyway [18:19] ehw: no official version. the .0.1 means it wasn't an official release [18:19] ehw: that means you've used --upload-tools [18:20] rogpeppe: which isn't *that* uncommon because of MaaS, people tend to do that instead of sync-tools, I think [18:20] jam: hmm [18:20] jam: they really shouldn't thoough [18:20] rogpeppe: mgz: well, I agree, but lack of egress means we exposed a way through (a jujud right here, use it) [18:20] at the least, it's probably an old local copy, instead of our latest 1.16 relase [18:21] jam, rogpeppe looks like it was deployed with ` juju bootstrap -v --upload-tools ` [18:22] although right before that, they did `juju sync-tools`, not sure why [18:23] ehw: did they try sync-tools but it failed because of lack of outbound network access? [18:24] jam: no error reported in the deployment doc; I'm going to add a comment to it and see [18:42] hi jam [18:46] mgz, ehw, rogpeppe: We never published a good document about collecting the tools and running sync-tools. I wrote a doc explaining how QA does it. We might use that doc as a base for official docs [18:46] sinzui: that sounds like a good idea [18:48] i'm going to have to leave soon [18:49] mgz: here's what i've got now FWIW - it does work, but feel free to discard http://paste.ubuntu.com/6475222/ [18:49] rogpeppe: thatnks [18:50] We haven't seen hp-cloud do a successful upgrade-juju since r2071. We might no care about this case since we will advise users not to upgrade envs and note that cli is incompatible with existing envs [18:51] g'night all [19:11] I'm off to bed myself, have a good night [19:12] sinzui: I was just pinging you to help get an answer for e-hw and tools [20:32] mgz: you around? [20:54] natefinch: yeah [20:57] mgz: are the backup & restore steps in here ready to be tested, do you know? https://docs.google.com/a/canonical.com/document/d/17Ougx-wbiUP8xsGskESyUX_bEbZApvXHHYibMBzDD0Q/edit [20:59] dimiter said he tested them, yeah [20:59] I'm not sure it covers *everything* [21:02] sinzui: you guys have any experience with https://wiki.jenkins-ci.org/display/JENKINS/Promoted+Builds+Plugin or something like it? [21:02] ^ abentley [21:04] rick_h_: No, but it might be something we'd use as we start to support per-environment testing. [21:04] abentley: k, thanks [21:29] natefinch, they should be -- the part that's not addressed is getting a backup of the full machine [21:29] natefinch, I'm afraid that part is down to whatever hackery matches your best guess [21:30] natefinch, "tar everything up and splat it over the new system" is inelegant but I think matches the reality pretty well [21:31] fwereade: cool. I'm waiting for garage maas to download all the ubuntu images.... it is distinctly slow. something on the order of 20 minutes so far. [21:31] fwereade: my previous virtual maas never completed setup somehow. Not sure why, and red squad is all european based, so I had no one to ask. [21:31] fwereade: so I started over === gary_poster is now known as gary_poster|away