[00:01] babbageclunk: we miss you [00:01] omws! [00:02] hangouts is being weird [00:35] wallyworld: it wasn’t me. ha! it was a merge error [00:35] lol [00:35] wallyworld: so sort of me. :-) [00:35] wallyworld: i’ll fix it [00:36] ta [01:22] thumper: that branch you commented on - it's targetted at a feature branch, not develop [01:22] there's a whole bunch of stuff we're holding off on [01:34] wallyworld babbageclunk: is the agent version stuff all done? I notice that we're still looking for agent binaries in streams for dev builds, is that expected? [01:35] axw: there's still stuff to commit. streams can contain alphas/betas etc. is that what you mean? [01:36] wallyworld: I mean if I build juju from source, should it still be looking for packaged binaries? [01:36] wallyworld: it adds 10 seconds to bootstrap for me to find out there's nothing packaged [01:36] it will unless you use --build-agent [01:36] would be nice to shave [01:36] ok [01:36] otherwise without that flag it doesn't know to use a local version [01:37] so it searches for a match [01:37] wallyworld: I thought we were using the fact that it's not a released agent (i.e. sha256 doesn't match version file) to skip that [01:37] the change from --upload-tools was that without --upload-tools it would fail [01:37] yes, that should be part of it [01:38] what is being worke don now is just the fallback bit [01:38] okey dokey [02:01] wallyworld: https://github.com/juju/juju/pull/8078 you were right, it was a small fix :) [02:01] looking [02:01] lol [02:02] axw: cool - don’t see 1 char fixes often [02:02] :-) [02:02] hml: :) [02:02] wallyworld: the test ended up being deceptively simple. ha https://github.com/juju/juju/pull/8077 [02:02] looking [02:05] hml: great ty [02:05] wallyworld: ty [02:06] g’night all [02:07] night hml [02:15] axw: sorry, was out looking at a house. what wallyworld said though [02:16] wallyworld: oh, ok, good [02:16] babbageclunk: cool [03:21] wallyworld: check out my WIP please? https://github.com/juju/juju/pull/8080 [03:21] babbageclunk: will do afterr meeting [03:21] wallyworld: working on the multiple results from GetMetadata now. [03:22] wallyworld: cool cool - have a nice meeting! [03:22] yay, meeting [04:22] babbageclunk: i left a few comments, not sure if they make sense [04:22] wallyworld: thanks, will have a look shortly [04:22] righto, ping if yo want to discuss [04:58] wallyworld: do you think that ValueParams needs a func that takes a stream name and returns the mirror content id then? [04:59] yeah, maybe, that could work [04:59] the flow is once a choice has been made, the actually download needs to come from the mirror [04:59] ok, I'll add that. [05:00] i think a mirror has product files and binaries [05:00] so the index entry determines the mirror to redirect to [05:00] or something like that, it's been a loooong time [05:01] ok, I'll check how MirrorContentId gets used and see if that change makes sense. [05:02] had a couple of other questions, the rest of your suggestions make sense. [05:07] babbageclunk: let me know if you want to HO or whatever [05:13] wallyworld: yeah, can we? in 1:1? [05:14] ok [05:54] jam: are you able (allowed?) to add lp:~axwalk ssh keys to the GUI MAAS box, so I can sshuttle? [05:55] axw: added: 2017-11-15 00:54:50,190 INFO Authorized key ['2048', 'SHA256:eVZBwvL1gnylHvyYS2+SAxyCXp1YKNVkZHA1otEtvuc', 'andrew.wilkins@canonical.com', '(RSA)'] [05:55] and some others [05:55] jam: thanks [06:15] wallyworld: axw: I just saw "unit-blah-0: juju.worker.uniter.remotestate update status timer triggered" does that have to do with CMR or is the 'remotestate' there just the fact that state is on the controller [06:15] I'm trying to figure out what 'remotestate' has to do with a local charm and no CMR involved. [06:15] but maybe i'm misreading something [06:16] jam: remotestate refers to the other side of the relation [06:16] that term existed before cmr [06:16] IIANM [06:19] jam wallyworld: not quite. remotestate is about what's in the state db [06:19] vs. local state, which is what the unit has done to date [06:19] the unit reconciles local state with remote [06:19] uniter* [06:20] ah right, yes [06:20] but yeah, nothing to do with CMR [06:20] i was thinking of "remote unit" [06:20] and remote unit data [06:24] axw: has your $ vs \$ been landed? [06:24] axw: I was testing anastasiamac's patch on action debug-hooks and thinking that maybe the "failed" values are because that was missing. [06:24] jam: yeah it has [06:24] jam: