[00:02] kelvinliu__: i left a comment on the juju-managed-units PR - just a small change needed [00:07] wallyworld, ah, u r right. just pushed the change, thanks. [00:07] nw, looking [00:08] kelvinliu__: what about the other bit? [00:08] the if numUnits > 0 [00:11] wallyworld, u mean the deployment controller? [00:12] kelvinliu__: yeah, a few lines down where it does "if numUnits > 0". that check will alsways be true now [00:12] so it's no longer needed [00:12] ie we always want the deployment controller [00:13] wallyworld, that's right. [00:23] kelvinliu: small fix for a code comment and then good to $$merge$$ [00:24] kelvinliu: actually [00:24] we don't need the cleanups slice anymore [00:24] maybe [00:25] ah, i think we do after all [00:25] so leave it as is [00:27] ic, wallyworld thanks === freyes__ is now known as freyes [01:36] veebers: ping [01:42] veebers: nm, found the issue [01:56] wallyworld, can i have ur 2 minutes? [01:56] sure === jhebden is now known as jhebden-afk [02:21] thumper: oh, please share :-) [02:31] veebers: I had golang 1.6 deb installed [02:31] ah, that'll do it [02:31] :-) [03:21] wallyworld and/or thumper: could you please take a look at https://github.com/juju/juju/pull/8680? [03:21] ended up being a bit bigger than expected [03:27] babbageclunk: sap you? [03:27] swap [03:27] wallyworld: Doh. Yup! [03:27] :-D [03:27] wallyworld: you mean 8700? [03:28] yup [03:30] cool cool [03:31] babbageclunk: i still wish we called the worker raft-unfucker :-) [03:32] wallyworld: yeah, that would be pretty cool sort of. [03:43] babbageclunk: does the raft backstop worker need to be a singular worker? [03:44] the manifold definition looks like it sytarts up in every controller? [03:46] wallyworld: that's right - it needs to run in each controller just in case that's the only one left standing. [03:47] Hmm. I guess it could be a singular worker. Except when that's being run by raft it couldn't be. [03:47] ok. i'll read more of the code. i am wondering how they all then coordinate and not get confused [03:47] Well, it only ever does anything when it gets told by the peer grouper that it's the only one left alive. [03:47] the whole idea of a singular work is that there's always one, no matter how many controllers there are [03:48] Right, but we can't use something that will be powered by raft to save raft. [03:48] ok, i'll read more of the code [03:48] The singular worker uses leases, which will be implemented by raft. [04:25] wallyworld: approved with some comments. [04:25] babbageclunk: ty, 75% done on yours [04:25] awesome [04:29] babbageclunk: i have a meeting, so left some initial comments for now [04:29] ok, thanks wallyworld [04:32] wallyworld: FYI percona-cluster and mediawiki bugs I filed re: CMR/upgrade issues https://bugs.launchpad.net/charm-percona-cluster/+bug/1771729 and https://bugs.launchpad.net/charms/+source/mediawiki/+bug/1768710 (I'll follow up the mediawiki one too) [04:32] Bug #1771729: When using CMR db username is too long [04:32] Bug #1768710: workload status does not survive juju upgrade. [04:32] veebers: great ty [05:23] babbageclunk: let me know to take another look at the PR so I can +1 [05:24] wallyworld: ok, will do - just replying to comments at the moment, then I've got some fixes to do. [05:24] ok, np, just want to be sure i'm not blocking you [06:05] wallyworld: just pushing review fixes now, also lots of argumentative replies to your comments ;) - can you take another look? [06:05] sure :-) [06:06] only slightly argumentative. [06:18] babbageclunk: +1 with the main followup required being using the machine id rather than tag string [06:23] wallyworld: if u get a chance, PTAL https://github.com/juju/juju/pull/8717 [06:23] :) [06:24] sure [06:27] anastasiamac: lgtm, awesome [06:27] wallyworld: \o/ [07:30] join [09:33] hi guys, I am unable to login to openstack from dashboard, the error is "Unable to establish connection to keystone endpoint." [09:33] apache error log shows the openstack dashboard is trying to access keystone on the localhost instead of the keystone host [09:34] local_settings.py has the OPENSTACK_HOST set to the keystone hsot [09:34] how can I debug this? [09:34] can some one help? [12:07] or may be is