=== shang_ is now known as shang | ||
rvba | rbasak: I'm enlisting ARM nodes all right, but when I try to deploy something with juju on them, juju is never able to connect to them. | 12:47 |
---|---|---|
rbasak | rvba: do you have serial output on the ARM nodes? You should be able to get this with ipmitool or freeipmi | 12:48 |
rvba | rbasak: I'm using the package in the dailyppa on precise. | 12:48 |
rvba | rbasak: this is happening in the lab | 12:48 |
smoser | rvba, ping. | 15:58 |
smoser | i'm playing aroudn with some stuff, and i'd liek a general bucke where i could put some information on start | 15:58 |
smoser | is there anything like that ? | 15:58 |
smoser | right now i can post to the api key value pairs with the start reqeust, but only 'user_data' and 'series' are paid attention to. | 15:59 |
smoser | i'd like for other stuff i post to get somehow attached ot the node (probably removed on 'release') | 16:00 |
rvba | smoser: hey | 16:03 |
smoser | the end goal would be for me to be able to get at those key/value pairs inside the templates | 16:04 |
rvba | smoser: if you have a look at the api code for start (http://paste.ubuntu.com/1414975/), we pass the 'user_data' to the start_nodes method and the 'series' is attached to the node. I don't know what you're trying to achieve really but looks like you might want to create new data attached to a node. | 16:04 |
smoser | i'd like a general bucket | 16:05 |
smoser | rather than "user_keys" or something. that everything got shoved into | 16:05 |
smoser | in my installer templates, i can then read stuff from there | 16:05 |
smoser | the specific motivation would be to pass something like: | 16:06 |
smoser | installer=fast_pass | 16:06 |
smoser | or | 16:06 |
smoser | installer=d-i | 16:06 |
smoser | that specific thing might make sense to have maas know about, but there are surely lots of other things that maas can just shuffle through. | 16:06 |
rvba | One solution would be to create a new JSONField on the node with a meaningful name (:)) and use that to store your stuff. | 16:07 |
rvba | This field would contain a json object that you could exploit in the installed templates. | 16:08 |
rvba | (you would have to make it available to the templates of course) | 16:08 |
rvba | smoser: maybe a JSONField is overkill if you just want to have one string to store installer parameters that can be stored as one line of text. | 16:11 |
smoser | json is just text. | 16:11 |
smoser | binary blob is most "I dont care" | 16:11 |
smoser | second is text that "I dont care" | 16:12 |
smoser | and if i'm suggesting that maas not care, then i'd suggest it really not care. | 16:12 |
smoser | (ie, not insist on json) | 16:12 |
smoser | rvba, i dont *hvae* to make it avaliable in the templates explicitly. as I can execute arbitrary python code there. | 16:32 |
=== matsubara is now known as matsubara-lunch | ||
melmoth | is there a way in Maas to have several different preseed file used when a machine is being deployed ? | 16:39 |
melmoth | like, on some box creating a new partition on the drive (so we can use them to deploy a swift storage or nova-volumes charms) ? | 16:40 |
melmoth | (talking about maas on 12.04 btw) | 16:40 |
smoser | melmoth, why would you not do that via user-data ? | 16:41 |
smoser | what would be the value of doing it in the installer ? | 16:41 |
smoser | you *can* accomplish this (even in 12.04 most likely) by having the template execute whatever python code you want and call out to another service even more data). | 16:41 |
melmoth | i have no idea about the user-data thing. | 16:41 |
smoser | ? | 16:41 |
smoser | when you deploy a node in maas, it takes user-data ( just like on ec2/openstack ) | 16:42 |
melmoth | i understand user-data for vm, but for bare metal, once lcoud init runs , the machine is alreadt installed, so you cannot shrink / can you ? | 16:42 |
smoser | and cloud-init acts on that user-data. | 16:42 |
smoser | well, you can shrink filesystems, yes. | 16:42 |
