| AskUbuntu | Enlisted MAAS nodes will not PXE boot for commissioning | http://askubuntu.com/q/291472 | 12:35 |
|---|---|---|
| === kentb-out is now known as kentb | ||
| AskUbuntu | Can jitsu deploy multiple service units like juju's "add-unit" command? | http://askubuntu.com/q/291579 | 15:59 |
| Dr{Who} | http://askubuntu.com/questions/263849/info-bootstrapping-environment-error-no-matching-node-is-available so I need more free nodes to even be able to get juju working? I am getting this error and dont have any nodes in ready state. | 19:07 |
| roaksoax | Dr{Who}: what does /var/log/maas/maas.log say? | 19:09 |
| roaksoax | Dr{Who}: can you also share with me a screenshot of the info of a node? | 19:09 |
| Dr{Who} | http://pastebin.com/BceDkAic | 19:13 |
| Dr{Who} | I have a few nodes setup but all are marked Allocated although only 1 is realy alive the other 2 are dead just not easy to delete them | 19:13 |
| Dr{Who} | you want a screen shot of the "Edit Node" page? | 19:14 |
| roaksoax | Dr{Who}: yeah, the node that you want to use but juju says no node available | 19:20 |
| Dr{Who} | i did not want to use any nodes yet. Just want to have it setup so I can move to the next step of using Juju and MAAS to build an openstack cluster. So I need nodes free first correct? | 19:23 |
| Dr{Who} | this error came from my MAAS box itself where I am trying to setup juju | 19:24 |
| Dr{Who} | ok out for lunch.. bbl. | 19:24 |
| roaksoax | Dr{Who}: right, so the nodes *must* be in 'Ready' state in order for you to use juju to deploy | 19:25 |
| roaksoax | if they are 'Allocated to user XYXZ' that means that the nodes are not available | 19:25 |
| roaksoax | for juju to use | 19:25 |
| Dr{Who} | ok so JuJu will actually take over a system and install stuff on it. Juju needs its own dedicated system? | 20:21 |
| Dr{Who} | ok so next question. I guess I have to delete the nodes I have setup so far and get them to the "ready" state. Any wiki page on doing this via the cli? I do not see a way through the UI | 20:22 |
| roaksoax | Dr{Who}: yeah there's a way | 20:31 |
| roaksoax | Dr{Who}: give me a sec | 20:31 |
| roaksoax | Dr{Who}: look into node > release http://maas.ubuntu.com/docs/maascli.html | 20:32 |
| roaksoax | Dr{Who}: so when you juju bootstrap, it will select a 'Ready' system, mark it as allocated, and deploy ubuntu on it | 20:32 |
| Dr{Who} | it will create just one node it looks like. I presume I would call thsi juju in maas node name. | 20:33 |
| roaksoax | juju will use *any* of the nodes in the pool which are in 'Ready' state | 20:34 |
| Dr{Who} | k I am going to provision a vmware virtual for this as I dont have much metal for testing. | 20:36 |
| Dr{Who} | ok i just created a new node called it 'juju' and it should start to build it now. | 20:36 |
| Dr{Who} | i did 13.04 same as my maas box. | 20:37 |
| Dr{Who} | I want to save what metal I have for openstack compute nodes etc. so if its a management box like MAAS or this new juju server they can be vms imho bare metal is a waste for such a small task. | 20:38 |
| roaksoax | Dr{Who}: yeah they can be in VM's | 20:39 |
| Dr{Who} | how can i delete an allcoated node? I get an error. Dont see any command in the api to "unallocate" | 20:49 |
| Dr{Who} | maas release from cli found it. I think. | 20:58 |
| Dr{Who} | yep all good.. Added to my wiki | 20:59 |
| roaksoax | Dr{Who}: cool, so now that the node is released and back to 'ready', you can delete it | 21:01 |
| Dr{Who} | yep. my garbage is now gone from testing :) yay. | 21:02 |
| Dr{Who} | ok I have a node called juju that is marked ready. the vm itself turned off. I presume its donw? | 21:03 |
| Dr{Who} | done. | 21:03 |
| Dr{Who} | dont have ipmi just wake on lan and not sure its right so I have been restarting as needed manually. | 21:03 |
| roaksoax | Dr{Who}: yeah you need to manually turn it on | 21:04 |
| roaksoax | Dr{Who}: maas doesn't support vmware VM management | 21:05 |
| Dr{Who} | ok so should I try the juju bootstrap now while the vm is off and maas has it as ready? or ... | 21:05 |
| Dr{Who} | or in some specific order power it up manually. | 21:05 |
| Dr{Who} | ya just boot strapped and it allocated that node. starting it up now | 21:07 |
| roaksoax | Dr{Who}: so if you juju bootstrap and then manually start the it has chosen, then it will be installed with ubuntu and the juju releated stuff | 21:08 |
| Dr{Who} | yep its install ubuntu now. time for a 3 hour break :( i need to get some of the repo's mirrored locally if that is possible with maas? | 21:09 |
| mwhudson | roaksoax: hey, have you thought about fpi for arm systems at all? | 21:10 |
| roaksoax | mwhudson: hey! So... if we have a ephemeral image for amr, then we sdon't think there would be any difference | 21:10 |
| roaksoax | arm* | 21:10 |
