=== menn0 is now known as menn0-afk [00:22] waigani: i'm looking at your style guide PR [00:22] have you pushed the latest version ? [00:23] davecheney: give me 5 - just about to push latest [00:23] davecheney: just writing an example [00:24] waigani: ok, protip, don't comment on the review saying "i've done this", unless you've pushed the code [00:24] it's confusing [00:24] * thumper agrees [00:25] sorry, i know this is yet another github fail [00:26] and it requires you to juggle responding to comments and pushing branches concurrently [00:26] but it is what it is [00:43] thumper: https://github.com/juju/juju/pull/244/files [00:43] i need advice on how to handle this [00:44] * thumper loks [00:46] davecheney: kick it up [00:46] davecheney: email alexisb and mramm [00:46] they will say what is good or not [00:46] I'm not sure what our agreement is [00:46] thumper: will do [00:46] same [00:51] school holidays are cramping my style [00:51] * thumper will be in and out for the afternoon === psivaa_ is now known as psivaa-off [00:56] thumper: consider yourself nagged about a policy of using the juju/errors package [00:57] davecheney: I just pass that nagging on to fwereade [00:57] :) [01:02] thumper: then everything is going swimmingly [01:02] heh [01:04] hey, being a reviewer does suck [01:04] you cant' edit the title or the labels [01:04] :( [01:09] davecheney: but its very much appreciated :) [01:11] davecheney: sorry for the confusion. It's up now. [01:11] waigani: oh man [01:12] you just undid all the stuff I said to drop [01:12] shit man [01:12] i wanna land this today [01:12] i'm going to write some comments [01:12] you're going to do what i write [01:12] then we're going to merge this [01:12] ok ? [01:12] davecheney: ok === menn0-afk is now known as menn0 === thumper is now known as thumper-afk [01:26] davecheney: I wish you hadn't hit merge on my PR... I was about to push an extra unit test (as per my comment) [01:31] menn0: sorry [01:31] davecheney: it just means I have to do another tiny PR with the test change [01:32] menn0: ok [01:32] davecheney: https://github.com/juju/juju/pull/273 [01:35] menn0: what does \d{14} do ? [01:35] i'm not used to that form [01:35] "\d" == "[0-9]" [01:35] 14 digits [01:35] {14} means "exactly 14 of the last thing" [01:35] \d will match utf8 digits as well, in most regex engines [01:36] you can also do {A,B} which means between A and B of the last thing [01:37] yeah, thati's what I thought [01:37] but how does \d{14} assert that it's 7 bit ascii [01:38] davecheney: the fact that we're checking the entire string exactly with a regex that only allows characters that are 7-bit ascii [01:38] I hadn't heard of the UTF-8 digits thing [01:39] it's not mentioned in the docs: https://code.google.com/p/re2/wiki/Syntax [01:39] menn0: change looks good [01:39] but the comment doesn't match the description [01:40] would it be simpler to say, the backup file must be juju-backup_NNNNNNNNNNN [01:40] or something like that [01:40] davecheney: yeah, I did consider that the comment could be confusing ... I just want to make sure that future changes ensure that the filename remains 7-bit ascii [01:41] maybe I'll add a more explicit test... [01:41] menn0: you keep saying 7 bit ascii [01:41] but don't you mean juju-backup_14decimals ? [01:41] davecheney: juju-backup_14decimals happens to also be 7-bit ASCII :) [01:41] that's what I'm getting at [01:41] obviously it's not clear enough [01:42] let me rework the PR a little [01:42] ok [01:42] there are 117 other valid ascii symbols that also fit in the 7 bit range [01:46] https://code.google.com/p/re2/wiki/Syntax has an [:ascii:] character class for that. [01:46] waigani: did you push that PR again ? [01:48] e.g. \d matches １ [01:48] [0-9] does not match １ === Guest41889 is now known as wallyworld [01:53] davecheney: bug 1336891 - can that be fixed by changing our code like in other cases so as not to have to have the backported compiler fix? [01:53] <_mup_> Bug #1336891: intermittent panic: runtime error: invalid memory address