smoser | but generally having the installer not take all for / if you wanted to use less makes more sense here. | 16:42 |
melmoth | but only for some machine. | 16:43 |
Aram | hi, I need a little hint. My maas controller has two ethernet interfaces. I want to use maas-dhcp and maas-dns. how do I tell maas-dhcp only to use eth1? I don't need a full answer, just a manual page name. | 17:25 |
=== matsubara-lunch is now known as matsubara | ||
Aram | hmm... /etc/dhcp/dhcpd.conf does not get created by maas-dhcp. I was under the impression maas generates this files, do I need to create if by hand? | 18:51 |
Aram | I solved the eth1 problem, btw | 18:51 |
melmoth | any idea why bootstraping fail ? http://pastebin.com/gF72WSzq summary: The requested URL /MAAS/api/1.0/tags/ was not found on this server. | 18:52 |
Aram | I guess I could use maas-provision generate-dhcp-config to generate it, but what omapi key to use? | 18:54 |
matsubara | Aram, if you configure the controller through the web UI you'll be able to choose which interface is running the dhcp server | 19:05 |
matsubara | Aram, or you can use the maas-cli | 19:06 |
Aram | matsubara: thanks. found it. I feel stupid now. I didn't think of using the GUI at all. | 19:06 |
Aram | dhcpd seems to work now, but it can't find the tftp server | 19:07 |
Aram | do I need to manually enable that as well? | 19:07 |
matsubara | Aram, nope, the dhcpd.conf file is in /etc/maas/ can you paste that? the next_server parameter there is where the nodes will look for the tftp server | 19:08 |
Aram | ah, yes, the next server is wrong. it's on a different network interface | 19:08 |
matsubara | so, once you install maas, if you need to run dpkg-reconfigure maas to set the proper address for the pxe interface | 19:08 |
Aram | hmm... dpkg-reconfigure mass doesn't do anything | 19:09 |
Aram | maas | 19:09 |
Aram | exists with status 0 | 19:09 |
Aram | no output | 19:09 |
Aram | same with -p low | 19:10 |
Aram | ok, it was maas-region-controller, not maas, but dhcpd.conf is still wrong. | 19:17 |
Aram | well, when running dpkg-reconfigure maas-region-controller, I did get an error. | 19:18 |
Aram | this is the error | 19:19 |
Aram | http://paste.ubuntu.com/1415395/ | 19:19 |
Aram | don't know if the error is fatal or not, but it does rememer what I've set. | 19:20 |
Aram | but dhcpd.conf has the old name | 19:20 |
Aram | also the dns is wrong in dhcpd.conf, not only next-server | 19:21 |
Aram | this is my dhcpd.conf: http://paste.ubuntu.com/1415399/ | 19:22 |
Aram | matsubara: any idea about that error ^ ^ | 19:31 |
matsubara | Aram, you can change that file manually and run sudo service maas-dhcp restart | 19:33 |
matsubara | and check if the node will boot properly | 19:33 |
Aram | matsubara: I did that, am I correct in assuming my changed won't be overwritten? anyway, now it finds the TFTP server but it stops at 'Trying to load: pxelinux.cfg/default' | 20:21 |
Aram | s/changed/changes/ | 20:21 |
Aram | Unable to locate configuration file | 20:23 |
matsubara | Aram, did you run maas-import-pxe-files? | 20:24 |
Aram | yes. | 20:25 |
Aram | but in /var/lib/maas/tftp/amd64/generic there's only precise | 20:25 |
Aram | and I set quantal in the web ui afterwards | 20:25 |
matsubara | maybe it failed to download the quantal images? | 20:30 |
Aram | it only tries to download precise | 20:31 |
Aram | ran it again | 20:32 |
Aram | precise node seems to boot via pxe, whohoo | 20:32 |
Aram | but I'd like quantas as well | 20:32 |
Aram | quantal | 20:32 |
Aram | hmm, the precise node didn't auto-enlist | 20:34 |
Aram | yep, nodes boot alright but with no auto enlistment | 20:39 |
Aram | am I doing something wrong? | 20:40 |
=== lifeless_ is now known as lifeless | ||
Aram | yes they do | 20:44 |
Aram | it just took some time | 20:44 |
Aram | I see two nodes | 20:44 |
Aram | so now only the quantal problem remains | 20:44 |
Aram | why does it take so long for new nodes to appear in the maas interface? | 22:01 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!