| bigjools | bjf: yeah it's a packaging bug, is filed somewhere | 01:59 |
|---|---|---|
| bjf | bigjools, cool, just wanted to let you know in case you wanted me to file a bug for it | 03:02 |
| smoser | roaksoax, tomorrow. sorry. | 03:50 |
| === mwhudson is now known as zz_mwhudson | ||
| bigjools | hi :) | 04:26 |
| Lord_Set | Hello | 04:26 |
| Lord_Set | Getting some awesome pxe errors | 04:26 |
| Lord_Set | 2014-02-11 20:01:17-0800 [HTTPPageGetter,client] Logged OOPS id OOPS-9168ac1b2e875425f6b7ee91899da19a: No exception type: No exception value | 04:26 |
| Lord_Set | 2014-02-11 20:01:17-0800 [HTTPPageGetter,client] Logged OOPS id OOPS-e3d0e778908b133237b872870492049b: ParserError: while parsing a block mapping | 04:26 |
| Lord_Set | in "/etc/maas/pserv.yaml", line 60, column 5 | 04:26 |
| Lord_Set | expected <block end>, but found '<block mapping start>' | 04:26 |
| Lord_Set | in "/etc/maas/pserv.yaml", line 68, column 6 | 04:26 |
| bigjools | did you modify pserv.yaml? | 04:27 |
| Lord_Set | I uncommented the lines that shouldn't have been commented out like the tftp server lines | 04:27 |
| Lord_Set | Didn't touch anything else | 04:27 |
| bigjools | it uses defaults, you don;t need to touch them | 04:28 |
| bigjools | can you paste me your config? | 04:28 |
| Lord_Set | Sure | 04:28 |
| bigjools | or better, revert and try again and see if it workd | 04:28 |
| bigjools | works | 04:28 |
| bigjools | Lord_Set: can you re-comment those lines | 04:34 |
| bigjools | they don;t line up with the block so it's confusing the yaml parser | 04:35 |
| Lord_Set | Sure | 04:36 |
| Lord_Set | That worked! thanks | 04:38 |
| bigjools | cool | 04:38 |
| Lord_Set | Not sure why it had those uncommented | 04:38 |
| bigjools | me neither | 04:38 |
| Lord_Set | Doing a big test deploying on a full rack for Openstack deployment | 04:44 |
| Lord_Set | err deployment | 04:44 |
| bigjools | nice | 04:45 |
| Lord_Set | The rack is at SwitchNAP in Vegas :) | 04:45 |
| bigjools | roaksoax: I took the liberty of implementing the packaging change for the tool renaming: https://code.launchpad.net/~julian-edwards/maas/packaging/+merge/205901 | 06:54 |
| bigjools | waiting for your review | 06:54 |
| === CyberJacob|Away is now known as CyberJacob | ||
| rvba | bigjools: I'm using the package from the daily PPA and a node enlistment just failed with: http://paste.ubuntu.com/6918675/. I'm guessing maas-enlist needs updating… ? | 07:55 |
| Lord_Set | Quick question for you guys... What do you reccomend for remote power management and provisioning of MAAS nodes? WoL? IMPI? ILO 4? | 09:09 |
| Lord_Set | We are running most HP 5th and 6th generation servers I believe with ILO 2 natively built in. | 09:10 |
| Lord_Set | Also outside of that any suggestions or recommended paths for openstack, hadoop, and savannah deployment once the NAAS nodes are deployed? | 09:11 |
| Lord_Set | Our original test platform was with Mirantis but decided to move away from it due to its inability to provision across geographical locations easily. | 09:12 |
| jtv | Lord_Set: generally IPMI. Don't know about HP specifically — no experience with the ILO driver, but it's supposed to have some peculiarities. | 09:21 |
| bigjools | Lord_Set: ipmi2 will work OOTB | 09:57 |
| bigjools | on HP | 09:57 |
| rvba | jtv: well, I guess what we've done must be standard for packaged Django applications. But if you think we can do better, it's worth talking to Andres about it. | 10:16 |
| jtv | Well seeing now that we didn't use peer auth, I would guess that something in the infrastructure doesn't support it. In fact it rings a vague bell. | 10:17 |
| rvba | jtv: just filed https://bugs.launchpad.net/maas/+bug/1279304 | 11:39 |
| ubot5 | Launchpad bug 1279304 in MAAS "Node commissioning results are not displayed in the UI" [High,Triaged] | 11:39 |
| rvba | jtv: btw, did you resolve your uvtools issues? | 11:52 |
| jtv | Yes! | 11:56 |
| jtv | It's a libvirt problem — you now need to ensure that the "default" network is started. | 11:56 |
| rvba | Is this something that will be done by uvtool or does this mean we need to update maas-test? | 11:57 |
| === CyberJacob is now known as CyberJacob|Away | ||
| rvba | jtv: testing a Saucy MAAS server now… | 13:59 |
| jtv | rvba: keep an eye on the console. | 14:01 |
| jtv | The traceback happens before the JSON data dump. | 14:02 |
| rvba | jtv: I don't have a monitor hooked up to the node. | 14:02 |
| rvba | jtv: but the test should fail correct? | 14:02 |
| jtv | I don't know. | 14:03 |
