[02:37] hpidcock: "juju-run python3" doesn't work with upgrade-charm PR [02:37] so still something to follow up on [02:51] wallyworld: can you link me the charm you are using and how you are deploying it please on the bug [02:52] will do, just reviewing PR, and also have a zoom meeting at 1 so after that [02:52] it's just my mariadb-k8s charm [02:56] babbageclunk, wallyworld: just thinking... does backup work with k8s controllers? And if so, do we have any idea how we're going to restore? [02:56] I was thinking about this when thinking about mongo 4.0 restore [02:57] thumper: I haven't tried backup on a k8s controller [02:57] backup should work. restore no, falls into the HA category for next cycle [02:57] wallyworld: have you discussed with babbageclunk the mongo 4.0 restore question? [02:58] yes [02:58] Was trying to think about how k8s restore would work - the core restore is the same, but the stuff around the outside (stopping and starting agents and rewriting configs) is all different [02:58] exasctly [02:58] * thumper nods [02:58] probably want to have a check somewhere [02:58] to say "sorry" [02:59] yeah true - I'll add that [02:59] I mean, it'll be hard to follow the directions anyway - you can't scp the backup and juju-restore to the controller machine [03:00] :) [03:00] yeah... we'll need another whole set of instructions for restoring a k8s controller [03:43] babbageclunk: any thoughts on the last comment in this bug? https://bugs.launchpad.net/bugs/1858693 [03:43] Bug #1858693: ANARCHY!!!!!!! Entirely leaderless application spotted in the wild [03:46] thumper: yeah - that sounds like an entirely different problem. (Or I guess the same symptoms caused by something totally different.) [03:46] * thumper nods [03:46] thumper: reading some code... [03:46] why would it timeout enqueing? [03:47] I think that's coming from the raft apply command - presumably the raft engine loop is blocked somehow? [03:47] hmm... [05:06] hpidcock: https://github.com/juju/juju/pull/11443 if you got 5 minutes [05:08] sure [05:09] wallyworld: got time for HO ? [05:12] tlm: just finishing a PR, give me 5 [05:13] ll [05:13] kk [05:15] hpidcock: finally +1 on PR. i think we'll need a followup to deal with the new local attributes and controller state [05:16] tlm: ready for HO now if you're free [05:17] k [05:17] hang on, stupid 2fa etc [05:23] wallyworld: the local state attributes are resolver local state not operation local state. Resolver local state is not saved. [05:27] hpidcock: ah no worries [06:07] wallyworld: when you have a spare two seconds https://github.com/juju/juju-db-snap/pull/1 [06:07] sure [06:08] hpidcock: so 4.0.18 was officially the latest? [06:08] and AGPL not SSL? [06:08] yep, but either we choose 17 or 18 and need to test it [06:09] its confusing because SSL is the mongo license but we include both mongo and mongo-tools (which is apache) [06:09] yeah [06:09] IANAL :-) [06:10] well the charm itself has a license [06:10] sorry the snap [06:10] the snap code is under whatever license we want [06:10] tup, just about to say that [06:11] technically the snap contains more than just mongo stuff [06:11] openssl etc [08:05] manadart, I got it working, just wrapping up the tests for a review [08:19] here is a casual observation, if you're creating mocks for anything outside of your package, but within juju, then you're doing it wrong [08:27] stickupkid: Will look in a mo'. OTP with thumper. [08:27] stickupkid: any interesting examples of this happening? [08:28] achilleasa, as in, doing it badly (got lots of those), doing it well, not so much [08:28] stickupkid: more interested in the doing it badly ;-) [08:30] achilleasa, anything to do with networking common, model backend... [08:32] achilleasa, if you're taking on a bigger interface from somewhere else, then you're tied to every other place that uses it. You know longer have a vertical boundary and are locked stepped to that other component/entity etc. [08:35] stickupkid: ah, so you are basically talking