[01:55] davecheney: look again please: https://github.com/juju/juju/pull/273 [01:55] wallyworld: no [01:55] wallyworld: we need the compiler fix to land [01:55] davecheney: I've removed mention of 7-bit ASCII [01:55] menn0: LGMT [01:55] davecheney: the test now checks what is actually important and only accepts a minimal set of characters [01:56] that amkes much more sense [01:56] davecheney: ok, i was asked to check because john "fixed" other issues like that by tweaking our code not to tickle the bug [01:56] wallyworld: matt sent me the dmesg output [01:56] which one of you guys is working on LXCs? [01:56] this is the runtime bug we were hitting several months ago [01:56] bigjools: that's thumper-afk [01:56] ta [01:56] davecheney: cheers. merging now. [01:56] davecheney: ok, thanks, will update the bug [01:56] wallyworld: are we talking about different bugs ? [01:57] davecheney: the bug i typed above [01:57] i just put an update on the bug i think we're talking about [01:57] wallyworld: same bug [01:57] see comment #9 [01:57] davecheney: ah sorry yes, my browser copy was out of date [01:57] thanks [01:57] :sad chrome [02:01] jrwren: are you sure \d matches unicode digits? http://play.golang.org/p/fyInDC26-W [02:02] jrwren: you've got me curious :) === thumper-afk is now known as thumper [02:33] bigjools: whazzup? [02:33] thumper: there's a new API call coming in the next release that allows you to get a static IP [02:33] maas release, I mean [02:34] so I figured you might want that for containers [02:34] bigjools: nice, but that is networking, so dimeter :) [02:34] heh [02:34] but handy, thanks [02:35] * bigjools watches hot potato flung to Malta [02:35] * thumper blows on his hands [02:35] phew [02:35] did somebody make some benchmarking into the lxc conatiners created by juju-local ? [02:35] *containers [02:35] sebas5384: at some time... yes [02:35] why? [02:35] thumper: hey o/ [02:36] because i'm running php using a drupal charm [02:36] and I did a benchmarking script [02:37] better say, i execute a benchmarking script in php [02:37] and it seems that its at least 4 or 5 seconds more than the host [02:38] which version? [02:38] http://www.php-benchmark-script.com/ [02:38] php5.3 [02:38] version of juju [02:38] same version in the two machines [02:38] ohh [02:38] the last stable one [02:39] 1.18? [02:41] that one [02:42] its strange because there are no constraints [02:48] davecheney: pushed again now [02:49] thumper: maybe is something about filesystem ... [02:49] waigani: ta [02:49] food, then review [02:50] sebas5384: the lxc containers are being treated exactly like machines, they do apt-get update/upgrade at the start (for that version) [02:50] sebas5384: not surprising they are the same [02:51] well yeah, but why then php is slower into the lxc containers :/ [03:03] davecheney, actionWatcher refactor related to our discussion yesterday : https://github.com/juju/juju/pull/275 [03:13] jcw4: ta [03:13] davecheney: pushing one quick change in 15 seconds === vladk|offline is now known as vladk [03:14] davecheney: okay; updated === jcw4 is now known as jcw4|afx === jcw4|afx is now known as jcw4|afk [03:38] jcw4|afk: sorry mate, thats a -1 from me [03:38] do i pass a 0 to ToMachineSpec in the ServiceDeploy api in order to have it automatically deploy to a new machine? [03:38] or do i need to add-machine and deploy to that machine via the api [03:40] stokachu: well, i'd start with "" [03:40] and i'd open a bug that the documetation on that method is crap [03:40] or will setting NumUnits to 1 make it deploy to a machine whether allocated or not? [03:40] ok lemme try these options first [03:40] then ill create a bug [03:41] stokachu: you cannot specify a number of units less than 1 [03:42] so right now if i set the numunits to 1 the ServiceDeploy will just deploy the charm without assigning it to a machine [03:42] new or one allocated (local provider) [03:42] stokachu: sorry, i can't help you [03:42] i tried reading the documentatoin but is it crap [03:43] lol it is kind of difficult to use the api with the docs :( [03:44] on a brighter note ive got a python3 juju client working pretty well https://github.com/Ubuntu-Solutions-Engineering/macumba [03:45] \o/ [03:45] gonna use this for the openstack installer [03:47] davecheney, hah, setting ToMachineSpec = "" worked :| [03:48] possibly if you don't supply that tag in the json message, you'd get the same effect [03:49] i think i tested that and it didnt work, lemme try again [03:50] ah my mistake that does work [03:50] davecheney, ty [03:51] ha [03:51] if cannot afford to provide ToMachineSpec then one will be provided for you. [03:51] lol === jcw4|afk is now known as jcw4 [03:59] davecheney: it sounds like my description of the solution is the problem right? [03:59] davecheney: let me try again and see if it makes more sense [04:01] ok [04:09] davecheney: https://github.com/juju/juju/pull/267 has merge conflicts [04:09] according to gh [04:11] waigani: ta [04:11] i'll repropose [04:11] someone must have landed something over night [04:11] not surprising [04:24] davecheney: I updated the PR message, should I update the actual git commit message with the same details too? [04:25] jcw4: nah, nobody reads those [04:25] well, maybe not nobody [04:25] davecheney: lol [04:25] but it's fn hard to get access to them [04:25] from the PR screen [04:28] jcw4: good PR message [04:28] davecheney: tx [04:30] my repl set tests are getting worse and worse [04:30] [LOG] 0:37.995 DEBUG juju.testing tls.Dial(127.0.0.1:47092) failed with dial tcp 127.0.0.1:47092: connection refused [04:30] replicaset_test.go:163: [04:30] s.dialAndTestInitiate(c, root, getAddr(root)) [04:30] replicaset_test.go:69: [04:30] c.Assert(err, gc.IsNil) [04:30] ... value *errors.errorString = &errors.errorString{s:"cannot get replica set status: can't get local.system.replset config from self or any seed (EMPTYCONFIG)"} ("cannot [04:30] get replica set status: can't get local.system.replset config from self or any seed (EMPTYCONFIG)") [04:30] never seen that error before [04:30] i assume it's a race setting up the repl set === vladk is now known as vladk|offline === vladk|offline is now known as vladk [04:39] tx davecheney ; I'll address the comments that I think are correct. If you'd like to review my responses on the others and let me know if they make sense to you? [04:40] jcw4: ok === vladk is now known as vladk|offline === jcw4 is now known as jcw4|ZZzzzz [05:04] davecheney: Backups in environment storage: https://github.com/juju/juju/pull/278 [05:05] davecheney: reviewing your PRs now [05:05] ta [05:15] davecheney: LGTM for #265 [05:17] menn0: ta [05:17] menn0: i added a lot of snotty comments on your PR [05:17] in general it looks good [05:19] davecheney: thanks for looking. I'm just responding now. [05:29] davecheney: I've responded to your snotty comments. Some answers from you are required. [05:30] davecheney: :-p [05:43] davecheney: thanks. any thoughts on the first comment? [05:55] menn0: waiting for you to push a new revision [05:56] also, i just fixed a bug in my PR [05:56] when someone calls [05:56] back.Backup("some password", "", ...) [05:56] that is user is "" [05:58] dimitern: around? === uru-afk is now known as urulama [06:01] bigjools, yeah, barely (need coffee :) [06:01] dimitern: well get some coffee and then I can show you Cool New Stuff™ [06:03] davecheney: I see the fix in your PR. Should that mongo connection field even be called Tag? [06:03] it's all about the user that's connection to mongo [06:03] is it really a tag? [06:03] s/connection/connected/ [06:05] dimitern: dogwalking, speak in a bit [06:06] menn0: it's not a user [06:06] well [06:07] you say user, you think human [06:07] but we say tag [06:07] as in 'some actor in the juju universe' [06:07] now, some tagable things don't make sense to have credentials to talk to mongo [06:07] like networks and environments [06:07] but machines and units do [06:08] menn0: i agree this is apeshit crazy [06:08] davecheney: I get what tags are about... I guess what I'm missing is what our mongo DB usernames look like [06:08] which is why i'm trying to fix it [06:08] "user-admin" [06:08] "machine-0" [06:09] "unit-mysql-1" [06:09] ^ these are mongo usernames [06:09] right. so we are using tag strings as mongo username [06:09] I didn't realise that. [06:09] it gets so much more horrible [06:09] "" < this is the admin user [06:09] your change makes sense now [06:09] davecheney: I've pushed updated revs for my PR [06:09] * menn0 nods. [06:11] davecheney: related topic: does it annoy you that it's Backup(password, tag, ...) and not Backup(tag, password, ...)