what failed values? [06:25] axw: you can hook into an action request, but you exit 0 and it reports "failed" as the action result [06:26] jam: yeah, if my commit's not in the branch, that would explain it [06:26] I'll try rebuilding merging dev and see if that changes [06:26] jam: hm, anastasiamac's branch does have it [06:29] axw: ok, I think its because you have to actually run the action and have it report back something, vs just the return code [06:30] doing ./actions/ping and then exit seemed to give me a completed value [06:30] okey dokey [06:30] seems a little backwards maybe [06:31] jam: https://github.com/juju/juju/pull/8082 fixes the MAAS issue, QA'd on jujugui MAAS [06:31] jam: could you please take a look? [06:31] I'm taking the kids to martial arts shortly, will be back in a couple of hours [06:33] looking [06:35] axw: does your patch break "juju deploy --to zone=" ? [06:35] or is that being resolved in the value that is passed into StartInstance ? [06:35] jam: shouldn't, that's covered by DeriveAvailabilityZones [06:35] I'll verify that when I get back [06:36] axw: it seems overbroad to turn all provisioning failures into an AZ failure === frankban|afk is now known as frankban [08:24] axw: had a question about leases an 2.2 and "juju models" not sure if you have context, but when I'm done with standup, I'd like to chat a bit [08:43] axw: your patch effectively deals with bug #1706462 doesn't it? [08:43] Bug #1706462: juju tries to acquire machines in specific zones even when no zone placement directive is specified

[10:14] externalreality: have you been changing configuration on jujugui maas recently? [10:15] Yes, I have changed the configuration of 2 machines nuc.8 and nuc.7 I think the machines are called. Generally, I changed them to have 3 partition 1 mounted as root and 2 configured to Raid0 for reproduction of a bug [10:15] No other node was changed [10:17] I have also modified tags [10:17] wpk, have a broken something you were doing? [10:17] These changes were made a few days ago. [10:27] wpk, Is everything ok? [10:32] externalreality: I'm getting weird bootstrap error [10:32] 11:15:46 ERROR juju.cmd.juju.commands bootstrap.go:515 failed to bootstrap model: cannot start bootstrap instance: cannot run instances: cannot run instance: No available machine matches constraints: [('zone', ['azone']), ('mem', ['3584']), ('agent_name', ['a5bd9766-22fa-4d81-8748-c719f85fd1cb'])] (resolved to "mem=3584.0 zone=azone") [10:32] and I wonder if that's something I broke === freyes__ is now known as freyes === freyes__ is now known as freyes === frankban is now known as frankban|afk === TikTok is now known as michealb [20:51] can I get a review of https://github.com/juju/juju/pull/8086? wallyworld, thumper, this is the second piece, patches [20:53] balloons: won't this monkey with our working copies when we build using the makefile? [20:54] babbageclunk, it will indeed. comments welcome.. I had it as a seperate makefile target at one point. I monkeyed with it a bunch, but figured I'd just toss out the PR and see what people think [20:57] balloons: for one thing, it will mean that mgo and other patched packages are always dirty after a build, so make godeps will fail. [20:59] 1. it won't work if there are no patches [20:59] yes, godeps will fail [21:00] 2. I'd make it a separate target, with 'undo' (-R) option [21:00] balloons: We could mitigate that by having something to unapply the patches? But that seems risky - what if someone was making a change to the package? [21:00] https://github.com/juju/juju/pull/8084 anyone? (for rc1) [21:01] we could have a 'release-build' target. For the idea of rollback is also fine. patch supports reversing [21:02] balloons: I like release-build [21:03] balloons: I guess as we get more patches (I didn't realise there were so many now!) the divergence of behaviour between patched and unpatched juju might be a problem. [21:08] (for devs doing local testing, I mean) [21:13] that's why I fell on the side off being heavy handed with always building them [21:14] and to make you feel the pain a little I guess so we get a better solution :-) [21:14] but honestly it could be annoying to have your sources mucked with [21:18] yeah, I think we really need to carve out some time to implement the fork support in godeps. [21:19] One of the patches is against a package where the last commit was in April 2015 [21:20] Or in that case we should maybe just fork it and actually update the imports. [21:23] babbageclunk, well, I