[14:12] <roaksoax> rvba: around?
[14:13] <roaksoax> rvba: i'm seeing this: http://pastebin.ubuntu.com/5676753/ any thoughts?
[14:13] <rvba> roaksoax: yep…
[14:13] <rvba> roaksoax: btw, you still need to merge https://code.launchpad.net/~andreserl/maas/ipmi_usercreation_ilo_versions_trunk/+merge/148579 in trunk right?
[14:14] <roaksoax> rvba: yeah so the reason why that did land in 1.2 and not trunk is because julian thought it would be better to have the testing of that in trunk,s ince it didn't make sense for 1.2
[14:15] <roaksoax> rvba: and the above might be due to the nodegroup change?
[14:15] <rvba> roaksoax: right.  Julian and I agreed that we don't really have time to add the testing so let's land this as is and file a bug about the lack of test for this.
[14:15] <roaksoax> rvba: ok sounds good to me
[14:18] <rvba> roaksoax: the error you pasted is due to the fact that 2 nodegroups match when we're trying to determine to which nodegroup a node should be attached.
[14:18] <roaksoax> rvba: so same network?
[14:19] <rvba> Yeah, the IP of the connecting node belongs to a network to which for nodegroups are connected to.
[14:19] <rvba> Thus we cannot determine to which is should be attached.
[14:19] <rvba> You can probably fix this by removing the network in question on one side.
[14:20] <roaksoax> yeah makes sense, I was jus testing whether your changes do fix the issue, but I just realized I was managing interfaces in both nodes
[14:20] <roaksoax> so makes sense
[14:27] <roaksoax> rvba: so this is a result of the revert you provided me with yesterday, righ? Since the current escenario I'm testing was working before that change
[14:27] <roaksoax> rvba: and this means that we cannot have 2 clusters in the same network
[14:27] <allenap> roaksoax: I've just had a chat with rvba, and we have a plan...
[14:27] <rvba> roaksoax: exactly
[14:28] <rvba> roaksoax: you can have two clusters in the same network, but you have to manually remove an interface (in MAAS) on one of the nodegroups.
[14:28] <roaksoax> rvba: ok I know how to setup an environment with 2 different clusters in different networks that's quite easy to do with kvm :)
[14:28] <roaksoax> rvba: I see
[14:28] <roaksoax> allenap: cool! what's the plan :)?
[14:29] <rvba> roaksoax: to tell to MAAS: this network, although physically attached to 2 nodegroups, is actually managed by *this* nodegroup.
[14:29] <roaksoax> right! makes sense
[14:29] <roaksoax> rvba: since the above error shows up even when both networks are *unmanaged*
[14:29] <allenap> roaksoax: I'll change the find_nodegroup() function to prioritise managed interfaces, but fall back to non-managed interfaces. Additionally, if more than one non-managed interface matches, we blow up, suggesting that either one interface is marked as managed, or to remove the other interface.
[14:29] <roaksoax> err both cluster interface are unmanaged
[14:30] <rvba> roaksoax: indeed, you need to *remove* one interface on one of the nodegroups/
[14:30] <rvba> In your case, you don't have any managed interface.
[14:30] <roaksoax> rvba: right, say the master running in the region controller should not have any interface in the same network as the other cluster (bnoth managed/unmanaged)
[14:31] <rvba> roaksoax: exactly.
[14:31] <roaksoax> allenap: makes sense to me. But wouldn't it make better sense that if there's 2 cluster in the same network, to simply use 1 of those 2?
[14:31] <roaksoax> rvba: ok makes sense. Ok will setup an environment and test
[14:32] <allenap> roaksoax: That might be okay. I can't think of an issue with that offhand. rvba?
[14:33] <rvba> allenap: seems reasonable, as long as this only happens when no managed interfaces on this network are found I think that's ok.
[14:34] <allenap> rvba: If there's a managed interface, we will choose that one anyway. What if we find two managed interfaces for the same network?
[14:35] <roaksoax> if we found 2 managed interfaces in the same network that most likely would be the case that i'm using different ranges for DHCP. (which is what I was doing right now).
[14:35] <roaksoax> i guess it would make sense to try to detect from which one the node pxebooted from
[14:35] <roaksoax> if not fallback to the first one
[14:36] <allenap> roaksoax: When an interface is managed, the DHCP ranges are defined by the netmask, iirc, so you can't have two cluster controllers managing the same network.
[14:37] <rvba> Indeed, I don't think that's possible in MAAS right now.
[14:37] <roaksoax> allenap: right, but I can have 2 DHCP servers saying "Server IP between 192.168.1.100 - 200" and another one that will server "192.168.1.1 - 99"
[14:37] <roaksoax> s/server/serve
[14:40] <allenap> roaksoax: You /can/ do that, but it's not a supported configuration: the interface should be marked as unmanaged.
[14:40] <roaksoax> alright :)
[14:42] <allenap> Finding a node's nodegroup (aka cluster controller) in a nutshell: 1 managed interface on the network = choose this one; >1 managed interfaces on the network = misconfiguration; >1 unmanaged interfaces on a network = choose any
[14:43] <allenap> Version 2 of Finding a node's nodegroup (aka cluster controller) in a nutshell: 1 managed interface on the network = choose this one; >1 managed interfaces on the network = misconfiguration; 1 unmanaged interface on a network = choose this one; >1 unmanaged interfaces on a network = choose any
[14:49] <roaksoax> allenap: sounds good to me
[14:49] <allenap> Cool.
[14:51] <roaksoax> allenap: oh btw... i just made this change to python-tx-tftp in raring. Spads was getting his connection dropped and him and Daviey investigated and came up with the solution: http://pastebin.ubuntu.com/5676874/
[14:53] <allenap> roaksoax: That's cool :) I'll submit that upstream (and also press for landing of the other patches).
[14:54] <roaksoax> allenap: that would be great!
[15:29] <AskUbuntu> Maas Web Interface not reachable after new installation! | http://askubuntu.com/q/277874