=== thumper-lunch is now known as thumper === defunctzombie_zz is now known as defunctzombie === wedgwood is now known as wedgwood_away === defunctzombie is now known as defunctzombie_zz [04:37] Sander^home, not at the moment [04:38] Sander^home, the future notion is that you could preconfigure a whole set of services and relations as blueprint/stack and reuse that. === defunctzombie_zz is now known as defunctzombie === defunctzombie is now known as defunctzombie_zz [12:41] Hi [12:41] I have a problem with new machines staying in pending state forever (>1 day) [12:41] 35: instance-id: pending [12:42] The are new units to existing services [12:42] They [12:43] loooks like juju cannot create new vm. if you are using openstack as a backend, do check you can start new vm with nova. [12:46] melmoth: it is aws [12:47] I don't see the pending instances in their admin console [12:50] i never used aws, but it looks like juju was not able to start a new vm. [12:50] you mayt want to try to fire up a new vm manually, just to check it works, and may be use --verbose with juju, to spot error ? [12:53] how do I do that? And is there somewhere I can get logs? In node 0? [12:58] not sure how you do that on aws. for the logs, i use --verbose, i m not sure where to look on node 0 [13:08] "--verbose add-unit" gave no interesting information. The process probably happens asyncronly [13:08] juju debug-log neither === wedgwood_away is now known as wedgwood === bcsaller_ is now known as bcsaller [14:17] melmoth: hmm, now I see the problem probably. Machine 0 has: "agent-state: not-started" [14:17] Then not possible to SSH into, so I guess I need to force a restart of it === benji___ is now known as benji [14:38] I've got some open questions with bounties if someone is looking to help out: http://askubuntu.com/questions/267409/could-not-internally-obtain-zookeeper-handle [14:38] http://askubuntu.com/questions/267689/agent-state-pending-in-juju-node-null-public-address-associated [14:50] marcoceppi: hah, new gitlab release [14:50] didn't juan just promulgate your older one? === defunctzombie_zz is now known as defunctzombie [14:58] jcastro, no someone else took it over [14:58] They're releasing once a month now, and this new release completely drops gitolite as a dependency [15:01] gotcha [16:14] <_mup_> Bug #1158841 was filed: fails to notice (or retry) if there's an error connecting to launchpad < https://launchpad.net/bugs/1158841 > [16:31] What to do about persistent "agent-state: not-started" on the master machine? === deryck is now known as deryck[lunch] === salgado is now known as salgado-lunch === salgado-lunch is now known as salgado [17:29] How to add Landscape monitoring to machine 0 in juju | http://askubuntu.com/q/271298 === deryck[lunch] is now known as deryck [18:09] FYI, juju was just removed from Debian testing because of this FTBFS bug http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=702791 [18:11] IMO thats probably a good thing.. you don't want Debian users to be finding such an old'n busted juju for the next 2+ years [18:18] I agree, I prefer to try again with 2.0 [18:19] What to do when Juju machine 0 has got "agent-state: not-started" state | http://askubuntu.com/q/271312 [18:19] Ideas about the above question? My environment is completely broken now :/ [18:21] yeah, I think when the head node goes boom ... [18:21] Then you are lost? [18:21] Recreate environment is the only solution? [18:21] That must absolutely not happen if Juju is going to be ready for production use [18:22] yea, it's on the list to fix, as of now, the bootstrap node is not HA. :-/ [18:22] there may be a way to recover it though [18:22] hazmat: ideas? [18:23] Ssh to its IP works, but the SSH key is wrong [18:23] I'm not sure which ssh key "juju ssh" uses [18:23] jcastro: his head node isn't boom [18:23] SpamapS, 0.5.1 ouch.. [18:23] its working fine [18:23] the machine agent just didn't start [18:23] or died [18:23] hazmat: I know [18:23] hazmat: thats how Debian rolls [18:24] SpamapS: oh awesome, maybe we can recover [18:24] tho, experimental still just has 0.6 :-P [18:24] * SpamapS would love it if an actual juju user wanted to join the effort to maintain in Debian... :) [18:26] needs to support debian as a launched OS for people to care [18:27] Juju really should get out of the game of defining everything needed to run its agent [18:27] thats why Heat is more useful to me. [18:28] the agents needed in heat are like.. 200 lines of python. [18:28] and you can emulate them w/ curl calls [18:28] * SpamapS lunches [19:07] koolhead17: ask.openstack.org looks great dude! [19:08] jcastro, we made it finally [19:08] :D [19:09] jcastro, can we RT it from our ubuntu cloud handle :P [19:09] yeah [19:09] link me to your announcement please! [19:09] Am one of the mods there & would love if we get some of our hackers there [19:09] https://twitter.com/koolhead17/status/315177524005048320 [19:11] got it [19:15] jcastro, can we make an internal announcement to in our list? [19:16] the ubuntu server list? sure, why not! [19:17] jcastro, shoot it then. we need our folks in there too :P [19:49] jcastro, will you be at OS summit? [19:49] yep [21:18] jcastro :3 [21:33] https://juju.ubuntu.com/docs/subordinate-services.html#usage [21:34] what happens at the "juju deploy rsyslog" point? [21:39] sarnold: not much... charm is looked up, validated, cached in bootstrap node... sort of "staged"... but nothing's deployed on any service units until relation-time [21:39] m_3: cool, thanks :) [21:39] sarnold: /me wants to see that `juju deploy rsyslog --with ` [21:40] sarnold: but bigger fish atm [21:41] m_3: would that deploy the primary with the rsyslog already in place? or deploy the primary and then rsyslog? [21:41] sarnold: desired workflow for me would be `juju deploy primary && juju deploy sub --with primary` [21:42] m_3: aha [21:42] sarnold: but either way would work... just want fewer no-op steps [21:42] aternatlively: `juju deploy primary --with sub1,sub2,sub3` [21:42] but I prefer the former [21:42] makes it easier to split concerns between primaries and "aspects" [21:43] it's funny you mention that, I keep thinking of the (little) I've read about aspect-oriented programming every time I re-pick-up the juju documentation.. [21:44] sarnold: yeah, the paradigm really fits with infrastructure [21:44] peeps focus on primary infra [21:44] but there's a _lot_ of work on aspects that cut across all primaries [21:45] backups, log aggregation, authentication, etc [21:45] heh, I was just going to ask about backups, logging, updating, audit logs.. [21:45] firewalling? apparmor etc MAC configs? [21:45] sarnold: also comes up a lot in thinking about the gui... how to cull and/or slice and dice your live whiteboard of your infra [21:46] m_3: ooof [21:46] :) [21:46] m_3: I'm sure there's something beautiful there to show the layers of aspects on each service .. just good luck finding it :D [21:46] haha [21:46] yeah [21:46] but worth poking at over time [21:48] sarnold: it's interesting to consider how this concept of aspects fits into integrating juju with existing config mgmt tools [21:48] i.e., audit logs managed by puppet living together (orthogonally) in a juju-based infrastructure [21:49] but that's still open (and hard) [21:52] might be the only practical integration story though [21:53] it mght be the inevitable destination, but my instincts suggest it'll be herculean task to get it right [21:53] ack [22:09] openstack-dashboard login | http://askubuntu.com/q/271382 [22:53] there's a missing bullet point for "symlinks must be self contained within a charm" here https://juju.ubuntu.com/docs/policy.html === wedgwood is now known as wedgwood_away