[00:00] katco: thanks, sorry about that. little bit of rain and the network goes out [00:01] thumper: good time off? [00:01] rick_h_: yeah, a real nice break [00:01] rick_h_: didn't get all I wanted to get done done... [00:02] but them's the breaks [00:02] thumper: never do [00:02] the kids have decided they like D&D [00:02] lol [00:02] and I'm kinda required as a DM [00:02] which takes time... [00:02] but of course [00:02] thumper: awesome! [00:02] to do properly [00:02] axw: no worries at all [00:02] thumper: have time today for me to bug you with a few notes on things? [00:02] they were real frustrated that I had to read up on the campaign first [00:02] lol, yea been a long time here [00:03] I'd have to go reread a lot of stuff heh [00:03] rick_h_: yeah, at a cafe now so I can focus on design without kids around [00:03] the rules changed since I last played [00:03] over 20 years ago [00:03] thumper: nice :) I never had anyone interested in playing D&D growing up [00:04] thumper: k, ping if you get free/bandwidth and if not will bring it up later but had some things come up friday/today figured I'd mention [00:04] axw: same here.... somehow i was a huge nerd and never played even once [00:04] steve jackson books were there for me though [00:04] rick_h_: need to mention with voice? or irc would work? [00:04] thumper: we can try irc if you want [00:05] rick_h_: my mind is pretty good at reading your writing in your voice [00:05] lol awesome [00:05] can also pick up most of the sarcasm [00:05] :) [00:06] rick_h_: if we find it now working, we could do a call later [00:06] thumper: here, PM, or other side? [00:06] rick_h_: if it is sensitive PM or other [00:06] otherwise, I'm fine here [00:06] I don't think it is but do that just to be safe I guess [00:06] sure === kadams54-away is now known as kadams54 [00:13] katco: it's JUJU_DEV_FEATURE_FLAGS [00:13] you're missing the "DEV" [00:21] axw: thx [00:23] katco: left a few more comments. all the panics need to go, and init needs to move, otherwise trivials/suggestions [00:23] axw: yeah i was going to ask about the panics... wasn't quite sure what to do there [00:23] axw: i think i ran into a circular reference last time i tried to address the registration? i'll let you know here in a sec [00:24] katco: there should be no references from storage->openstack, only the reverse [00:29] axw: i see my misunderstanding... i think i was trying to register it in storage/provider/ something or other [00:38] axw: thanks for the great reviews. fresh PR up. [00:42] and like that you can ruin someone else's day bwuhahaha :) [00:42] katco: it was suggested that without the _DEV_ bit, it would be too tempting for clients to try to use them in prod settings [00:42] which they would do [00:43] thumper: i completely agree with that lol [00:52] thumper: I forgot a dinner tomorrow, can I move out sync back a few hours? [00:52] rick_h_: back as in later...? [00:52] thumper: yes [00:52] later in the day 3hrs [00:53] 3.5? [00:53] or is that too later? [00:53] that's peachy [00:53] ok [00:53] * thumper moves [00:53] aunt's b-day, not on my work calendar doh! [00:53] :) [00:54] katco: thanks, LGTM with a few more small fixes [00:54] axw: k thx... getting the dummy charm to allocate some storage will be a milestone [00:55] katco: did you see the email I sent out, with the reference to the hacked version of postgresql? [00:55] katco: or did you just want to test with something a bit leaner? [00:56] axw: just wanted to start lean so i can iterate quickly [00:56] sure, SGTM [00:56] axw: i will move onto postgres afterwards since i know that's what we would like to demo with [00:56] good practice to write charms anyway :) [00:56] axw: haha yeah :) [01:12] hey axw, do you know how to get the output of the cloudinit script like is in bug 1439447? [01:12] Bug #1439447: tools download in cloud-init should not go through http[s]_proxy

[01:13] cherylj: one moment, I think I know, just checking [01:15] cherylj: the instance's cloud-config is /var/lib/cloud/instance/cloud-config.txt [01:15] it's not exactly a shell script, but has each of the commands in a YAML file [01:15] axw: Awesome, thanks!! [01:15] nps [01:17] cherylj: BTW regarding your last comment on that bug: yes, there will almost certainly be people that want it either way. In particular, manually provisioned machines may need to go through a proxy [01:17] I think otherwise they'd be going direct tho [01:18] axw: yeah... kinda hard to deal with that where we have a mixed environent [01:18] unsure just yet [01:19] axw: Thanks, providing an option is proving to be a bit difficult to implement. [01:36] thumper cherylj: I was just thinking, don't we currently expect all nodes to communicate directly? for the API