[01:41] kelvinliu: we already have a really old version of a python juju client in the stable juju packages archive https://launchpad.net/~juju/+archive/ubuntu/stable/+packages [01:42] we should be able to publish a newer "python-libjuju" ppa [01:42] to that juju stable archive [01:42] they were pushed 5yrs ago [01:43] yeah, realy old [01:43] so we just need to push our new "python-libjuju" ppa [01:43] yeah [01:43] to that juju stable archive [01:44] can do that and update the release process doc for libjuju [01:44] that one was manually pushed [01:44] yeah, so updating process doc after doing it manually is all we need for now [01:45] yep [02:05] wallyworld: got a question about the quoteStrings in config process, free to have a quick HO? [02:05] sure [02:40] wow it is raining hard here today [02:40] makes me pleased that the stadium has a roof [02:42] elton had to abandon a concert here in OZ due to gale for rain, wrecked all the sound gear etc [02:43] now we have gail force wind [02:54] Best day of February so far, nice and cool [04:08] wallyworld: should I just continue to change to the spec in the doc? [04:09] what was the final outcome of the discussion? i didn't hear it due to sound issues [04:09] i disagree with splitting it as it sucks for the user [04:09] wallyworld: stdup? [04:09] ok [09:27] manadart: around to talk about rename-space? 2 review comments I am not sure [09:27] https://github.com/juju/juju/pull/11143#discussion_r373477221 (missing information) and https://github.com/juju/juju/pull/11143#discussion_r373573802 (I tried to solve, not sure if thats what you meant) [09:41] nammn_de: I am in Daily. [09:42] manadart: 2 min, cannot find phone for 2fa [10:16] jam: manadart regarding our chat on side-effects yesterday, if my open/close ports operation uses the same Ports model and I stack an [open, close] operation, the ports slice of the model will end up in an inconsistent state (open appends an entry, close replaces the port list with a filtered version) [10:17] I could tweak the done logic for close to make it work but not sure if it's worth the effort given that the Ports model will be thrown away [10:18] (e.g. have close track the entries to be removed instead of the remaining entries and filter the port list in Done) [10:27] achilleasa: Seems reasonable. We discussed how the object curation for consistency with state post-DML execution is kind of a waste when we they are so short-lived. [10:28] manadart: And probably less relevant than ever if the particular entities live in the model cache. [10:28] * manadart tags himself like an idiot. [10:29] manadart: actually, now that I look at it again, it's a bit more complicated. open ports issues an updatePortsDocOps whereas close ports issues a setPortsDocOps [10:30] my guess is that the first txn.Op block will succeed but the later will not due to the revnos and cause the txn to fail [10:30] manadart: quick ho? [10:30] achilleasa: Yeah, give me a mo' [10:30] perhaps I need an "opencloseports operation" [10:49] manadart, got a sec? [10:49] i'm either being stupid or stupid [10:51] stickupkid: In Daily with achilleasa. [10:51] i'll gate crash [10:54] https://jenkins.juju.canonical.com/job/nw-deploy-focal-amd64-lxd/24/console === mup_ is now known as mup [11:22] manadart: incorporated your feedback + updated qa and should now be ready . https://github.com/juju/juju/pull/11143 [12:06] nammn_de: Will look soon. === wpk_ is now known as wpk [12:17] achilleasa: sounds like we want to have a $pullAll instead of a $set [12:20] jam: turns out you cannot compose the two individual ops together without jumping through hoops. As a workaround, I made them into a single op that calculates the final port range list and either does insert or $set. Original tests pass and I am adding some extra ones for composed open/close [13:14] manadart rick_h: do we allow the deletion of provider provided spaces? (Maas) [13:14] nammn_de: No. [13:21] +1, nope the world is the world we have to live in it [14:14] can anyone explain to me why this test works? https://github.com/juju/juju/blob/develop/state/unit_test.go#L1594 I don't see any unit-is-alive (there is one for model-is-alive though) assertions in the open/close txn code https://github.com/juju/juju/blob/develop/state/ports.go#L199 [14:20] rick_h, fyi: https://github.com/lxc/lxd/issues/6833 [14:20] stickupkid: <3 [14:42] Small mechanical change: https://github.com/juju/juju/pull/11181 [14:42] manadart, me look [14:48] manadart, I looked, also saw you may have fixed a runtime panic? [14:48] stickupkid: We've never hit it AFAIK. [14:48] good piece of code then [14:49] kill it? === sfeole is now known as sfeole_away [15:02] stickupkid: Volume attachment plans are used by the OCI (Oracle) provider for attaching volumes to instances where the local machine needs to take some actions, I believe. [15:02] manadart, but we know that would never have worked, so isn't exercised [15:04] manadart, whilst we're at CR sharing https://github.com/juju/juju/pull/11182 [15:10] stickupkid: Yep. Can you look at the commit I just pushed to mine? [15:29] jam: got it updated and pushed together a bundle with the bits in it https://discourse.jujucharms.com/t/monitoring-juju-controllers/430 [15:32] manadart, I think we should add a comment and it's done :) [15:37] stickupkid: OK. [15:38] something of the lines, "we have no idea what we're doing, but this should fix it" [15:38] haha [15:40] nammn_de: I am going over your patch again. I was thinking that given the number of files touched for these, we should break future ones up. [15:40] We can add the API server, client and cmd parts in separate patches. [15:44] manadart: agree, should makes everything easier to review [15:47] manadart: while running through it. Should we be able to rename-space `alpha`? I did not catch the case if someone wants to rename. i just did, because I thought that we hold by the spaceID and not spaceName [15:49] manadart: it looks like there may have been work done on the duplicate subnets already, but some of it is a work around. background in https://bugs.launchpad.net/juju/+bug/1733266 [15:49] Bug #1733266: Autodetection of subnets prevents bootstrap when there are duplicate subnet ranges

[15:50] nammn_de: We should prevent it. There are some conditions interrogating the name. It can be done in the cmd. [15:50] manadart: oh, I need to add this then. [15:51] manadart: I will let you finish review first and add the alpha case in the cmd [15:51] hml: Thanks, I'll take a look. I still need to work out why I am not seeing subnets from another network in the same tenant. [16:09] manadart: does it work with dev branch? [16:10] nammn_de: Actually, we need to check at the API server, in case someone is using say, pylibjuju. [16:11] nammn_de: QA is looking OK here. Go ahead and add that change. [16:11] manadart: will do. [16:11] achilleasa: The network thing? I am using develop HEAD. [16:12] manadart: I promise my next pr is gonna be puny :D [16:12] Well, HEAD-ish. [16:20] manadart: added [16:38] manadart: I am thinking of modifying SettingsGroup in state/settings.go to implement ModelOperation instead of having a Write() method (func equiv to buildTxn+apply). I added this type when I introduced the UpdateNetworkInfo call in uniter. IIRC you recently added a settings op type. [16:38] I can probably switch the code to use that and remove SettingsGroup. Thoughts? [16:39] nammn_de: I think you used the ^^ type in one of your recent PRs [16:45] I wonder why we never used middleware for auth of facades [16:46] used composition rather than logic checks [16:46] * stickupkid thought of the day [16:51] achilleasa: So I used ModelOperation for the PRs [16:53] so no real throughts on the settingsgroup refactor. Their write would comply with applyoperation. Modeloperation I think [16:53] but it does make sense. As they comply both [16:53] nammn_de: can you point me to that PR? [16:54] achilleasa: https://github.com/juju/juju/pull/11143/files#diff-c2f3a22eae3088d2abd0727df01272e4R2 [16:54] opfactory.go and rename.go [16:55] thanks! [16:55] stickupkid: yeah, would be really cool. [17:19] manadart: nammn_de actually the SettingsGroup is still required because my use-case requires persisting multiple *Settings in a single txn. I will make it into an operation for consistency [17:26] rick_h: hey there :) [17:26] * roadmr about to pester Rick again heh [17:26] roadmr: howdy [17:26] pester away [17:27] rick_h: so the list of urls that juju hits, that you obtained from logs [17:27] rick_h: it has method, url, query string basically. right? [17:27] roadmr: right [17:28] rick_h: would it be possible to get user-agent as well? [17:29] rick_h: with that we can first weed out non-juju clients and have a better idea of what juju itself needs [17:29] roadmr: all those are a generic "go client" user agent [17:30] rick_h: also lets us see how the api access patterns differ between our targeted juju versions (1.25 and 2.x IIRC) [17:30] roadmr: and juju and the charm command are the only ones I am aware of [17:30] rick_h: ahh that's super unfortunate :( whaat [17:30] rick_h: ah but the charm command is python, right? most likely not a go client string [17:30] rick_h: I'd settle for being able to weed out non-jujus [17:30] roadmr: so the charm command is a mix of go and python. [17:30] argh :) [17:31] roadmr: trying to recall, I thought I weeded out obvious charm commands [17:31] rick_h: ah yes, you said "exact requests we're getting form go clients (juju)" [17:31] roadmr: as far as different Juju's, my first instinct is forget juju 1.2X as the only folks we care about that are internal and they use offline charms with mojo ime [17:32] rick_h: ok, that sounds good [17:32] roadmr: not to say I did it perfect but do recall trying to really focus it in and did a lot of cleaning of that data [17:33] rick_h: got it... doh I'm dumb, mind if we move this to #snapstore so facu can also be in the loop? [17:34] roadmr: yea, or I'd suggest the canonical #juju channel so other folks on the team can see as well [20:42] rick_h: do you know who to chase for https://github.com/juju/charmstore/issues/895#issuecomment-576562396 [20:44] thumper: mhilton was working with rt's to folks around these when they happened. I thought this one got corrected [20:44] maybe the bug didn't get closed? [20:44] * rick_h checks rt email history [20:44] maybe [20:45] thumper: yea, marked as resolved === timClicks__ is now known as timClicks [22:04] rick_h: thumper: test passed this time no problem [22:05] we have seen issues if the slaves are busy [22:06] thumper: btw, https://github.com/juju/juju/pull/11172 is ready for looking at