=== CyberJacob is now known as CyberJacob|Away === mwhudson is now known as zz_mwhudson === zz_mwhudson is now known as mwhudson === CyberJacob|Away is now known as CyberJacob === mwhudson is now known as zz_mwhudson === jam1 is now known as jam === zz_mwhudson is now known as mwhudson === mwhudson is now known as zz_mwhudson [09:56] jtv1: default zone branch is up for review https://code.launchpad.net/~rvb/maas/default-zone/+merge/200635 [09:56] jtv1: I'm reviewing your maas-test branch now [09:56] Thanks. I'll have a look at yours. [09:57] Ta. [10:55] jtv1: gmb there is still something left to do with the AZ stuff: update the documentation (to account for the default zone thingy) and add a node saying this feature is only available in the version published in Trusty. [10:55] s/node/note/ [10:57] Ah great. [13:29] hi regarding )tomixxx, https://help.ubuntu.com/community/Installation/FromUSBStick#live-usb from yesterday: does this mean i have to INSTALL ubuntu on my nodes ? === jtv1 is now known as jtv [13:32] tomixxx: no, you shouldn't have to do that. Misunderstanding? [13:32] On your servers though, yes please. :-) [13:33] roaksoax: turns out we got unlucky and cut the Trusty package just when we had that one script briefly in the wrong package. What do I need to do to set that right? [13:34] jtv: i have one server and 2 nodes. the server has both, windows 7 and ubuntu server tls 12.04.3 installed, the 2 nodes have only windows 7 installed [13:35] jtv: i already have the fix for it. it is just a simple Conflicts/Breaks [13:35] tomixxx: maas installs ubuntu on the nodes remotely (by netbooting them into install images) [13:35] jtv: y, but woke-on-lan is not working correclty, so my maas-server cannot boot the nodes [13:36] That's annoying. If they can still netboot, you can turn them on manually — but it's not going to be as nice. [13:37] (Wake-on-LAN also has the problem that it can't power nodes down again) [13:37] but if i start the nodes manually, nothing happens, except windows 7 is starting [13:37] and the nodes tay "commissioning" in maas web interface [13:37] Are they set up to boot off PXE? [13:38] good point [13:38] What happens there is: MAAS has tried to power the node up, and expects the node to netboot off the server. [13:38] i understand, i will check boot order of my nodes [13:38] Very important. :) [13:38] BTW be aware that that will install the system right over the nodes' existing OS. [13:39] So make sure there's nothing on those nodes' disks that you want to keep! [13:39] kk [13:42] hmm i cannot select booting from network in bios/boot options [13:42] Ouch. [13:42] Does it say why not? Or is the option just not there? [13:43] option is not there [13:43] Maybe the BIOS has a separate option where you need to enable netbooting first, before it shows up in the boot order at all? [13:43] kk, will check [13:46] there is an option "onboard lan boot rom" -> "disabled" [13:48] oh, sth going on here after "enabled" this option [13:49] ? [13:49] New entry in boot order? [13:50] jtv: is it ok if i have now sth like the following: 1.) ubuntuI686nov10, 2.) ubuntuI686feb10,...5.) Acronis Partition.... boot: [13:51] That's the boot loader? [13:51] What matters is the boot order in the BIOS though. [13:51] (btw, there was no new entry in boot order but i disabled all other and there is an option "enable other boots option" [13:51] ok, so this does not come from maas? [13:52] I don't think so... Are you seeing that list in your BIOS? Or in your boot loader? [13:52] no, normally not [13:52] btw: which power type i should select when i turn on the nodes manually? [13:52] currently, i have "wake-on-lan" selected [13:52] Keep that one. [13:52] ok [13:53] I'm still curious where you got that list... you said you have something like 1) ubuntuI686nov10, 2.) ubuntuI686feb10 etc. Is that in the BIOS boot-order settings? Or elsewhere? [13:53] ah [13:54] i just shut down the node, turn on again and now i can select "network: MBA v.12.2.5 slot 0200" in the screen "please select boot device" [13:54] Ahh [13:54] tomixxx, nope, I was suggesting that you use the live usb stick to test the wake-on-lan in your network before you try again with MAAS [13:56] matsubara: kk, but jtv told me that it might work with manually booting the nodes, so i try it now manually [13:57] Which reminds me... I bet there's a BIOS option for enabling wake-on-lan as well. [13:58] when i choose booting from "network: MBA v.12.2.5 slot 0200" i come into the same selection than before:1.) ubuntuI686nov10, 2.) ubuntuI686feb10,...5.) Acronis Partition.... boot: [13:58] but its not the "please select boot device" screen [14:00] a screen that tells about the "SystemImager autoinstalls system v3.8.2" [14:00] it seems these come from other servers of my university network [14:01] You've got other DHCP servers on the same network? [14:01] y [14:01] in my maas i have the interface "unamanged" [14:02] Then you also need to do some configuring on the DHCP server. [14:02] so i have to configure DHCP in the maas-webinterface? [14:03] Definitely the best option, yes. [14:03] It does mean that your nodes have to be on a network of their own, where MAAS provides them with their DHCP. [14:03] But if you're running multiple bootp servers for different purposes on the same network, I imagine things could get a bit confused. [14:04] architecture is, that my 2 nodes are connected to my one maas-server with a SWITCH and the SWITCH is connected with the network of the university [14:06] It should be possible to configure the DHCP server to say "when you netboot, please netboot from the maas server." But it sounds as if that might interfere with your network setup. [14:07] a question: which MAC address do i have to enter in the "wake-on-lan" dialoge when editing a node [14:07] ? [14:07] the mac address of the node or the mac address of the maas-server? [14:12] tomixxx: the node's. [14:12] ok [14:16] so, should i try to set DHCP settings in maas-server? [14:17] If you have a network that you can serve DHCP on, yes. But be careful not to serve DHCP arbitrarily on the university network! [14:18] Sysadmins don't take kindly to that sort of thing. :) [14:18] Everything may seem to work just fine, but then machines elsewhere start renewing their DHCP leases, and getting ones from your server instead of the one that's meant to manage the network. [14:19] Or a machine that's meant to netboot off the departmental server tries to boot off the maas pxe server, and won't boot... [14:20] ouch ok [14:20] A "rogue" DHCP server (as it's called) is a really serious problem, unless the network is carefully managed to prevent it. [14:21] so, i guess its better to let cluster interface "unmanaged" ? [14:22] The best thing to do is to have the maas on a separate network, behind a router, and let maas manage the interface. [14:23] If that's not possible, be aware that you'll need to tweak the configuration of the DHCP server that's already there. [14:23] do i have to enable wake-on-lan in bios only on the nodes or also on the maas-server? [14:23] You only need that on the nodes. [14:23] ok... [14:23] The server is really a pretty conventional thing. [14:26] so, what could be the best next step now? [14:27] Talk to the network admins. What you need is either: [14:27] (a) A network where your maas server can act as the DHCP server, without affecting the network. [14:27] Or [14:28] (b) *If* it isn't a problem for the network, a configuration change on the existing DHCP server, to tell clients that want to netboot that they should netboot off the maas server. [14:29] kk [14:30] Unfortunately you're in a kind of gap between two kinds of use: "private network" and "dedicated network." [14:31] k [14:36] ok, following hotfix idea: [14:36] what is, if i remove the network cable which go out of my swith to the uni network? [14:37] then i have my private network, not? [14:37] The nodes will have to install packages. [14:37] k [14:37] and therefore, the need i-net connection... [14:37] Now, if you could do it so that your maas server was still on the uni network, it'd work. [14:37] Because the nodes should only talk to a proxy running on the maas server, never directly to the internet. [14:38] and is it possible to pre-install the OS on the maas-server while internect connection is available, and then put away the outgoing network cable? [14:39] In principle, yes. But not easy yet. There is something we could try. [14:39] For commissioning, as I recall, the nodes only need some lldp packages. [14:39] jtv: did you hit the upgrade bug? [14:40] roaksoax: not me, but Raphers did. [14:40] cool [14:40] jtv: note that the file will now be present in both maas-region and maas-cluster i think (because python-maas-provisioning server will be isntalled in both) [14:40] tomixxx: so, if on any machine you could set the http_proxy to be the MAAS server, and then install those same packages, you should be able at least to get through commissioning without hitting the internet... In principle; [14:41] roaksoax: that's fine, thanks. We don't actually use it yet for anything inside maas itself, but we need it for maas-test. [14:42] tomixxx, how many network interfaces do you have in the MAAS server? [14:43] matsubara: maas-webinterface is showing me one cluster-controller with two interface: eth0 and another one, without a name [14:43] Ahh yes, good one! The server could act as a bridge. [14:43] tomixxx, so, connect one of the interfaces to your uni network, the other to the switch and both nodes to the switch [14:43] that way you can test your MAAS stuff without breaking hell into the uni's network [14:44] jtv: you moved it from maas-dhcp right? [14:44] Yup. [14:44] Thinking nobody would be releasing it juuuuust then. :) [14:45] matsubara: if i do ifconfig in terminal, the second interface is called "lo" and link encap is local loopback [14:45] so, is this really a interface? [14:45] jtv: in reality this would only be a problem if someone installs what's currently in trusty and upgrades to something else (and technically, upgrades to something that's in trusty). [14:45] since I'm not expecting anyone using trusty just yet... that shouldn't really be a problem [14:45] but just in case, i'm adding this fix [14:45] I figured there's probably a testing setup somewhere that could be hurt by this. [14:46] If not, yes, I'm perfectly willing to forget the whole thing! You'll know more about this than I do. [14:47] jtv: https://code.launchpad.net/~andreserl/maas/packaging_upgrade_fix_dhcp/+merge/201937 [14:47] tomixxx, ifconfig -a, I think will show you all the interfaces. even the ones that are down [14:47] Grr... why do developers never listen to my suggestion for tab password completion? It would save so much time! [14:47] you probably want eth0 and eth1 or something like that [14:48] matsubara: i only have "eth0" and the mentioned "lo" [14:48] after executing ifconfig -a [14:49] So not enough to build your own private kingdom. :/ [14:49] tomixxx, yeah, only one interface then :-( [14:49] kk [14:49] Still, ethernet cards in a large network shouldn't be hard to get! [14:49] indeed, even an ethernet usb adapter should be pretty easy to find [14:49] I'm not suggesting that you open the machine of a colleague who's not in right now and take one, but... [14:50] in this room, there are only 3 nodes ^^ ofc, i additionally i have my laptop if this help sth... [14:50] See if the admins have one lying around? [15:13] ok, i have got a network interface card from my sysadmin [15:13] i will plug it now into the maas-server [15:24] ok, card is plugged in now [15:29] ok, maas server and 2 nodes are now connected to the switch. the cable from the switch to the uni network is UNPLUGGED. a new cable goes from the 2nd interface of the maas-server to the university network [15:29] tomixxx: hrhr, this is the exact same setup that we run as well [15:30] nat all the things [15:31] tomixxx: which university are you at? [15:31] university of klagenfurt [15:31] in austria [15:32] cool, university bremen here (northern germany) [15:32] hehe [15:33] turned on the maas-server again, do i have to do sth when it says "waiting for network configuraiton" ? [15:33] ah sorry [15:35] if i do "ifconfig -a", i can see eth0 and eth1 [15:37] tomixxx, you can run dpkg-reconfigure maas-region-controller to reconfigure MAAS. it'll ask what's the address of the pxe address, complete it with the value you get from the interface not connected to the uni network [15:38] matsubara: it seems, network cards now have no IP so far [15:38] hmm but before you do that you might need t configure your internal network appropriately first [15:38] i cannot connect to i-net [15:39] but when i set dhcp settings in maas-server, i have already configrued my internal network, not? [15:42] tomixxx, right, so first you need to configure the network. Open /etc/network/interfaces and do something like this: http://pastebin.ubuntu.com/6762617/ [15:42] assuming eth0 is the interface connected to the switch and eth1 is the interface connected to the uni network [15:43] tomixxx, then you sudo service networking restart and check with ifconfig if the IP is properly set for each interface. [15:44] ok, first i have another problem: if i start windows i can see there is no driver installed for the new network card [15:45] (i have both, windows and ubuntu installed on the maas-server ;) [15:46] tomixxx, Linux in general have pretty good support for network interfaces, so usually you don't need to install anything else. [15:46] tomixxx, so, booting up into windows doesn't really help configure the MAAS server [15:46] kk [15:47] i was a windows user all the time, so when things get unfamiliar, i tend to boot windows again [15:48] tomixxx, https://help.ubuntu.com/13.10/serverguide/networking.html has plenty of information on how to configure the network [15:48] but a question: what is if i dont know if eth0 is the card which is connected to switch? [16:00] tomixxx, well, you can just try and if doesn't work, you reverse the order. The basic idea is to have the interface that's connected to the uni network set up to use dhcp and the interface connected to the switch with a static address [16:01] hmm ok, currently i have edited /etc/network/interfaces and both set to dhcp [16:01] but i-net not working so far [16:04] tomixxx, does it get an IP at all? Also, another thing to keep in mind is, if your uni dhcp server is serving IP address in the same range of the static address things won't work. [16:05] i have started now sudo ifup eth1 resp. eth0 [16:05] i guess i need this? [16:06] when i say sudo ifup eth0, response was: "interface eth0 already configured", sudo ifup eth1 still not finished... [16:08] jesus, what about the promissed 20-minute-install time for the whole cloud as written on the ubuntu website ;) [16:10] interesting: when i do ifconifg -a, i cann see sth like "eth0:avahi" and "eth1:avahi" [16:10] additionally [16:10] to the 2 interfaces [16:12] but i have still no internet on my maas-server :( and it seems the interfaces only have ipv6 addresses [16:17] tomixxx, I'd suggest get some help from the sysadmin that got you the new interface card. The uni's admins knows how the uni network really works and once you get that configured we can get back to MAAS [16:19] y, u are right, this has nothing to do with maas what iam posting here... [16:29] interesting thing, if i plug network cable from the uni to the onboard-ethernet-card, internet works [16:40] tomixxx: btw you might want to be carefull, when activating dhcp [16:41] the sysadmins might come and hurt you if you break university wide dhcp, because you broadcast in the same range they do [16:41] you want to set up nat [16:41] https://help.ubuntu.com/community/Router [16:42] you can check which card is which by calling ifconfig in the terminal [16:42] at the moment, i have the problem that one of my two network cards does not find correct drivers [16:47] ok guys [16:47] eth0 ---- dhcp to uni, internet works [16:47] eth1 ----- static ip as suggested 192.168.1.1 [16:52] in maas-web-interface, i guess i have to delete eth0 from cluster-controller and add eth1, right? [16:54] matsubara, jtv, someone online? [16:57] tomixxx, I'm here but a bit busy at the moment. Would you mind bringing this up in the maas mailing list or askubuntu? That way the solution will be archived and others might be able to jump in to help as well. [16:58] tomixxx, but you're right, delete eth0 from the cluster-controller and add eth1 and make it manage dhcp for you. === bjf[afk] is now known as bjf === zz_mwhudson is now known as mwhudson [20:37] greetings, I am going to be using maas for our deployments for our clusters and wanted to see if there is a more "architecture" friendly document for maas [20:37] I really want to understand the regional controller / cluster/ cluster controller better [21:05] CarlosC: this might help: http://www.roaksoax.com/2013/05/getting-started-with-maas-and-juju [21:07] roaksoax: very nice, thanks [21:07] let me digest this and I'm sure there will be more questions to follow ;-) [21:21] ok...so a cluster is the smallest unit for managing nodes...how is HA achieved at the region controller level? === CyberJacob is now known as CyberJacob|Away