there a place where I can post these questions? [12:09] I'm not that familiar with openstack but I assume most experts are offline right now, what timezone do you live in srihas? You could always try the mailinglist if it's urgent [12:32] TheAbsentOne: CEST [12:32] thank you [12:32] but I have fixed it by upgrading the charm with 258 version [12:34] but we have issues with neutron-api charm when neutron_plugin: aci [14:29] charmers, I have a little competition running with my interns where they've been tasked to build a dashboard over the charmstore API using opensource software of their choosing [14:30] the entries are being judged on, usefulness, ease of install and look and feel. We want it judged by some 3rd parties to remove bias, anyone want to volunteer for judging in a few weeks time? [14:34] magicaltrout: i'm in! [14:35] thanks kwmonroe [14:36] i'm sure i can convince rick_h_ to waste some cycles on it also [14:49] magicaltrout: sure [14:54] cheers egghead [14:58] you're just such a great guy, i cant find anything to tease you about [15:02] :D [15:08] magicaltrout: :) do I get a gavel? I want a gavel. [20:09] i hope https://bugs.launchpad.net/juju/+bug/1771885 is an easy fix... [20:09] Bug #1771885: 2.4-beta2 - lxd containers missing search domain in systemd-resolve configuration [21:33] morning peeps [21:37] babbageclunk: hey, you aware of this? http://ci.jujucharms.com/job/github-check-merge-juju/1418/testReport/junit/github/com_juju_juju_worker_raft_raftclusterer/TestPackage/ [21:40] wallyworld: yup - it bit my PR last night. It's my top priority [21:41] wallyworld: then your "no tags anywhere" vendetta [21:41] :-) [21:41] babbageclunk: also, did you see my comment about the ref count? i forgot about the global vs model collection thing [21:42] wallyworld: ooh, no - just looking [21:43] ah yeah - that makes a lot of sense [21:43] wallyworld: did you see my comment about the select-clause priority thing? [21:43] oh, no, looking :-) [21:48] babbageclunk: huh. maybe that changed. maybe it was always like that. but so much of our code is written to assume that the source order determines which one of multiple ready channels gets the result first [21:48] I can see some confusion because of the evaluation order thing [21:49] What do you think, should I still simplify it? I'm not sure the priority argument really holds water anyway. [21:54] wallyworld: that's the last thing I've got, if you're ok with it I'm landing the sucker! [21:55] babbageclunk: if you think the order doesn't matter then simplification would be gr8. we don't seem to make the distinction in other owrkers [22:00] thumper: wouldn't mind a teddy bear for something if you are free at some point [22:00] wallyworld: I'm using babbageclunk for that just now [22:01] be gentle with him [22:09] wallyworld: how urgent is your teddy bear needs? as I'd like to complete this bug fix [22:09] and I need to talk to veebers to do that [22:09] thumper: not too urgent, i'll relocate soon and maybe we can chat after [22:09] ok [22:09] veebers: got a few minutes? [22:09] thumper: sure do [22:10] sweet [22:10] veebers: in our 1:1 [22:40] whenever I add --destroy-all-models to a command I picture Bender saying "destroy all humans". I wonder if I can alias that argument [22:45] thumper: Am I right in thinking that Catacomb.Err() will never return Catacomb.ErrDying()? [22:47] I couldn't say without looking [22:50] thumper: it looks like it from my reading - tomb.Kill(err) will never store tomb.ErrDying as the reason it's dying. [22:54] babbageclunk: https://github.com/juju/juju/pull/8720 [22:55] thumper: looking [23:07] thumper: approved with minor comment. [23:07] babbageclunk: thansk [23:09] reasonable comment [23:11] babbageclunk: although wrong [23:11] ? [23:11] babbageclunk: we need to check against a nil initial error [23:11] because a dial error is better than nothing [23:12] thumper: oh, right [23:12] I had the sense wrong. [23:12] :0 [23:12] all good [23:13] Will other ever be nil? [23:13] nope [23:13] initial is only ever nil the first time combine is called [23:13] ok, in that case fine. [23:13] assuming that combine returns non-nil [23:13] it is only called when there was an error returned [23:14] cool cool