anyway? [01:36] so... maybe just disabling the proxy isn't a problem [01:37] I think even for manual we require things to be directly routable atm [01:37] that was my reasoning too [01:41] axw, thumper: cool, so I'll just disable the proxy when we're downloading tools for the non-bootstrap node [01:41] weird, just got disconnected [01:42] natefinch: I just assumed you were done :) [01:42] cherylj: sounds good. I can't recall if there's any case where a non-bootstrap node downloads tools directly from the Internet, but if so then don't disable for that [01:42] cherylj: I don't *think* there is such a case though [01:42] thumper: not quite :) [01:43] axw: we were looking at ONLY changing the curl command for acquiring tools [01:43] axw: yeah, I don't think there is [01:43] axw: there is a command line option '--noproxy=*' that /should/ work [01:43] thumper: yep, SGTM [01:43] cherylj: any luck getting it working from the machine not using cloudinit? [01:44] axw: did my last two comments go through, about the review? [01:44] natefinch: looking [01:45] axw: it was just 15 minutes ago... possibly I was already disconnected and the client didn't tell me [01:45] natefinch: last comment I saw from you was from 9.5h ago [01:45] ah nope [01:45] ok :) [01:47] axw: I tried moving the set password stuff in the agent code directly but I couldn't find the right time in the startup sequence where it would work. Did you have an idea of where is appropriate? I tried just above where I'd put the converter code, at the beginning of the state worker, and pretty near the beginning of all the workers. I got different errors for each, but unfortunately didn't copy all of them down. [01:48] natefinch: lemme see, one minute [01:54] natefinch: did you try modifyin OpenAPIState? add an "else" after the "if usedOldPassword", and call entity.SetPassword(info.Password) in it [01:54] modifying* [01:54] modifyin' [01:55] axw: nope, but that's a good idea [01:55] natefinch: I expect the StateWorker would bounce until that passes, but I think it should keep retrying? [01:55] axw: it should [01:55] Bug #1440940 was opened: xml/marshal.go:10:2: cannot find package "encoding"

[01:56] natefinch: initially I was thinking stateStarter could not start StateWorker until the API has connected, but I'm not sure if that'll cause problems with API server initialisation when there's only one server ... [02:01] * natefinch tries it. === kadams54 is now known as kadams54-away [02:03] thumper: It seems to be a problem using the * wildcard. If I do --noproxy *, it still tries to go through the proxy. But if I explicitly list the state server, it bypasses the proxy [02:03] haha [02:03] ha [02:03] hmm [02:03] I know what it is [02:04] oh? [02:04] the whole command line is being pushed through bash [02:04] * is expanded to all the filenames in the directory [02:04] I figured it was some substitution like that [02:04] try wrapping it in quotes so bash doesn't expand it [02:04] --noproxy "*" [02:05] that worked, yay! [02:05] w00t [02:05] let me try it in the actual code now [02:05] ok [02:11] thumper: it worked! huzzah! [02:11] awesome [02:12] I'm going to turn in for the evening and write the tests for the change in the morning. [02:12] kk [02:17] axw: when i do juju storage pool, i don't see "cinder" as an option... what have i done wrong? [02:17] axw: juju storage pool list rather [02:18] katco: you won't see any pools; there's an implicit pool for each provider, but it's not listed [02:18] (perhaps we should change that) [02:18] Added charm "local:trusty/hello-world-charm-3" to the environment. [02:18] ERROR cannot add service "hello-world-charm": reading pool "cinder": settings not found [02:18] [02:18] huh [02:18] that was from: juju deploy local:trusty/hello-world-charm --storage="foo=cinder,1MB" [02:19] katco: you *bootstrapped* with the feature flag enabled right? [02:19] yeah [02:20] export JUJU_DEV_FEATURE_FLAGS=storage [02:20] and then juju bootstrap --upload-tools [02:20] this is on canonistack [02:21] i'll try tearing it down and retry just in case [02:21] katco: that's weird, the error annotation implies that the error is not a NotFound error, but the cause suggests it is [02:21] katco: see storage/poolmanager/poolmanager.go [02:21] axw: k [02:25] katco: oh, there's a bug in provider/openstack/init.go -- not sure if it's related [02:26] katco: there's an existing call to registry.RegisterEnvironStorageProviders [02:26] katco: remove the first one, which is saying that openstack supports no custom storage providers [02:26] hmm don't think it's related tho, looks like it'll accumulate [02:29] * thumper heads home [02:30] axw: lol stale binary >.