=== lifeless_ is now known as lifeless === charlie_ is now known as Guest48100 === reco is now known as crazy_charlie [12:00] query.. what does juju expose actually do from an lxc/lxd perspective ? [12:22] Dwellr: nothing, expose is meant to update firewall rules to enable ports the charm needs to open properly. On LXD and providers without a built in firewall setup (security groups or the like) there's nothing there to work against [13:01] is there any reason why the ceph charm does not implement the ceph-admin interface? [13:03] i would like to try out ceph with kubernetes with a minimal setup as a poc, not having to deploy ceph-mon x3 and ceph-osd x3 if possible [13:04] SimonKLB: not sure. Have to check in with the OS folks. Cool experiment to try out. [13:07] jamespage icey cholcombe ^ [13:08] SimonKLB: the ceph charm is officially deprecated; as such it does not always grow the same features as ceph-mon and ceph-osd do over time [13:08] jamespage: got it, is there any way to have a ceph deployment smaller than 6 machines? [13:08] only reason its still in the charm store is we've not found a satisfactory migration approach for existing ceph deployments. [13:09] SimonKLB: what provider are you using? [13:09] jamespage: aws [13:09] SimonKLB: hmm no not really [13:09] jamespage: alright, then ill go with that! [13:09] with MAAS you can of course place ceph-mon on LXD containers, and ceph-osd alongside k8s [13:10] ah, would that not work on aws though? [13:11] ceph-mon on 3 LXDs and ceph-osd on one worker each would be pretty neat [13:11] just to try it out [13:19] SimonKLB: AWS doesn't have the overlay networking required to get packets into the ceph-mon's in containers [13:19] icey: ah too bad! [13:24] (that problem will be solved in 2.3) [13:27] wpk: im running 2.3-alpha1.1, that fix is not there yet? [13:29] SimonKLB: no [14:19] wpk: you are my hero [14:20] great news [14:42] does anybody know what is up with the bigtop repos? [14:42] kafka charm is currently failing because of 403 forbidden [14:42] http://bigtop-repos.s3.amazonaws.com/releases/1.2.0/ubuntu/16.04/x86_64 [14:50] kwmonroe: ^ [14:51] hm - not sure tvanhove, but i'll send a "wat?" to the dev list [14:57] wondering if (after installing conjure-up kubernetes-core) I should be using iptables to route traffic to the node InternalIP .. or if I should be figuring out how to add an 'ExternalIP' to the node [15:07] tvanhove: not that you needed it, but i verified the 403 on a different arch as well. looks like something is afoul with all 1.2.0 repos: [15:07] E: Failed to fetch http://bigtop-repos.s3.amazonaws.com/releases/1.2.0/ubuntu/16.04/ppc64le/pool/contrib/k/kafka/kafka_0.10.1.1-1_all.deb 403 Forbidden [15:13] yeah we were setting up for a demo next week and noticed the failure in our juju storm deployments with kafka [15:26] tvanhove: mail sent to dev@bigtop.apache.org. i'll keep you in the loop as soon as we figure out what's up. [15:27] tvanhove: until then, one possible workaround would be to manually set the repo on each affected unit to the CI builders. to do that, you'd edit your apt sources like this: http://paste.ubuntu.com/25445303/ [15:29] i've verified an apt-get update / install works from those repos. [15:30] buuuuut, that's upstream vs the official 1.2.0 release. so don't go to production with that ;) [15:34] it's just for demos right now [15:34] no production [15:34] thanks :) [15:34] "it's just for demos" <-- that's what they all say ;) [15:34] ;) [16:03] o/ juju world [16:07] \o stormmore [16:13] tvansteenburgh: do you recall if stub's "hookenv.principal_unit" fix was the only thing in ch-0.18.1 (vs 0.18.0)? wanna make sure i'm reading this right: https://code.launchpad.net/~charm-helpers/charm-helpers/devel [16:14] b 40 [16:15] kwmonroe: yeah it was just that one commit [16:15] ack, thx tvansteenburgh [17:56] hml: is there a way to remove/stop the auto subnet scanning neutron subnets for network-space mapping? it keeps finding embarrassing duplicates and crashes the whole install [17:57] xarses: i’m not sure, let me see what I can find out. [17:58] rick_h: ^^^ any ideas? [17:58] * rick_h erads backlog [17:58] https://gist.github.com/xarses/307a07d290fcc9f48008b3ae1d192f05#file-duplicate-neutron-subnets [17:59] hml: xarses is that in the neutron charm itself? Or in juju trying to figure out what's up? [17:59] rick_h: i think it’s juju investigating subnets [17:59] juju bootsrap on a openstack cloud fails with a duplicate subnet in neutron [17:59] yea, gotcha [18:00] xarses: oh, try bootstrapping with the network uuid instead? [18:00] normally that's the bootstrap path ^ [18:00] xarses: i thought it was a different issue [18:00] no, its scanning networks in the output of `openstack subnet list` [18:01] the network to boot the instance selector is not the problem here [18:01] it's found 2 duplicates on me so far, and I'm guessing I have another 4 based on what happened here [18:02] xarses: sounds like bootstrap is failing when discovering subnets for juju - where juju list-subnets etc would be used [18:02] but the bootstrap haults on reading the duplicate [18:08] xarses: while we look for a workaround on this… can you file a bug on this please? [18:08] sure, on lp? [18:09] xarses: yes: https://bugs.launchpad.net/juju [18:09] xarses: I have an idea, but not sure if it’ll work - let me test out first [18:10] cool [18:13] yay, it only took 3 days, but I finally have a controller installed [18:13] xarses: w00t! sorry it took so long. openstack is one of the harder to bootstrap unforunately. [18:13] tvanhove: bad news! see the [IMPORTANT] thread here: http://mail-archives.apache.org/mod_mbox/bigtop-user/201708.mbox/browser. tl;dr, kafka was removed from the repos due to licensing. [18:13] hml: dont worry about the work around, I was more easily than expected able to remove the duplicates [18:14] xarses: good news - [18:17] ya, I wish it wasn't so stubborn, the main issue where with the image and tools metadata process being overly weird to the uninitiated [18:19] xarses: usually the tools aren’t an issue - the images are, because they are specific to your openstack. we can’t just download the info - it’s not an issue with other clouds. :-/ [18:19] xarses: that said, we’re looking to improve the docuementation on this [18:19] it was the offlining that I had to do for the tools to get them through whatever was wrong with the network that was causing them to time out [18:20] The network guys still haven't come back to me on that [18:20] xarses: Somewhat related is this bug (though that is concerning the generic subnet that is created for HA routers): https://bugs.launchpad.net/juju/+bug/1710848 [18:20] Bug #1710848: Bootstrapping Juju 2.2.x fails on a Openstack cloud with Neutron running in HA.

[18:21] oh, well that is about what I was to report [18:22] close enough anyway [18:27] I wacked it back to new, while confirmed might be just as valid, its not my project [18:28] so it doesn't languish in some incomplete filter [18:28] xarses: Great. Feel free to hit the "affects me too" button too. [18:28] Duplicate subnets is a pretty common scenario in OpenStack so Juju will need to handle that anyway [18:33] yep [18:39] xarses: good thing the subnets were easy to remove - my idea didn’t work. :-( [18:50] happens [20:53] thanks for your help hml, wouldn't have gotten throug this w/out it [20:53] xarses: glad I could help! [20:54] now I can continue with the getting started videos [20:54] =) [20:54] :-) [20:54] there’s a juju show on youtube also with different topics - maybe that’s what you’ve found? [20:56] https://www.youtube.com/watch?v=ovsBVZsQqtg [20:56] for 1.25 [20:56] which means a bunch of these commands aren't around anymore [20:56] 1.25 is a bit different than 2.0 [20:56] the concepts have changed too [20:56] ya, finding useful videos is is hard [20:57] most are a billion years old [20:57] I've ran into a few that appear to be 0.x [20:58] here’s one of the bi-weekly juju show: https://www.youtube.com/watch?v=YoZsP7TDyZI [20:58] let me look for me [20:58] more [21:00] ya, I've watched most of that and felt lacking from it [21:01] ah [21:03] that’s more on-going juju news rather than getting started [21:07] ya [21:58] zeestrat: hmm, a possible work around is to hide the duplicate networks from the user bootstraping the controller