| mwhudson | roaksoax: well xinstall has a bunch of stuff about grub, for example | 21:11 |
| mwhudson | which i don't think is used on the arm server i have access too | 21:11 |
| mwhudson | *to | 21:11 |
| roaksoax | mwhudson: I see, then I guess it would not work. You'll need to check with smoser since he wrote the installer | 21:12 |
| smoser | mwhudson, it'd take some work for sure. | 21:17 |
| mwhudson | smoser: would it make sense for xinstall to support this, or would it make more sense to have a different installer for arm (or even highbank, i dunno) and some kind of parameterization? | 21:19 |
| smoser | mwhudson, i'm not opposed to having xinstall figure that out. | 21:24 |
| smoser | longer term, i really want "xinstall" to be much dummer about the stuff its installing. | 21:24 |
| smoser | i'm open to s uggestions on how to support arm or the like | 21:24 |
| smoser | but where i'd like to go is basically 'dd if=diskimage of=blockdevice' | 21:24 |
| smoser | and expect that to result in somethign that boots. | 21:24 |
| mwhudson | i'm so glad to hear you say that :) | 21:25 |
| mwhudson | there might need to be some partition resizing | 21:25 |
| mwhudson | but that can happen on first boot i guess | 21:25 |
| mwhudson | smoser: why does xinstall _not_ do that now? the nature of the images that you are working with? | 21:25 |
| smoser | wmwell, i suppose that is the case, yeah. | 21:26 |
| smoser | mwhudson, the thing that initially moved me away from "image based" to "tar based" was that the filesystems werent necessarily aligned with disk secotro size. | 21:28 |
| mwhudson | ah yeah | 21:28 |
| smoser | but that is arguably a a.) filesystem | 21:28 |
| smoser | problem. | 21:29 |
| smoser | or b.) image problem | 21:29 |
| smoser | can be solved though. | 21:29 |
| Dr{Who} | ok juju is setup and seems to be ready for action. Now openstack. The page https://wiki.ubuntu.com/ServerTeam/OpenStackHA shows I need 23 systems? but then it also talks about Virtual-MAAS so I guess next question is what is the minimum I need to start a test openstack environemnt. | 21:34 |
| Dr{Who} | what would be nice is to understand why 26 servers was mentioned here. What does each system do out of the 26 | 21:36 |
| Dr{Who} | it sure seems like a lot of these process can run on the same metal. https://wiki.ubuntu.com/ServerTeam/OpenStackHA?action=AttachFile&do=view&target=local.yaml Not sure how to adjust the config to do that. still reading | 21:43 |
| AskUbuntu | relation error between nova-compute and nova cloud controller | http://askubuntu.com/q/291736 | 21:47 |
| mwhudson | smoser: ok | 21:50 |
| mwhudson | smoser: i appear to have some highbank images (eg https://snapshots.linaro.org/quantal/pre-built/highbank/301) which i think are just dd-able | 21:51 |
| mwhudson | smoser: do you have in mind where a simpler path in xinstall would go? | 21:54 |
| smoser | mwhudson, if you're looking at the code there, i have no issue with you doing something siplistic at the moment like just figuring out arch and not doing that. | 21:55 |
| smoser | or, having some way that just disables all "mucking" | 21:56 |
| smoser | mwhudson, sorry to be somehwat vague here... | 21:56 |
| mwhudson | heh no worries | 21:56 |
| smoser | i have to think about it more, but at least in informal thoughts, i've had the folowing idea: | 21:56 |
| mwhudson | i'm doing at least four things concurrently | 21:56 |
| smoser | * images shoudl be "full disk" (dd'able) | 21:57 |
| smoser | * installer wil "communicate" with installed image via a partition that it sees of a certain format | 21:57 |
| smoser | this is not entirely different that basically shoving in a "config-drive" | 21:57 |
| smoser | with data like "here is your maas credentials" and "here is your network information" | 21:58 |
| mwhudson | ah yeah | 22:01 |
| mwhudson | lava (which is the other thing i've used that does this) assumes serial console access for that sort of thing | 22:01 |
| Dr{Who} | so just learning the big diff between the different cloud systems. It seems like OpenStack is "not" monolithic a good and a bad thing. It can be massively distributed "good" but can it be combined for smaller clouds. I planned on dedicating 3-4 systems to this cloud test. I current run everything about 100 virtuals on 3 vmware boxes running older xeons and 32g of ram each. | 22:17 |
| Dr{Who} | I am not sure this can be done. Juju does not seem to want to deploy onto a box multiple times so adding ceph onto 3 systems then adding rabbit mq on all 3 mysql on all 3 etc seems out of the scope of most docs I have found. | 22:18 |
| === kentb is now known as kentb-out | ||
| roaksoax | bigjools: around yet? | 22:40 |
| roaksoax | bigjools: http://paste.ubuntu.com/5639917/ | 23:33 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!