< [02:30] katco: :) [02:31] axw: so it's pending... it should eventually show up? [02:32] katco: yes, once the instance is created, the storage provisioner should try to create it [02:33] katco: we need to add proper status support to storage, it's a little bit difficult to debug what's going on at the moment [02:35] axw: agent is started, running hook config-changed [02:36] axw: and storage is still pending. hm. [02:38] axw: i'm getting some decent errors in debug-log at least [02:38] volume "0" not provisioned [02:38] getting storage source "cinder": requisite configuration was not set: auth-url not assigned [02:46] axw: ah i see... there are some config options i need to set to tell it how to authenticate to canonistack. where do i do that for storage? i.e. how does it get passed into VolumeSource(...)? [02:46] katco: hm, why are those pool config attributes? shouldn't they just come from the env config? [02:46] I think I glossed over that in my review [02:47] katco: sorry, in the VolumeSource method, it should be using environConfig to get the credentials to open a nova/cinder session [02:47] i'm getting them from *storage.Config [02:47] axw: ah k [02:48] katco: see ec2/ebs.go for inspiration if required [02:48] i shall seek inspiration :) [02:49] axw: fyi the whole provider/* vs. storage/provider/* screwed me up for a long time [02:49] axw: i kept wondering if you guys just hadn't checked in the ec2 provider [02:49] katco: sorry. the reason it lives in provider/ is because it's closely tied to the environ provider [02:50] non IaaS storage providers will go in storage/provider/ [02:50] axw: totally my fault [02:50] axw: i need to turn in, but this looks like we're pretty close. thanks for the help :) [02:50] katco: cool :) no worries, talk to you tomorrow [02:51] axw: have a good day! [02:51] cheers, good night === kadams54 is now known as kadams54-away [03:00] axw: btw, that suggestion to put it in OpenAPIState works great. [03:00] natefinch: awesome [04:00] axw: added some tests and finished the cleanup & suggestions you had. Would love to have you re-review. The tests aren't super thorough, but they're similar to what already existed, so I don't feel too bad (I do still feel bad, but I also need to get this committed). [04:00] going to bed. Good night all. [04:01] natefinch: yep, just about to hit the button. thanks for the updates [04:01] axw: thanks for the review. It's been a big help. [04:01] no worries [04:46] morning all === kadams54 is now known as kadams54-away === urulama__ is now known as urulama [07:34] morning all [07:40] dimitern: moeninf [07:40] dimitern: morning rather [07:41] ok, the coffee hasn't hit [07:41] voidspace, dooferlad, morning! [07:43] o/ [07:43] coffee is a good idea [07:48] morning o/ [07:57] o/ europe! [07:59] davecheney: heya, Mister Vienna [08:33] dooferlad, I've landed your branches, but we need to clean up a few bits [08:33] davecheney: what are you doing in Europe - just here early for the sprint or at a conf? [08:34] dimitern: yea, I spotted that. Thanks. [08:36] voidspace, I've tried to land yours but one had conflicts after the first one landed [08:36] dimitern: yep, fixed and retried [08:37] dimitern: just now [08:37] dimitern: creating versions of those fixes for master [08:37] voidspace, sweet! [08:37] dimitern: first one applied cleanly and PR created [08:37] dimitern: second one the patch didn't apply cleanly, looking at now [08:37] dimitern: thanks for landing the other one [08:38] voidspace, I'd appreciate if you have a look at bug 1439880 to see if my analysis is correct [08:38] Bug #1439880: Container's interfaces are all on private networks instead of host's eth0 network

[08:38] oh god [08:38] dimitern: ok [08:38] voidspace, that's most likely the same issue (not found subnets: []) [08:38] which you fixed already [08:38] dimitern: ok, I hope so [08:38] dimitern: looking [08:40] TheMue, heya, any update on your maas issue? [08:41] voidspace, LGTM on http://reviews.vapour.ws/r/1391/ btw [08:42] dimitern: thanks [08:42] dimitern: I'm an dialog with rvba, lp issue is created. now I at least want to propose my little fix that doesn't stop with an error but returns the default values instead [08:42] dimitern: yeah, those logs make it seem pretty likely that this is the same issue [08:42] dimitern: fetching an interface with no subnet and attempting to use that [08:43] dimitern: we'll see when he tries latest 1.23 [08:43] voidspace, yep, sounds good [08:43] TheMue, ok, but please follow through with rvba on the issue [08:44] dimitern: yes, sure, just added a comment [08:45] voidspace, dooferlad, and BTW I'd appreciate a review on http://reviews.vapour.ws/r/1390/ [08:45] dimitern: *click* [08:47] TheMue: the screenshot you pasted on bug 1439831 helped a great deal… but do you have access to the full console log? It's very