[00:38] axw: you see this at all, go vet error: provider/vsphere/internal/vsphereclient/createvm_test.go:455: possible formatting directive in Fatal call [00:39] maybe fix as a drive by in the other pr? [00:43] ah, already merged, i'll fix [01:03] axw: here's a followup PR which finishes the new suspend/resume processing https://github.com/juju/juju/pull/7857 [01:08] wallyworld_: doh, thanks. looking [01:09] ta [02:21] * babbageclunk goes for a run [02:36] thumper: axw: do we have plans to support temporary aws credentials? bug 1714022 [02:36] Bug #1714022: Juju failed to run on aws with authentication failed [02:37] anastasiamac: it has been asked for, but AFAIK nobody has put time into designing a solution [02:37] axw: k, i guess I'll triage to wishlist then :) [02:38] anastasiamac: it would be relatively straightforward for non-JAAS, and support for multiple credentials (models with separate credentials) [02:38] triaged [02:39] supporting JAAS and multiple creds (controller and model machines in different accounts) makes things more complicated [02:53] Bug #1714130 changed: juju reports units in non-existent OpenStack availability zones [03:09] thumper: babbageclunk: i think ur work last week (or so?..) addressed this too :) bug 1597490 [03:09] Bug #1597490: juju 2.0-beta9.1: juju relation status PROVIDES and CONSUMES confused [03:42] anastasiamac: true! [03:45] babbageclunk: \o/ nice... do u remember if it went in 2.2.3? [03:45] anastasiamac: yup - just commenting. It was fixed in August, so it'll be in 2.2.3. [03:45] https://github.com/juju/names/pull/82 needed for bundlechanges work [03:46] babbageclunk: k, i'll do milestone magicking... [03:52] wallyworld: yucks... another failure where juju is going to "dev version" for no apparent reason :( bug 1714526 [03:52] Bug #1714526: juju 2.2.2.1 unable to add a unit to an existing service, no matching agent binaries found [03:55] it's a dev version because a non-streams agent was used [03:57] wallyworld: that's not the point [03:59] it is the cause of that error message [03:59] if it weren't a dev build, the error wouldn't happen [04:12] wallyworld: it's in IS, bootstrapped with released 2.2.2, then something happen... according to the bug, they r not sure what... the point is - "this should not happen" or "we should know why it's happening and fix it" [04:12] agreed. i was merely commenting on the error message being valid if agent is dev version [04:13] it happends because there's a bug in upgrade-juju [04:13] which is on the todo list [04:16] wallyworld: of coure. i was not asking for the cause. I was saying "yet again..." [04:16] ok :-) [05:36] axw: i've updated the pr with leadership checks etc. was a little messy to weave it into the uniter. pushed as another commit. i had to add an IsLeader attribute to the hook info (not serialised to state though). [05:36] wallyworld_: ok, will take a look after I finish lunch [05:36] no rush, ty [06:08] wallyworld_: I think it might be better if you put the SetStatus call in uniter/operation/runhook.go (specifically, runHook.afterHook) [06:08] wallyworld_: there's already code in there that sets status after Start and Stop hooks [06:08] could do, ok [06:08] wallyworld_: then you can use the HookContext directly [06:08] no need to add IsLeader to hook.Info [06:08] i did look there and discounted for some reason [06:09] no idea why off hand [06:20] axw: oh i know why - the runner has no access to the relation so i'll have to add an extra method to the context [06:20] probably cleaner though === akhavr1 is now known as akhavr === frankban|afk is now known as frankban [08:25] axw: i've moved the set status to the afterHook callback [08:28] axw: not sure if you saw my msg above as you disconnected; i've moved the set status to the afterHook callback [08:28] wallyworld_: thanks, I didn't. will take a look in a moment [08:28] no worries, no hurry [09:06] axw: what do you mean by "token is propagated to setStatus"? [09:07] wallyworld_: see my earlier comment, "this token should be passed down to the SetStatus call. state.setStatusParams has room for a token." [09:08] ah, didn''t see that comment [09:15] axw: huh, nothing sets that token param at the moment that i can see. whoever put it there didn't wire it up [09:16] i think because the SetStatus interface method doesn't allow for it [09:16] set status in application facade also doesn't use it [09:17] wallyworld_: the application facade? only thing setting status in there is your code isn't it? [09:18] wallyworld_: I think it was added by William, but he didn't finish plumbing it through [09:19] wallyworld_: in the case of Relation.SetStatus, I don't think we're using an interface, so it should be straight forward to add an optional leadership token [09:19] we may not always want to set from the app leader, hence optional [09:31] wallyworld_: please see my reply about application facade. I'll sign off, just want to make sure we're on the same page [09:36] wallyworld_: in your other PR, I was thrown by the info.Info[0]. why do we assume that there's only one? [09:55] wallyworld_: ping? did you see this? I'm OK with that, since it's non critical. Please file a bug and leave a TODO(wallyworld). [09:55] FWIW, I don't think it's necessary to pass a token in the application facade (I assume you're referring to SetRelationsSuspended), because that's not being done by the application leader, but rather a client. And that code's meant to be going away anyway, right? In fact I think this PR is meant to be replacing it... except you're not setting the Joined status yet. [09:55] wallyworld_: (I saw a !!build!! but no response) [10:23] axw: what makes it harder in jaas? it's on the jaas tools to make that work mirroring the controller work. Is it because in the jaas clase it multiple clouds as well so the api is more complicated? [10:45] axw: sorry, didn't norice ping :-( [10:45] i'll add a todo [10:45] i just pushed a change to remove an api call [10:46] axw: i'm referring to SetStatus() [10:47] axw: with the network-get stuff, the code has historically used the first one for --primary-address [10:59] rogpeppe: hey, were you interested in the walking tour in NY? [10:59] wallyworld_: hiya [10:59] wallyworld_: possibly! i'd let it slide by without actually looking. let me take a look. [11:00] ok, just trying to get numbers as wit will affect what we do [11:07] wallyworld_: i reckon i'm up for it [11:46] rogpeppe: ok, great, i think we have 6 [11:46] i'll see what I can tee up [11:46] wallyworld_: cool [11:47] wallyworld_: just so long as it doesn't interfere with my Irish trad session at 10pm :) [11:47] it won't :-) [12:01] rick_h: IIANM, for it to work you'd need to use IAM roles. one controller with multiple credentials would make that complicated, I think [12:02] wallyworld_: network-get can use the first one, but that doesn't mean we need to hide all the others up in the apiserver code. make that a frontend decision [12:03] axw: pretty much agree, was just being cautious about changing the semantics. since it was first written, network-get has returned Info[0] for primary-address [12:04] i can add them all [12:04] wallyworld_: I'm saying you don't need to change network-get. make the decision in the network-get code, not in the apiserver side. [12:05] right, but there could be another api user, but probs not [12:05] i'll add all of them [12:07] wallyworld_: was the *apiserver* code previously returning only the first one? I couldn't see it filtering out before [12:07] i thought it was, but maybe i am misremembering [12:07] wallyworld_: AFAICS, only in NetworkGetCommand [12:08] wallyworld_: I'm approving the relation one now, but in case you miss it: "See SetRelationsSuspended in the application facade, specifically the TODO you left." [12:08] that's what I was referring to, re setting Joined/Suspended [12:08] ok, ty [12:08] wallyworld_: I'm off now, will take another look at the network-get PR on Monday. have a nice weekend [12:09] axw: tyvm for reviews [12:09] you too === beisner is now known as Beisner === cory_fu_ is now known as cory_fu === frankban is now known as frankban|afk === Beisner is now known as beisner