[05:06] <melmoth> where to look at log if nothing happens when i pxe boot a node for the first time ? By nothing happen i mean, nothing happen on the weeb interface
[05:06] <melmoth> the node get the image, seems to runs some stuff, then power itself off
[05:06] <melmoth> at this stage, i was expecting to see it as 'declared' in the web interface
[05:06] <melmoth> but nope, nothing.
[05:06] <melmoth> and i do not find obvious error messages.
[05:11] <kurt_> melmoth: this can sometimes be due to clock sync errors
[05:12] <kurt_> one host is too far off time
[05:13] <melmoth> hmmmm
[05:14] <melmoth> that s going to be fun to check
[05:14] <kurt_> bios?
[05:15] <melmoth> do you know roughly how much time change is suppose to be ok (secondes, minutes, hours), or even 1 sec could be problem ?
[05:15] <kurt_> from the problems I've seen, it has to be significant, like hours.
[05:16] <kurt_> but I'm not an expert in this area
[05:16] <kurt_> I've not tested how many is too much
[05:27] <bigjools> IIRC it's an hour or so
[05:28] <bigjools> however
[05:28] <bigjools> this bug was fixed ages ago, if you are using any up-to-date maas it won't be a problem
[05:28] <bigjools> melmoth: if there's nothing in the maas logs, can you see anything on the node's console?
[05:37] <melmoth> hmmm, let see what happen after tftp is actually listening on the right interface....
[05:37] <melmoth> (and i wonder now that i realise it was not, how the node was able to download somethign to boot from)
[06:02] <kurt_> bigjools:  do you know how to force juju to deploy to particular nodes?  I'm thinking from the MAAS perspective and need to run multiple charms on a single node.
[06:03] <bigjools> kurt_: juju cannot deploy multiple charms to a single node
[06:03] <bigjools> there's some work coming up in juju-core that allows lxc creation which will do it though
[06:04] <kurt_> How does one create a "controller" node with all of its related services for openstack?  Just not possible?
[06:06] <bigjools> sorry I don't understand what you mean
[06:06] <kurt_> Have a look at the network diagram here:
[06:06] <kurt_> https://github.com/mseknibilel/OpenStack-Folsom-Install-guide/blob/master/OpenStack_Folsom_Install_Guide_WebVersion.rst
[06:07] <kurt_> Notice the compute node, network node, controller node all run consolidated services
[06:10] <bigjools> you're probably better off asking on the #juju channel.
[06:12] <kurt_> Another question please - do most people NAT their connections for the MAAS clients via the region controller?
[06:12] <bigjools> you mean api clients?
[06:13] <bigjools> or are you talking about the nodes?
[06:13] <kurt_> nodes
[06:13] <bigjools> you can do what you like with them, it's down to your own network setup
[06:13] <kurt_> I'm struggling with some network aspects of running openstack on top of MAAS
[06:14] <bigjools> there's a charm to deploy openstack on maas somewhere,  but again, #juju :)
[06:15] <kurt_> currently, I have the maas nodes NAT via the region controller (single node).  That gives them the internet access they need, but it causes further problems down the chain when trying to make quantum work
[06:16]  * bigjools has to go
[06:16] <kurt_> thanks anyways
[10:34] <MACscr> so i just installed maas, install finished and now im trying to get to the gui and i get a generic internal error
[10:35] <MACscr> though i dont see any apache logs, lol
[10:35] <MACscr> oh, nvm, got it
[10:35] <MACscr> had to sudo su to find them
[10:40] <MACscr> guess mass wont start: DBusException: org.freedesktop.DBus.Error.NameHasNoOwner: Could not get owner of name 'org.freedesktop.Avahi': no such name
[10:54] <MACscr> got it, was an rlimit issue in the daemon
[12:25] <AskUbuntu> Can we get Xenserver VMs provisioned by MAAS? | http://askubuntu.com/q/337425
[15:33] <kentb> argh! so, what's the best way to debug why ipmi no longer works on my maas setup?  I did an update to 1.2+bzr1373+dfsg-0ubuntu1~12.04.2 and now I can't power up any nodes from the WebUI, although I can run ipmipower and bring the nodes up and down at will :-/
[15:44] <kentb> ugh. nevermind, had to restart maas-cluster-celery, maas-region-celery, and maas-pserv to get it going again
[18:42] <addisonj> so, getting started with maas, is it possible to do anything without dhcp supporting pxe? If I manually add a host via mac, can I commission it that way?
