[00:01] Anyone able to interpret the output of jam's script from https://bugs.launchpad.net/juju/+bug/1680683/comments/3 for me? https://pastebin.canonical.com/185461/ Looks like something went wrong, possibly because it was run on a non-primary node? [00:01] Bug #1680683: Poor "juju create-backup" performance [00:04] anastasiamac: looks like the "juju models" crash is related to a change in the prototype branch I'm working on. [00:04] anastasiamac: probably nothing to worry about [00:05] menn0: what about destroy model? r u still seeing the failure? even with my fix :( ? [00:05] anastasiamac: still testing that [00:05] * menn0 is juggling 4 things at once [00:05] * anastasiamac still holds breath then [00:07] anastasiamac: gah! i'm not even able to reproduce the problem without your fix now [00:07] ah, worked it out - needs to be run on the primary [00:07] anastasiamac: ^ [00:08] anastasiamac: i'll let you know if I see the problem with your fix in place, but let's assume it's all good for now :) [00:08] * menn0 has to go [00:09] menn0: excellent, i'll breath again then [00:09] anastasiamac: it's actually fairly likely I was seeing it in a controller that didn't have your fix [00:10] menn0: awesome \o/ i'd suprised if u'd see it again :D [00:10] (with my fix i mean) [01:31] babbageclunk: standup? [01:42] anastasiamac: I like your confidence :) [01:43] menn0: if m not confident in mymself, what m doing here?... [01:43] * anastasiamac sighs.. if only i could type [01:43] true :) [05:16] Bug #1681287 changed: juju should retry failed update-status hooks [05:22] Bug #1681287 opened: juju should retry failed update-status hooks [05:34] Bug #1681287 changed: juju should retry failed update-status hooks [07:10] axw: if vsphere vms have extraConfig metadata that can hold tags, why do we still also need folders? [07:11] wallyworld: AFAIK you can't see the metadata in the UI [07:11] wallyworld: folders are the standard way of organising [07:11] ok, so purely a user facing aide, that seems fine [07:12] wallyworld: it also makes various DestroyController more efficient, since you don't need to trawl through all the VMs then [07:12] not a big deal, but it's something [07:12] yeah, i noticed that [07:13] btw, my isp is having issues atm, so i can't get to launchpad or streams.canonical.com :-( [07:13] sigh, makes it hard to look into 2.2 bugs [07:16] axw: pr lgtm, we'll need to ensure CI tests are updated etc [07:18] wallyworld: thanks. CI should add a test for migrate & kill-controller, I think everything else is covered [07:18] axw: i was thinking specifically of verifying folder structure etc [07:19] wallyworld: ok. I don't know what we normally test in that regard, but seems reasonable [07:19] yeah, especially since it's a user facing artefact we are relying on === frankban|afk is now known as frankban [08:23] wallyworld: I've made a small change so that the upgrade step no longer removes the old metadata. it causes an error when the value is empty [08:23] (despite what the docs suggest...) [08:24] ok [08:24] i doubt it will matter much in practice [08:25] wallyworld: just a little messy, no big deal [08:25] nobody sees it anyway [08:25] yep [08:33] who's ready for a question? [08:33] wallyworld, you for example? [08:33] o/ ;) [09:54] anyone around that can look at https://github.com/juju/txn/pull/28 [09:55] wpk: ^^ its not your area but maybe you'd like exposure :) [09:55] I was hoping wallyworld or axw might still be around [09:55] jam: I am but need to go pick kids up from my mother in law's [09:55] oh sure, play the mother-in-law card :) [09:56] axw: np [09:56] jam: if I get a chance I'll look later on [09:56] it is small [09:56] its a bug that hasn't landed in juju-core proper yet, cause I was testing it before landing [09:56] but lack of testing in juju/txn [09:56] jam: yeah that's trivial, I can review now [09:57] jam: LGTM [10:07] Quick juju question - will 2.2 have the same behaviour as 2.1.2 in terms of not automatically bridging interfaces? [10:23] joedborg: yes, the behavior will be the same as far as I'm aware. You'll need to specify interfaces you want setup on containers. [10:23] cheers rick_h! [10:41] axw: ping? [10:41] or menn0: ping? [10:42] babbageclunk: hi (not working though) [10:42] menn0: just a really quick question, promise [10:43] babbageclunk: ok [10:43] menn0: It seems like units on a machine could be bound to different spaces? [10:45] menn0: But then wouldn't they have different public addresses? Or is only one space able to have public addresses? [10:46] menn0: the code only supports (zero or) one public address for the machine and unit.PublicAddress() just delegates to the machine, so I guess I'm wrong. [10:47] babbageclunk: AFAIK they could have different public addresses [10:47] babbageclunk: i'm not sure, but I think the idea of single public address was a overly simplisitic idea which I think we're moving away from [10:47] menn0: Hmm. [10:48] jam would know better [10:48] menn0: ok, thanks. [10:48] machine/unit.PublicAddress() is probably problematic [10:48] menn0: I just realised jam would be the person to ask. [10:49] and it's his afternoon so hopefully he's about or will be back soon [10:50] menn0: yeah, good call - sorry to distract you! Thanks for confirming that there's probably something fishy going on there, anyway. [10:50] babbageclunk: no worries - don't work too late :) [10:50] menn0: no, I'm about to crash out (or get shanghaied by a baby) [10:54] jam: I'm probably going to have to drop soon, but if/when you see this any insights would be welcomed! [11:06] babbageclunk: [11:06] if you're around, I can join a HO or we can IRC [11:07] was just on a phone call [11:07] jam: hey - still here, let's hangout! [11:08] babbageclunk: https://hangouts.google.com/hangouts/_/canonical.com/afternoon-jam?authuser=1 [11:35] jam: you froze [11:36] yeah, brb [11:37] trying to reconnect === frankban is now known as frankban|afk [21:02] anastasiamac: coming... [21:02] just restarting chrome === menn0 is now known as menn0-doctor [22:47] wallyworld: around? [22:47] babbageclunk: i am, but am waiting to head into a meeting [22:48] wallyworld: ok, I'll grab you after. [22:48] babbageclunk: was it a quick question? [22:49] wallyworld: maybe (or else it can be something to distract you in a boring meeting) - are you sure that a unit won't enter scope until it's assigned? [22:49] (to a machine) [22:49] wallyworld: there's code in the ingress address watcher that suggests otherwise, although I'm not sure whether that's defensive coding. [22:49] babbageclunk: i thought i was sure yes [22:50] it's defensive [22:51] ok - needing to handle that would make the stuff I'm doing quite a bit fiddlier (I think) [22:51] yeah [22:53] wallyworld: oh, one more quick one - the IAWatcher maintains a set of addresses (as the current output) and the known map (unit -> address) in sync. I'm tempted to generate the former from the latter when it's needed - is there any problem with that? [22:54] wallyworld: I'll get the tests passing the way it is and do the change in a separate commit to see if that breaks anything. [22:54] babbageclunk: not really, given the size of the lists etc, should be ok [22:55] cool cool [22:55] thanks [22:55] nw [23:19] is there something in juju that can tell me the OS name (not series) of a machine? like "ubuntu" instead of "trusty", "windows" instead of "win2008somethingsomething", etc.? [23:19] by something i mean go function in juju/juju/... [23:25] aha, juju/utils/series.GetOSFromSeries [23:42] anastasiamac: ty, i forgot about adding the pr to the bug [23:42] hml: nps :)