| jtv | I had test failures, but those might be for different reasons. | 14:03 |
| jtv | It's quite possible that the error gets ignored and things work more-or-less-OK. | 14:03 |
| rvba | That would explain the success testing stuff in the lab. | 14:04 |
| jtv | Don't we have any validation of Node.system_id anywhere? I'm not seeing it in the model or the form. | 14:04 |
| jtv | Sometimes I guess having really really slow hardware has its advantages. | 14:04 |
| rvba | Don't know if we validate system ids anywhere. | 14:05 |
| jtv | Anyway, I fixed the bug (probably mostly on intuition) for trunk in Burlingame, because the python 3 stuff looked more urgent then. | 14:05 |
| === frankban_ is now known as frankban | ||
| jtv | rvba: here's a stab at the form approach to the API call → https://code.launchpad.net/~jtv/maas/validate-network-nodes-filter/+merge/205972 | 14:41 |
| rvba | jtv: cool, I'll have a look in a moment. | 14:44 |
| jtv | Thanks. | 14:44 |
| === cmagina_ is now known as cmagina | ||
| === freeflying is now known as freeflying_away | ||
| tomixxx4 | hi, does anyone have a download link to a boot images which can be used from the nodes? | 16:13 |
| tomixxx4 | or can i simply install "ubuntu server" on the nodes too? | 16:14 |
| jtv | tomixxx4: the cluster controller imports those by itself. | 16:16 |
| jtv | There's a button in the UI. | 16:16 |
| tomixxx4 | jtv: ok but assume the network interface is "unamanged", how can the nodes get the images then? | 16:16 |
| jtv | The DHCP server has to be told to make the nodes netboot from the cluster controller. | 16:17 |
| jtv | That's the custom DHCP server configuration I mentioned in the past. | 16:17 |
| tomixxx4 | jtv: maas-dhcp-server does not work in my case | 16:17 |
| jtv | What's the problem? | 16:17 |
| tomixxx4 | jtv: when i boot the nodes, the nodes print out "TFTP prefix: unable to loacte configuraiton file" | 16:18 |
| tomixxx4 | jtv: so, i want to boot from a usb-stick. the question is, if i can simply install the same ubuntu server image than used for the maas-server | 16:19 |
| * jtv greps source | 16:19 | |
| jtv | Got me there. I don't know! | 16:20 |
| jtv | But netbooting is pretty central. | 16:20 |
| tomixxx4 | jtv: ok, the dashboard mention sth about "avahi boot" | 16:20 |
| jtv | I never looked into that... maybe rvba would know. | 16:21 |
| tomixxx4 | jtv: i have tried many things to get this dhcp - functionality to run, but i have still made no real progress. i have re-installed maas a few minutes ago ;) | 16:21 |
| tomixxx4 | my last hope is to ignore dhcp-server-funktionality, connect the nodes directly to the external-network (so that they have i-net-access) and voila. but first i need a boot image | 16:22 |
| jtv | No chance of running the built-in DHCP server? | 16:23 |
| tomixxx4 | no, it seems not. :( i never got thhis error before (TFTP prefix...) but since i tried to NAT in order to connect the nodes to the i-net, it does not work anymore. | 16:24 |
| tomixxx4 | even if i clear NAT-tables now, i get this error message | 16:25 |
| tomixxx4 | I have also tried to reboot ubuntu server and so one. | 16:25 |
| tomixxx4 | iptables after reboot was empty but the nodes were not able to download the pxe images... | 16:26 |
| tomixxx4 | jtv: the original problem was: if the nodes have no access to internet, juju does not work | 16:27 |
| jtv | The error does seem to mean that the node is trying to download its boot image over TFTP, which is good. | 16:28 |
| tomixxx4 | jtv: juju does not work, because i have to deploy multiple lxc containers on a single node and the lxc-container creation failed because of no internet | 16:28 |
| jtv | Argh. | 16:28 |
| tomixxx4 | yeah... | 16:28 |
| jtv | No way of giving the nodes access to the internet? | 16:28 |
| jtv | The TFTP connection should be straight to the cluster controller, so within the network you control — that part shouldn't require internet access, as long as the cluster has imported the images. | 16:29 |
| tomixxx4 | the idea was maas-dhcp-server + NAT, as the dhcp-server has two interfaces, one connected to the outside-internetworking-network | 16:29 |
| jtv | (By the way, if you want to look at the images, they're on the cluster controller under /var/lib/maas) | 16:29 |
| tomixxx4 | jtv: yeah, the images are there | 16:29 |
| tomixxx4 | resp. the were there | 16:30 |
| jtv | You should be OK running the DHCP server on the gateway — just a matter of making sure you configure the right interface. | 16:30 |