about dupping (or defining a restricted version of it) the interface in your package and mocking that? [08:37] achilleasa, yeap [08:37] achilleasa, my usual spiel [09:52] stickupkid: You can get 3 shims down to 2 by pushing the raw state shim into environs/space like this: https://pastebin.canonical.com/p/fRT68XZhSQ/ [09:52] But that damn modelmanager backend is a bastard. [09:53] manadart, yeah, I've not got there yet, as I'm writing tests for spaces reload API [09:53] manadart, I've got rid of my horrid life hack as well, which has made me really happy [09:54] Good morning/evening. [09:55] I've bumped into a problem I can't understand if it is a bug or something I can resolve for myself. I've sent a discourse post about it here: https://discourse.juju.is/t/error-cert-pool-creation-failed-cannot-parse-certificate/2927 ... hope to get some help on it so I can get back my controller. [09:56] In short, I get this: "ERROR cert pool creation failed: cannot parse certificate..." [09:57] Followed by...... asn1: syntax error: data truncated [10:16] Right, I don't know what has happened. But my resolution to ^ was to remove the controller entry from controller.yaml and then log back in as admin: "juju login -u admin 192.168.2.12:17070 -c snowflake-maas" After being prompted with a password challenge and a fingerprint accept, the controller is again available to my client. [10:17] elox: looks like you were hitting https://github.com/openssl/openssl/issues/1381 [10:18] elox: openssl asn1parse -inform der -strparse 40 -in $yourcert manages to parse the ASN1 data (fails without the strparse flag) [10:20] I'll have a look and update my post on discourse. [10:20] Thanx! [10:21] achilleasa: Can you review this small one? [10:21] https://github.com/juju/juju/pull/11445 [10:22] manadart: looking [10:23] @achilleasa: https://discourse.juju.is/t/error-cert-pool-creation-failed-cannot-parse-certificate/2927/2 [10:29] elox: great that you could get back the controller without having to do any cert surgery ;-) [10:36] achilleasa: Yeah, the horror. [10:39] manadart, updated in my last commit [10:39] Finally I was able to test the latest version of my Nextcloud charm on my own system and feel comfortable releasing it to stable: https://jujucharms.com/u/erik-lonroth/nextcloud/4 [10:40] stickupkid: OK. [13:01] stickupkid: rick_h proposed change for fixing the weird defaults when parsing charm meta: https://github.com/juju/charm/pull/309 [14:24] can someone please take a look at https://github.com/juju/juju/pull/11446? [15:12] manadart, ping you now or in the morning about the new stuff? [15:32] stickupkid: Morning. I've EoD'd [15:32] manadart, ha [15:37] stickupkid: re your wording comment; should I change it to 'as'? [15:37] achilleasa, yeah, that's better [15:43] achilleasa, tick [15:44] stickupkid: still need to patch some more stupid tests with the correct limit expectations :-( [15:44] achilleasa, 9 year old code, 9 years old [15:45] achilleasa, it's like clippy, "do you want to limit this?" <- no clippy, NO! [15:46] btw, what is the correct command for bootstraping a focal instance? [15:46] shouldn't --boostrap-series=focal --force do the trick? [15:46] achilleasa, you need --config image-stream=daily [15:47] achilleasa, not released yet, so you need to point the image-stream to the right place [15:47] aha! thanks [17:35] rick_h: got a focal with 4.0.18 from the candidate snap https://paste.ubuntu.com/p/kpNNWSQ9VV/. Will work on getting the conf settings through (bit harder since they need to be injected into the agent conf) but should have a PR ready for review by EOD tomorrow [17:41] achilleasa: sounds like grat progress ty! [18:40] I'm trying to help a friend adding AWS credentials to JAAS, and we can't do it. It complains about a credential tag. https://pasteboard.co/J3Xx4Pr.png [18:40] We have been trying to follow the guide, but it simply doesn't work [18:44] Can this be related to already existing credentials.yaml for him ? [19:21] elox, which guide exactly?