possible that the cause of the problem could be identified if we had all the log. [08:47] Bug #1439831: Missing lshw breaks cloudinit [08:49] voidspace: sprinting with aram for a week [08:49] to get some arm64 stuff done before the close of the change window [08:49] davecheney: cool [08:49] rvba: I really would like to. sadly I cannot do a cut'n'paste with the vmware console at that time. as far as I know the vmware tools have to be installed at a later time [08:50] davecheney: Vienna is on my list of "must visit sometime" [08:50] davecheney, hey, so arm64 support will be released with golang-gc 1.5 ? [08:50] dimitern: enjoy the sprint [08:50] I wish :) [08:50] rvba: but I'll see what I can do [08:50] Thanks. [08:51] TheMue, have you tried using the vmware CLI tools to capture the console log of the VM ? [08:51] it should be possible [08:51] davecheney:: enjoy the sprint [08:54] dimitern: they can be installed when then the OS is installed [08:55] TheMue, no, I don't mean the vmware tools that you install on the vm, but the command-line vmware client you can use from the host [08:56] dimitern: never heard of it, have to search for it [09:00] TheMue, vmware fusion seems to come with a vmrun command - https://www.vmware.com/pdf/vix162_vmrun_command.pdf which has some interesting features, like running a script inside the guest and copying a file from guest to host - do you have vmrun? [09:08] dimitern: just took a look, seems to be in the application package and I only have to set it up. sounds like a good way, indeed. [09:11] voidspace: it's nice here [09:11] very layed back [09:11] no forms [10:21] dimitern: yes, arm64 will ship in go 1.5 [10:21] that's what aram and I are doing this week [10:21] davecheney, awesome! [10:22] davecheney, do you know anything about native ppc64 support in gc-go as well? [10:23] dimitern: same, ppc64 will ship in go 1.5 [10:26] davecheney, \o/ great! I can't wait not to have to care about gccgo/ppc64 bugs :) [10:30] dimitern: i'm looking forward to your support when I fight for moving everyone up to 1.4 next week [10:32] davecheney, count on it! :) [11:08] dooferlad, thanks for the review btw [11:08] dooferlad, as you progress with the implementation of "space list" you'll need to add a type for Space in params/network (like I did for Subnet) [11:13] dooferlad, or actually, hold on.. you won't need that - you already have all the info (e.g. list of all spaces - like AllSpaces in cmd/juju/subnet), and the rest should come from a list of params.Subnet for each space's associate subnets [11:13] dooferlad, most of the rendering code (and tests) could be reused between space list and subnet list, but I'd suggest copying it first and when done [11:14] dooferlad, ...refactoring it to minimize duplication [11:23] dooferlad, voidspace, TheMue, last step of the subnets CLI, please take a look http://reviews.vapour.ws/r/1393/ (esp. proof-reading) === b is now known as Guest63632 [11:27] o/ Good Morning - there's a member in #juju looking for which ports are required to be open on the juju state server. Does anyone happen to know these ports off the top of their heads? [11:27] lazyPower, 17070 [11:27] tcp [11:28] lazyPower, that's the api server, not state server (mongo) which is 37070/tcp, but it's not accessible anyway [11:28] ok, so just 17070 and 22 - the rest should be fine in state: closed? [11:28] lazyPower, most commonly, yeah [11:29] right on, cheers dimitern :) [11:29] lazyPower, no prob :) [11:56] axw: katco: I've updated http://reviews.vapour.ws/r/1378/ for now I'm just commenting out the 1 non deterministic test. I'll try to work with william to get it tested again [12:28] dimitern: if I add issues and then hit "Ship It", does it automaitcally become "Fix it, then ship it"? [12:29] dimitern: as I can't specifically see an option for that [12:29] dimitern: anyway, you have a review [12:31] wow.... running "worker/uniter" tests causes the test suite to rebuild cmd/jujud/jujud which consumes about 700MB just for the 6l linker... [12:31] so much for running the test suite on a 1GB VM [12:31] dimitern: hmmm... no, it doesn't. Ah well. And I seem to have done it twice :-) [12:34] jam: so it rebuilds it everytime, even if there are no changes? [12:34] wwitzel3: I believe the issue probably has to do with JujuConnSuite building tools and "uploading" them to the environment as part of default setup. I don't quite understand why it always rebuilds jujud, but it is probably building everything in a temp dir (I would guess) === niemeyer is now known as niemeyer_ [12:36] jam: hrmm, wonder if there is a way to avoid that. Rebuilding everything is not an insignificant amount of time and with our test suite already being slow. [12:37] wwitzel3: well I only noticed