| jtv | If the interface is already configured when you set up MAAS, MAAS will pick it up automatically with its details, and you can just set up DHCP in the MAAS UI. | 16:31 |
| tomixxx4 | jtv: i got an bash script from gmb for the NAT thing so this should not be the problem. | 16:31 |
| tomixxx4 | jtv: interface-configuration looks as follow: http://pastebin.ubuntu.com/6920783 and bash script for NAT: http://pastebin.ubuntu.com/6920787 | 16:35 |
| tomixxx4 | whereas eth0 connects the server to the nodes and eht1 connects the server to the i-net | 16:36 |
| jtv | Right. | 16:36 |
| jtv | That part looks good. | 16:37 |
| jtv | A bit too tired now to understand the shell script... | 16:37 |
| jtv | (Timezones) | 16:37 |
| tomixxx4 | jtv: i have already attache dns-nameserver to the eth0, because otherwise there is a lookup problem later on in juju | 16:37 |
| tomixxx4 | jtv: 11:37 p.m. ? ^^ | 16:39 |
| jtv | Just about! | 16:39 |
| jtv | So... it sounds as if you're successfully doing NAT and DNS? | 16:39 |
| tomixxx4 | dont know, maybe this tftp-exception is caused by misconfiguration | 16:40 |
| tomixxx4 | jtv: at least, dhcp seems to work because the nodes get their ips and so one | 16:41 |
| tomixxx4 | jtv: but they cannot downlaod the boot images... | 16:41 |
| jtv | Where exactly was that TFTP error message, by the way? On the node's console? | 16:42 |
| tomixxx4 | yep, on the nodes console | 16:42 |
| jtv | Then something you might try is force the tftp server to log to a file. It produces lots and lots of debug output, really horrible, but I find it can help debug these problems. | 16:43 |
| jtv | Unfortunately I'm about to fall asleep and don't recall where that can be done. | 16:44 |
| jtv | We direct the output to /dev/null because there's just too much. | 16:44 |
| jtv | But if nothing else, it would tell you whether the TFTP server got the node's request at all. | 16:44 |
| tomixxx4 | kk, i will try this but first i will try to do sth witouth dhcp | 16:45 |
| jtv | I think it'll be harder! | 16:46 |
| tomixxx4 | what is the option "use the fast installer" ? | 16:47 |
| tomixxx4 | in the node menu | 16:47 |
| jtv | All MAAS can really tell the node is: "next time you wake up, netboot. Now, reboot." | 16:47 |
| jtv | The fast installer has some kind of ready-made image. | 16:47 |
| jtv | Doesn't always apply, but when it does, it installs the nodes faster. | 16:47 |
| tomixxx4 | jtv: and i do not need dhcp for this? | 16:47 |
| jtv | You probably still do, because DHCP is part of the netbooting process. | 16:48 |
| jtv | (DHCP is really an extended version of BOOTP as I understand it) | 16:48 |
| tomixxx4 | how does the maas-server know, when the node is "ready" ? | 16:55 |
| tomixxx4 | j4i: installing "ubuntu server 12.04.03" as node boot image does not work | 17:04 |
| === bjf is now known as [home-bjf] | ||
| jtv | tomixxx4: a node becomes Ready when it completes the commissioning phase. It reports that to the region controller through a web API. | 17:16 |
| tomixxx4 | k | 17:16 |
| jtv | At that point, it is available for users to allocate. | 17:16 |
| jtv | When a user has the node allocated, it will be in Allocated state. | 17:16 |
| tomixxx4 | i know, i had already nodes allocated :D | 17:18 |
| tomixxx4 | i was so far | 17:18 |
| tomixxx4 | question just arrised because i have tried to install boot image from usb-stick but nodes did not become "ready" after installation | 17:19 |
| tomixxx4 | DOES maas dashbourd confuse ethernet-names? | 17:21 |
| jtv | I don't understand that last question... | 17:23 |
| jtv | You can't just boot from a USB stick completely; MAAS boots the node several times, with different images and options. | 17:23 |
| tomixxx4 | kk | 17:24 |
| jtv | The images are (1) an ephemeral ("commissioning") image for enlistment & commissioning; and (2) an actual installer for deployment once you allocate. | 17:24 |
| tomixxx4 | how many times boot maas, because i always have to manually POWER ON my nodes | 17:24 |
| jtv | Oops, there's my message to stop working and come home. :) | 17:24 |
| jtv | Once for enlistment, once for commissioning, and once for deployment. | 17:24 |
| tomixxx4 | kk | 17:25 |
| jtv | Once it's allocated to you, it's yours to reboot as often as you like. :) | 17:25 |
| jtv | nn! | 17:25 |
| tomixxx4 | gn8 | 17:27 |
| === chuck_ is now known as zul | ||
| === CyberJacob|Away is now known as CyberJacob | ||
| === zz_mwhudson is now known as mwhudson | ||
| === CyberJacob is now known as CyberJacob|Away | ||
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!