[19:48] <roaksoax> addisonj: no, the nodes always have to PXE boot, there's no other way to tell the nodes what to execute if they don't pxe boot
[19:49] <addisonj> okay, figured as much, so antoher question, is there are problem with using a maas-managed dhcp range to also boot normal clients (laptops and the like)
[19:49] <roaksoax> addisonj: no not really. Though, the maas managed DHCP server will tell them to PXE boot
[19:49] <roaksoax> addisonj: unless you disable network booting from the BIOS
[19:50] <addisonj> right, and if I don't accept the nodes, it won't matter to much anyway, right?
[19:50] <roaksoax> addisonj: exactly. So you could just tell those nodes to look the HD as the first boot device too
[19:50] <roaksoax> and they will not pxe boot
[19:54] <addisonj> okay, so now, onto dns, does maas rely on DNS? or can it run just with IP addresses?
[19:56] <addisonj> hrm... so this is odd, my server is trying to pxeboot, getting a message: "trying to load: pxelinux.cfg/default" and this is hitting the correct dhcp server
[19:56] <addisonj> but just hanging
[20:15] <addisonj> ah, I had edited some settings in importing pxe files because I don't need ARM and also took at i386, turns out it defaults to i386
[20:46] <roaksoax> addisonj: it does rely on DNS
[20:46] <roaksoax> addisonj: yeah so you'
[20:46] <roaksoax> addisonj: yeah so you'll need both i386 and amd64 otherwise it will fail
[20:46] <roaksoax> there's a bug
[20:46] <roaksoax> bigjools: ^^
[21:23] <addisonj> okay, so, have my nodes going (hooray!) but I misnamed one of them. Doing everything as root, can't seem to delete or change the name
[21:37] <MACscr> is there any way to use diskless nodes with maas? Pretty much their storage would be nfs or iscsi (im using ceph block and object storage for the openstack cluster)
[23:45] <bigjools> MACscr: it's on the roadmap to go diskless at some point but no immediate plans. However you could implement it cheaply by setting the root filesystem as a kernel option on the node.  That's already done for the enlist/commission step using iScsi
[23:45] <bigjools> addisonj: you can edit the node's name in the web ui
[23:46] <addisonj> bigjools: when I try that, it tells me I can't because it is currently in use
[23:46] <addisonj> is there some way to take it back into the queue?
[23:46] <bigjools> roaksoax: yes there's an open bug for that
[23:46] <bigjools> addisonj: and is it in use?
[23:47] <addisonj> bigjools: it is commissioned and installed, but turned off
[23:48] <bigjools> addisonj: why is it off?  did you manually turn it off or did you try to release it from the assigned user?  in fact how did you start it?
[23:49] <MACscr> bigjools: is there any documentation on this kernel option?
[23:49] <bigjools> I am guessing that you started it from the UI and then turned it off
[23:49] <addisonj> I turned it off manually, but I actually turned it on manually as well. WOL is not triggering it for some reason (might have to do with some traffic filtering)
[23:50] <bigjools> what state does the ui say that the node has?
[23:51] <addisonj> "Allocated to root"
[23:52] <bigjools> ok then you need to use maas-cli to release it
[23:52] <bigjools> http://maas.ubuntu.com/docs/maascli.html#node
[23:52] <bigjools> maas-cli <profile> node release <node id>
[23:53] <addisonj> ah, so is "Allocated to <user>" the correct running state? or because WOL is failing I am getting some funky?
[23:53] <bigjools> allocated means that maas thinks someone is using it
[23:53] <bigjools> that happens when you click "start" in the UI
[23:53] <bigjools> or e.g. juju starts it up using the ap
[23:53] <bigjools> i
[23:54] <bigjools> "release" puts it back to the Ready state and then you can rename it
[23:54] <addisonj> so why don't I see a release button?
[23:54] <bigjools> because there isn't one
[23:54] <addisonj> ah
[23:54] <bigjools> bug :)
[23:54] <bigjools> brb
[23:55] <addisonj> many thanks, and, also a very cool project :)
[23:55] <addisonj> got it running in like 4 hours and 8 nodes provision, and that makes me happy
[23:56] <bigjools> MACscr: not sure, let me lool
[23:56] <bigjools> addisonj: great!