cause i did "go test -c" and then "./uniter.test & ./uniter.test" 10 times and my VM died to swapping :) [12:37] jam: heh [12:41] dimitern: in the "Juju Container Addressability" doc, is the greyed out section (from the bottom of page 5) [12:41] dimitern: there for historical reasons only, or do I need to go through that as well? [12:41] voidspace, it's historical only [12:41] dimitern: ok [12:41] voidspace, sorry, I just saw your earlier messages [12:42] dimitern: I deleted one out of date bullet point and I'm adding an additional one in the "common features" section about the addresser worker [12:42] voidspace, so AIUI "Fix it, then ship it" turns to "Ship it" once all issues are resolved/dropped [12:42] automatically [12:42] dimitern: right, but that wasn't the question [12:43] dimitern: it was how to post a "Fix it, then ship it" in the first place [12:43] dimitern: I posted a "Ship It" and it was just a "Ship It" [12:43] :-) [12:46] voidspace, ah :) well the "Fix it" part appears when you add any issues and tick the ship it box [12:47] dimitern: ah! [12:47] dimitern: thanks [12:47] I didn't post two "Ship It" reviews - wwitzel3 posted one within the same minute as me... [12:48] dimitern: hmmm... I have a horrible feeling [12:48] dimitern: we have the watcher, worker, apiserver ReleaseContainerAddresses method, api client method [12:49] dimitern: but it doesn't look *to me* like destroying a container calls the ReleaseContainerAddresses method [12:49] unless I'm missing something [12:49] I really thought I did that... [12:50] dimitern: the provisioner should be calling it in StopInstance [12:50] voidspace, how about adding a few ops in state around machine destruction? [12:51] dimitern: so when the machine is destroyed in state release the addresses [12:51] dimitern: then we don't need the api [12:52] and the logic is simpler [12:53] we have the machine ID, just find all IP addresses with that machine ID and mark Life as dead [12:53] no need for the unit agent to do it, so no need to check permissions [12:54] dimitern: I'll create an issue and a kanban card *sigh* [12:57] voidspace, thanks [12:57] voidspace, yes, that sounds like a better approach - the provisioner API RCA() can still stay (for now at least) [12:57] dimitern: heh, you just unassigned me from a three year old bug [12:58] voidspace, I think it's safer to do it in state though [12:58] dimitern: yep [12:58] voidspace, are you still working on it? :) [12:58] dimitern: I might have got round to it... [12:59] voidspace, well, in that case feel free to reassign yourself then :) === urulama_ is now known as urulama__ [13:00] dimitern: I think I'll skip it... [13:05] mgz, dimitern Can either of you review http://reviews.vapour.ws/r/1394/ [13:10] sinzui: on it [13:10] sinzui: lgtm [13:11] thank you mgz [13:12] aargh, spotify killed unity again [13:12] mouse no longer works, events blocked [13:12] I can alt-tab and type [13:12] but can't use a browser very well [13:12] will reboot and go on lunch, create tickets (and work on them) after that [13:25] sinzui: abentley: i see "inc-1.23-beta4" in the pipeline, and I'm trying to land the "remove leader-election flag for 1.23" as well. [13:25] It bounced once due to AddRemoveSet [13:26] but it should land [13:27] jam, yes, 1.23-beta3 was sent to the builders. we will release a 1.23-beta4 later this week or next if need be [13:28] dimitern, bug 1427814 should be High, or we should remove it from the milestone. We don't commit Medium bugs to deadlines. [13:28] Bug #1427814: juju bootstrap fails on maas with when the node has empty lshw output from commissioning

[13:28] sinzui, ok, I'll retriage it as high then [13:29] thank you dimitern [13:44] Bug #1427508 changed: cmd/jujud/agent: test failure

[13:44] Bug #1438168 changed: juju 1.23 doesn't release IP addresses for containers [13:44] Bug #1438683 changed: Containers stuck allocating, interface not up

[13:44] Bug #1438820 changed: IP address life field upgrade step and addresser worker don't play well together [13:58] ericsnow, wwitzel3: note the moved standup [13:58] (in an hour) [13:59] natefinch: k [13:59] ericsnow: are you not in pycon? [13:59] * perrito666 looks at the calendar and acuses it of lying [13:59] perrito666: leaving for the airport in 4 hours [14:00] ahh :D makes sense [14:00] my calendar, given my tz, has shown me things in the wrong day before [14:00] perrito666: the calendar's right, it just doesn't show today as partial :) [14:02] ahh, I see that now [14:02] hangout is the best max fan speed tool for Linux, even steam doesn't get it cranked like hangout [14:03] wwitzel3: get hardware accel for it [14:34] natefinch: one-on-one? [14:34] ericsnow: I think we can skip it unless there's something you need [14:34]