was pushing for moving to deps, which I think folks are aligned as a long-term vision to do [21:24] oh, I missed that - presumably it supports forking without changing imports? [21:25] ugh, too many go packages with variations on the name dep [21:25] which I guess shows that there's a problem [21:26] balloons: do you mean https://github.com/golang/dep? [21:27] babbageclunk, aye [21:27] cool, thanks [21:29] so it sounds like everyone is in favor of wpk's suggestions then. I'll just propose them [21:29] well your's too babbageclunk.. aka, release-build, and an 'undo' [21:30] balloons: thanks! wpk can have the glory. :) [22:01] babbageclunk: how are they hangin' ? [22:03] wallyworld: still making those changes we discussed yesterday. [22:03] ok, let me know if you get blocked so we can land today [22:11] wallyworld: do you have a few minutes for an HO? [22:12] hml: sure [22:12] standup one [22:12] wallyworld: on my way [22:15] axw, any idea on http://qa.jujucharms.com/releases/5946/job/add-cloud-vsphere/attempt/1298#highlight? [22:18] babbageclunk: https://github.com/juju/juju/pull/8084 ? [22:18] wpk: sure [22:19] tx [22:33] wpk: A couple of comments, but approved [22:39] hml: i just re-read that bug in more detail - he makes the point that the dasjboard download is for Ubuntu deployed openstacks, and that the openrc stuff is generic. it probably is worth considering, but i'll move the bug to a 2.3.x milestone [22:46] thumper: fyi, i created 2.3.1 milestone for things which miss the 2.3.0 cut [22:47] ta [22:48] thumper: moved one bug off the rc1 milestone due to churn in requirements and that it is more a micro feature; no need to rush a fix and regret it later [22:48] ack [22:49] wallyworld: making LookupConstraint.IndexIds return the stream as well... [22:49] ok i think :-) [22:49] wallyworld: when did we change the apiserver to connect to itself rather than a random other one>? [22:49] wallyworld: would it make more sense to have the interface have .Streams() and .IndexId(stream string)? [22:50] thumper: what's the context of the connection? [22:50] babbageclunk: maybe, i'd need to look at the code [22:50] apiservers would get wedged during upgrade as they tried to connect to a different server [22:50] and got told "sodd off, I'm upgrading" [22:50] but it meant that they couldn't upgrade [22:50] oh dear [22:51] um, i can't recall tbh [22:51] i rings a bell though [22:51] not sure if 2.1.x or 2.2.x [22:52] babbageclunk: thanks, fixd [22:53] babbageclunk: so IndexIds() is called in one spot inside GetProductsPath() - do we need the stream there? [22:54] wpk - you should name the error return so that the if err refers to that one. Otherwise there's a case where the err that the defer sees is nil even though the func is returning an error. [22:55] wpk: that might not be very clear. [22:56] wallyworld: we need it there (or somewhere around there) so we can make the mirror content id depended on the stream. [22:57] yeah figured as much, i just didn't see any reference to the mirror data near the call site, but i'm not 100% familiar with the code detail [23:03] wpk: added a (hopefully) clearer explanation on the PR [23:03] wpk: does it make sense? [23:06] wallyworld: I'll do it the way we talked about for now. [23:07] babbageclunk: whatever works :-) [23:24] babbageclunk: .. and that's why I wanted to avoid conditional defer :P [23:24] wpk: I mean, you could have said that. :) [23:25] babbageclunk: I won't make it for rc1, but I will have that in mind (it's only cosmetic, as the device will be eventually cleaned when the machine is removed) [23:26] wpk: ah, ok [23:26] the fact that { foo, bar := baz() } shadows outside scope bar is one of the few things I hate [23:26] babbageclunk: nothing will leak [23:27] wpk: yeah, scoping stuff can be pretty subtle and annoying. [23:41] wallyworld: agreed, openrc is more generic - though missing important info potentially [23:41] hml: yeah, we can prompt for missing if needed [23:41] wallyworld: just needs to be thought out. :-) [23:41] yup [23:45] jam: sorry, I saw your message about chatting yesterday after I got back, but then got distracted. can chat today if you still want to. [23:45] jam: my patch does not address https://bugs.launchpad.net/juju/+bug/1706462 AFAICT [23:45] Bug #1706462: juju tries to acquire machines in specific zones even when no zone placement directive is specified

[23:51] balloons: no idea. braixen/vcenter seems to intermittently shit itself, and I don't know why