=== scuttlemonkey is now known as scuttle|afk
[01:39] <Budgie^Smore> are we having fun yet?
[01:40] <rick_h> Budgie^Smore: but of course
[06:56] <kjackal> Good morning Juju world!
[07:10] <erik_lonroth_> kwmonroe: No, its stuck in "configuring spark in yarn client mode" and some of the spark tests included in the actions of the bundle fails related to spark. I'll try post it again soon as this is really close to work for us.
[07:29] <kjackal> hi erik_lonroth_ based on https://github.com/apache/bigtop/blob/master/bigtop-packages/src/charm/spark/layer-spark/reactive/spark.py#L167 in the juju logs you should be seeing something like that  "Configuring spark with deployment matrix" can you share this so we know why is the charm trying to do?
[08:05] <boolman> how do juju choose which IP to use as dns/agent-address? on my deployment it uses the wrong one and I end up with nodes that are unreachable
[08:27] <kjackal> boolman: I do not think you have a control over the IP unless you use the manual provider (aka create the machine/unit yourself and then add it to juju). There is however the concept of network spaces: https://jujucharms.com/docs/1.25/wip-spaces
[08:27] <kjackal> Havent used spaces myself but could help you specify your network requirements
[08:28] <kjackal> boolman: sorry this is the right url for spaces: https://jujucharms.com/docs/2.1/network-spaces
[08:39] <boolman> kjackal: the problem I'm having is that I have nodes with 4-5 subnets/vlans and I have no real control over what IP the juju agent gets. so You're saying that I need to configure the nodes myself? via curtain for example
[08:40] <kjackal> boolman: I think you should try configuring network spaces
[08:40] <boolman> I have configured network spaces, or what do you mean by that?
[08:41] <boolman> I have 1 space per purpose
[08:42] <kjackal> boolman: awesome, when deploying with the network space constraint you get an IP from a subnet you donot want?
[08:43] <boolman> where do I specify constraint, on the machine or on the services/units?
[08:43] <boolman> but does that actually force the subnet on which the juju agent uses?
[08:43] <kjackal> boolman: you should be able to do something like "juju deploy mediawiki -n 2 --constraints spaces=cms"
[08:46] <boolman> I'm not sure if I follow, on the services I'm configuring binding: space: myspace and I guess thats working. but since the nodes get the wrong DNS/IP the agents are unreachable
[08:47] <boolman> so its not deploying shit right now
[08:49] <boolman> if I delete that interface on the node, it chooses another random ip/subnet
[08:55] <kjackal> boolman: can you show me how you created the network spaces?
[08:55] <boolman> they are created from maas
[08:57] <kjackal> I see... so juju list-spaces and juju list-subnets gets you anything?
[08:58] <boolman> http://ix.io/xvt
[09:02] <kjackal> ok so if you do a "juju deploy mediawiki --constraints spaces=internal" you get a machine with an IP on 10.1.3.0/24 but it is not reachable by the juju controller?
[09:03] <boolman> i will get an IP of 8.8.8.0/24 range
[09:04] <kjackal> that is strange... so you get a machine on the front-end space although you requested the internal.
[09:06] <kjackal> boolman: can you try this other syntax: "juju deploy mysql --bind db-space" ?
[09:07] <boolman> this seems pretty dead on https://bugs.launchpad.net/juju/+bug/1473069
[09:07] <mup> Bug #1473069: Inconsistency of determining IP addresses in MAAS environment between hosts and LXC containers <2.0> <addressability> <bug-squad> <cdo-qa-blocker> <landscape> <lxc> <maas-provider> <networking> <uosci> <juju:Fix Committed> <Landscape Server:Invalid> <https://launchpad.net/bugs/1473069>
=== scuttle|afk is now known as scuttlemonkey
[14:27] <vds> I have a subordinate charm that failed on some units, https://paste.ubuntu.com/24864747/ , I know I can't remove the subordinate charm unit without removing the principal, but the system is working fine and the error doesn't really have any effect, what other option do I have to fix the deployment?
[14:39] <SimonKLB> vds: that's actually not the case, if you remove the relation to the subordinate charm it should be stopped and removed from the machine
[14:55] <kwmonroe> yeah vds, in the just released juju-2.2, removing sub relations should now remove the sub unit as well.  bug 1686696 for reference.
[14:55] <mup> Bug #1686696: Subordinate units won't die until all principal relations are gone <juju:Fix Committed by 2-xtian> <https://launchpad.net/bugs/1686696>
[14:55] <vds> SimonKLB kwmonroe thx!
=== frankban is now known as frankban|afk
[16:31] <jay_> Hello all, would like to know if anyone has ever tried taking a running docker container, and then adding it as a machine with juju? I gave it try, but come across some errors that don't really give my much insight into what might be wrong. The error just states something about an invalid entity name or password. Not much to work off of.
[17:17] <rick_h> jay_: so I think the big thing is that Juju expects to be able to download, install, and modify things in the container in such as way that normally you'd build a new image to launch.
[17:17] <rick_h> jay_: so I don't know that the juju way of "give me a machine and I'll get it all set up" works.
[17:18] <rick_h> jay_: now there's definitely some ideas around a different kind of charm or something down the road that might play along nice with other Juju managed applications but that's idea stuff atm
[20:01] <Budgie^Smore> O/
[20:05] <lazyPower> Budgie^Smore: \o