? [06:11] menn0: no [06:11] it annoys me that it creats a temporary file on disk when this shoud return a io.Reader [06:11] but that is for another day [06:13] davecheney: shouldn't it /take/ an io.Writer instead of returning a io.Reader? I don't see how it can return an io.Reader without either holding the entire backup in RAM or using a temp file. [06:14] bigjools, sure [06:14] menn0: you can do it either way [06:14] you have goroutines, remember [06:15] but you are correct, passing the http response write to backup would be simpler [06:15] i was thinking some odball io.Pipe thingy [06:15] davecheney: that PR has an extra rev for a one line fix to the existing code that I noticed btw === vladk|offline is now known as vladk === vladk is now known as vladk|offline [07:14] morning === uru_ is now known as urulama [09:26] dimitern: so got 5 mins [09:26] ? [09:26] bigjools, sure [09:26] dimitern: I added an API in maas 1.6 so you can claim an IP [09:27] apparently this is useful to you? [09:27] bigjools, great! so we can reserve container addresses explicitly [09:27] yes [09:28] will point you at a doc, one sec [09:28] bigjools, how does it work? does the api call return the address it allocated? [09:28] yes [09:32] dimitern: so api calls are: [09:32] GET /api/1.0/ipaddresses/ [09:32] to list ones you reserved [09:32] POST /api/1.0/ipaddresses/ takes op=release or op=reserve [09:33] and I shall dig up more concrete representations of the data you get back in a moment when I can find the damn thing [09:33] bigjools, thanks [09:33] bigjools, is this advertised via the capabilities/version api call? [09:34] dimitern: good question [09:37] dimitern: oh HA, I already sent you an email on the 2nd about this [09:38] with the API description [09:38] dimitern: the latest beta package for 1.6 is in ppa:maas-maintainers/testing [09:38] bigjools, sorry, it seems I've missed it [09:38] subject "proposed release 1.6.0" [09:39] I'll forward it again [09:39] bigjools, i'll look into the api docs, cheers [09:39] well there's no docs yet, it's just an email [09:40] * bigjools EODs [09:40] bye [09:44] dimitern, wallyworld, jam, davecheney: any chance of a review of https://github.com/juju/charm/pull/15 please? [09:46] rogpeppe: looking [09:47] TheMue: thanks [09:52] rogpeppe: nice one, +1 [09:52] TheMue: thanks! [09:53] * TheMue uses the time for reviews when his tests run [09:53] jam: ping [09:53] TheMue: yes? [09:54] jam: I’m still at the bootstrap/endpoint issue, because I came a step nearer [09:54] jam: but I don’t find a solution, maybe you’ve got an idea [09:54] TheMue: so we've had a second report of it, so it does sound like something that is a reasonably serious issue in 1.20. It might be associated with some people saying it took *minutes* for mongo to initialize [09:55] jam: I can simulate it, when I consume some time between ensureMongo and openState [09:55] wallyworld and axw think that mongo delay is because of the oplog, which they had a patch for [09:55] TheMue: so that triggers it? [09:55] jam: yes, tried other cases, but no error. but when adding just some seconds between those too steps I get the same error on local [09:56] TheMue: so the fix (IMO) is that we shouldn't let things like the API actually respond to requests until we have a backing DB [09:56] so something like the "afterUpgradeStep" stuff does today. [09:56] though you might run it by axw or wallyworld since I think they've touched the code most recently [09:57] jam: thought in this directy too, yes [09:57] jam: ok, will contact them [09:59] TheMue: btw, congrats to Germany... I don't know how they smashed it so hard, but kudos [10:01] jam: thanks, we had a lot of fun yesterday evening, and have been astonished on our own [10:01] jam: I expected a close fight, maybe 1:0. but 7:1, wow === vladk|offline is now known as vladk [10:45] Wow, need a few seconds, a weird JS almost killed my machine. [10:48] vladk, jam, standup? === urulama is now known as uru-food [12:14] jam: never done this on LP before. how do I merge https://code.launchpad.net/~niedbalski/golxc/fix-1329049/+merge/225584 and are any steps regarding our dependencies required? [12:15] TheMue: I believe the process for golxc (not managed by a bot) is that you: "bzr co lp:golxc; bzr merge $THEMUESCODE; go test ./...; bzr commit -m "merge $THEMUES most awesome of patches" [12:16] TheMue: actually, the first step you need is to create a kanban card :) [12:17] jam: ok, ok, will do this now :D [12:20] jam: so, can continue, card exists 8) [12:21] dimitern, jam: interfaces file in MaaS: https://docs.google.com/a/canonical.com/document/d/18okVil9JGJgGxNlig5Gk8wwtNDSoXtT-Qta6B7gxTI8/edit# [12:22] vladk, thanks, will take a look === uru-food is now known as urulama === jcw4|ZZzzzz is now known as jcw4 [12:23] TheMue: there is a slight "it depends" which matters on how you have Bazaar set up locally. [12:23] but basically, you want to turn your $GOPATH/src/launchpad.net/golxc back into a pristine trunk [12:23] merge your code into it [12:23] run the test suite, [12:23] commit it, and then push it back to launchpad [12:24] vladk: thx [12:51] Hello [12:52] is it possible that juju don't use containers on openstack? [12:53] jam: thx for your help, e’thing worked fine as wanted [13:10] TheMue: great [13:10] Egoist: "is it possible" you don't have to use containers with openstack, though openstack itself is doing virtualization, right? [13:11] (IIRC, LXC is one of the possible ways that openstack creates virtual machines.) [13:19] katco: #1319475 [13:19] <_mup_> Bug #1319475: Juju should support new signing format

