[00:38] * thumper is leaching free cbd wifi [00:38] wallyworld: ping [00:43] thumper: hey [02:10] wallyworld: https://github.com/juju/juju/pull/8141 adds the initial caasoperator API [02:11] wallyworld: I'm going to tack on another commit to move the caasprovisioner facade from facades/agent to facades/controller, OK? [02:42] dear wallyworld: I want to bootstrap a 2.3-rc1 controller with old audit logging turned on so I can test the upgrade step, but even though I'm specifying --config "auditing-enabled=true" at bootstrap, when I look at controller-config for the running controller it says auditing is off. What am I doing wrong? Is there some trick? Yours &c, Wondering in Whitby [02:42] axw: sgtm. do you have a dockerhub user id? [02:42] babbageclunk: it's pulling down the released tarball [02:43] as that is in streams [02:43] you need --build-agent [02:43] to force it to upload a local jujud and override the published version [02:43] in general, try not to develop on the same version as is released [02:43] sure - I want a controller with the old behaviour so I can upgrade to mine. [02:44] wallyworld: nope, I'll create one [02:44] in that case upgrade-juju --build-agent [02:44] axw: i have a jujud image pushed - i'll give access to we can hard code that one into juju for now [02:44] No, the bootstrapped controller (before upgrade) doesn't have the old audit logging running. [02:44] But I want it to. [02:45] wallyworld: docker ID: axwalk [02:45] ok [02:45] babbageclunk: rc1 controller will have old logging. and you want to upgarde to tip of develop right? [02:46] brb, restarting [02:46] Yes, but first I want to have a controller with the old logging, but the one I've got has auditing off. [02:47] sure, i don't get the issue though - just bootstrap an rc1 controller [02:47] I have [02:47] and upgrade with a local jujud [02:47] so what isn't working? [02:48] The one I have doesn't have audit logging turned on, even though it's the old version and I specified --config "auditing-enabled=true" [02:48] Can we do a hangout? [02:49] sure [02:49] 1:1 [05:40] wallyworld: did you share the image with me? I didn't get a notification or anything [05:41] axw: i added you as a collaborator; there was no option i could see to set anything else [05:41] wallyworld: what's the docker hub url to the image? maybe I need to navigate there directly. doesn't show up on my proifle anywhere I can see [05:43] axw: wallyworld/caas-jujud-operator [05:43] ta [05:47] wallyworld: have you pushed the Dockerfile somewhere? doesn't appear to be viewable there [05:47] i have pushed it [05:47] hmmmm [05:48] wallyworld: I can see https://hub.docker.com/r/wallyworld/caas-jujud-operator/, but no details. just a docker pull command [05:48] https://hub.docker.com/r/wallyworld/caas-jujud-operator/ [05:48] wallyworld: I might be able to push an image there... but I can't see the Dockerfile [05:49] i can't see the dockerfile either, i just pushed the image [05:49] i'm about to put up a PR with the dockerfile in it [05:50] wallyworld: OK, ta [05:56] axw: https://github.com/juju/juju/pull/8142 [05:56] wallyworld: thanks, looking. ICYMI, please take a look at https://github.com/juju/juju/pull/8141 [05:57] yeah, looking now [06:03] looks great, thanks for picking up the rename [06:23] wallyworld: https://github.com/juju/juju/pull/8143 - small one, gets things working iwht minikube [06:53] axw: did you run minikube from the snap? [07:01] wallyworld: no, I don't recall how I installed it, but it's not a snap [07:02] axw: i tried it - getting errors dialling address. someone else also had same issue using snap [07:03] wallyworld: minikube itself has issues, or juju has issues using k8s set up with snapped minikube? [07:03] minikube itself [07:04] damn same without snap === frankban|afk is now known as frankban [09:01] jam: https://github.com/juju/juju/pull/8144 and now I'm off to bed :) [09:01] night [09:01] thumper: night, [09:01] jam: if it's good, ask the bot to merge, if it needs fixing, it can wait until 2.3.1 [09:01] thanks [09:16] axw: just tested latest feature branch code - using image i just pushed to dockehub - it hangs together nicely :-) need to check python libs in the image etc but good start [09:17] wallyworld: sweet :) === freyes__ is now known as freyes [15:23] wpk, can you review https://github.com/juju/juju/pull/8086? [15:27] balloons: one suggestion, LGTM === frankban is now known as frankban|afk [21:34] wallyworld: I've revved the Admin API version, but I haven't made the v3 API do anything different with the CLI args - if they're passed in they'll get stored in the same way as they will for v4. [21:35] babbageclunk: righto, i think that's ok [21:35] wallyworld: On the client, do I need to start checking for API versions before the Login request? [21:36] i don't think so since things won't break [21:36] wallyworld: So an old controller would be ok with a client sending a v4 Login request? [21:37] if it doesn't have a params struct that knows about cli args, it will just drop them [21:37] But we explicitly say version 4 in the API call - will that cause a problem? [21:37] I guess I can try that... [21:38] otp, can't think, give me a few [21:38] ok, I'll do an experiment [21:51] babbageclunk: IIANM the request doesn't include any version - it just connects to whatever facade is there. it's up to the client to determine if that version is fit for purpose [21:54] wallyworld: but the call includes the API version: st.APICall("Admin", 4, "", "Login", request, &result) [21:55] wallyworld: and the LoginResult is where we get the facade versions to work out which version we should send. [21:56] wallyworld: so if I bump the version in the client login call then it can't talk to a controller that only supports v3. [21:56] babbageclunk: hmm, normally that version is set to BestAPIVersion() so i think Login is different [21:57] I guess we could try 4 then fallback to 3? Seems like a lot of work for this backwards compatible case though. [21:57] we could just set it to BestAPIVersion() like other api calls [21:57] But doesn't BestAPIVersion rely on having the API versions from the loging result? [21:57] login [21:58] oh yeah [21:58] chicken and egg [21:59] we used to handle login 1, 2 and 3 [21:59] but all that code was deleted [21:59] not sure how it used to work [21:59] How'd we do the negotiation? Just trying again? [21:59] I'll have a look [21:59] all that was in the 1.25 branch [22:03] wallyworld: yeah, it looks like it was just "try this version, no, try this one, no, try this one - ok". [22:04] wallyworld: If it's ok with you, I'm not going to put that in for this. [22:04] ok i think [22:04] just stick with v3 [22:07] wallyworld: cool, thanks. [22:26] wallyworld: can you put a tick on https://github.com/juju/juju/pull/8130 please/thankyou? [22:26] sure [22:26] cheers! [22:28] wallyworld: ooh, thanks for the comments on the other PR. [22:28] babbageclunk: no worries - i am a little confused about the subtle? differences between Call and Request [22:29] we can discuss at standup perhaps [22:30] wallyworld: yeah, you're right they're confusing. Call represents a top-level juju command (which will produce multiple requests/responses) [22:30] Maybe command or connection? [22:31] babbageclunk: is in, in the cmd/juju package, the CLI creates one or more facades and then makes various api calls on those facades [22:31] Yup [22:31] yeah, Call not so good then. let's brainstorm in standup [22:31] ok [22:32] I'll go through your other comments. [23:58] wallyworld: need to run QA still, but please take a look at https://github.com/juju/juju/pull/8149 later? [23:58] test failures have been driving my crazy