=== urulama is now known as uru-afk === viperZ28__ is now known as viperZ28_ [13:51] I just notices that I didnt see a few of natefinch comments [14:08] dimitern, jam : I sent you both a note on IPv6 I need a response asap [14:10] alexisb, sure [14:11] thanks dimitern [14:14] ok natefinch axw I think I addressed all the comments [14:16] perrito666: link me again? [14:16] wallyworld, ping [14:16] hi [14:16] natefinch: ln -s natefinch ./natefinch [14:16] natefinch: anything else? [14:16] https://github.com/juju/utils/pull/9/files [14:17] alexisb: i think i know what you're going to ping me about [14:17] Hi there wallyworld - well done to the team for whipping through a bunch of bugs [14:17] on this bug: [14:17] https://bugs.launchpad.net/juju-core/+bug/1338179 [14:17] <_mup_> Bug #1338179: juju 1.20.x slow bootstrap

[14:17] was there an actual fix? [14:17] given we couldnt repo? [14:18] alexisb: yes, the size of the mongo oplog file was reduced [14:18] (for local only) [14:18] are we sure that addresses the bug that was reported? [14:18] alexisb: it was deduced that that was the most likely cause of the extra 2 minutes of bootstrap time [14:18] alexisb: no [14:18] ok, thanks [14:18] but it was the most likely and plausible cause [14:18] just want to make sure I report back the status correctly [14:19] sure [14:19] wallyworld, I could use about 30 minutes of your time today if you can spare it [14:19] wallyworld: so it may also fix lp:1338511? [14:19] I need an education on bootstrap stuff for the SRE proposal [14:20] bug 1338511 [14:20] <_mup_> Bug #1338511: api-endpoints fails if run just after bootstrap

[14:20] TheMue: we are looking tat that now [14:20] alexisb: sure, any time [14:20] wallyworld: can simulate it on local when consuming some time between ensureMongo and openState [14:20] wallyworld, I will send you an invite [14:21] TheMue: thanks, we're looking to see if the previous fixes have helped and what else needs to be done [14:23] wallyworld: great, I’ve got to thank [14:23] TheMue: you mean the error occurs when you add a sleep between those two things? or adding the sleep stops it from happening? [14:23] I'm able to get the EOF just about time I bootstrap && api-endpoints [14:24] axw: on local I had it only once, since then no more. but when adding a sleep between those two I get it consistently [14:24] huh, weird. [14:24] TheMue: thanks [14:24] axw: yw [14:25] * TheMue listens and looks outside, we’ve got a warning for heavy thunderstorms [14:29] would love a quick LGTM on https://github.com/juju/juju/pull/163 -- it's a little overripe [14:30] ( mgz ?) [14:30] Now it starts [14:31] * bodie_ is jealous of TheMue [14:31] We had a great one earlier this week -- some of the largest bolts I've ever seen [14:32] one must have been about a block away [14:32] sinzui, when are we targeting the 1.20.1 release? [14:33] when the devs get their stuff together :-) [14:33] alexisb, Ahead of schedule...When I created the milestone, last week, I set 2014-07-17. [14:33] hopefully today we will have the final change committed [14:34] sinzui, I am not trying to push, I just want to set expectations from those asking [14:34] alexisb, I will release when CI passes the fixes. Maybe tomorrow, more likely Friday [14:34] we will not get the restore stuff done, but there's fixes that we must release asap [14:34] sinzui, ack [14:34] alexisb, We can say before Tuesday...and I may have jury duty that day [14:34] lol [14:35] wallyworld, is the plan to release the restore stuff in 1.20.2? [14:35] alexisb: yeah, unless we can somehow make it done for 1.20.1 [14:35] alexisb: ordinarily we could delay 1.20.1, but the fixes we need to release are critical [14:35] ok, cool, thanks for the info guys, sorry I am behind :) [14:36] wallyworld, I agree that 1.20.1 has enough critical fixes we need to get it out as soon as we can [14:36] alexisb: especially since the users can't even bootstrap a maas env etc [14:36]