Baribal | Hi. IS there a way to set up an already installed machine as a node without having to reinstall? | 01:46 |
---|---|---|
Baribal | Also, is it possible to use a machine both as region-controller and a node? 'Cause, why waste a good machine? (Especially if you only have one...) | 01:48 |
bigjools | Baribal: no that's not possible | 02:10 |
bigjools | you can share a machine for region and cluster controller though | 02:10 |
jtv | Small consolation, perhaps. :) | 02:10 |
bigjools | well it's a bit chicken and egg isn't it :) | 02:11 |
Baribal | Depends on what MAAS does exactly... But are you considering to lift that limitation? | 02:15 |
Baribal | Also, what about making an already installed machine a node? | 02:15 |
jtv | As far as I know, that's not under consideration, no. | 02:16 |
bigjools | it's not a limitation that we have any control over | 02:16 |
bigjools | you have to provision nodes before you can use them - that's what maas does | 02:17 |
Baribal | How about running a VM on the host which is used as a node? | 02:18 |
jtv | You mean use the VM as a node? | 02:19 |
lifeless | you can do that, but its not a MAAS task to provision VMs | 02:19 |
bigjools | also tools such as juju assume exclusive access | 02:21 |
Baribal | Inhowfar exclusive? | 02:21 |
jtv | It's assumed that the entire install is basically disposable. | 02:22 |
jtv | When you allocate a machine through juju, it normaly assumes that that machine is only used for the application that juju puts on it. | 02:23 |
Baribal | Just for clarity: That'd be the units juju deploys? | 02:27 |
jtv | Yes... I guess a unit is both an instance of an application on the one hand, and an entire machine on the other. | 02:28 |
Baribal | Wait, so per machine, only one unit gets deployed? | 02:29 |
bigjools | normally yes, but you can override that | 02:31 |
bigjools | but the point I am making is that this is one application that assumes that it has total control of nodes | 02:31 |
Baribal | Okay... How do I override it? 'Cause all I have to run MAAS on is two 8right now one :( ) VM. | 02:33 |
bigjools | are you using juju? | 02:33 |
Baribal | Yes. | 02:34 |
bigjools | I can't remember the config offhand | 02:36 |
bigjools | let me see if I can find it | 02:36 |
bigjools | unless jtv remembers | 02:36 |
jtv | Don't even know what config you're referring to... doesn't juju also need a bootstrap node? | 02:36 |
bigjools | yes but you can make juju dispatch charms to the bootstrap node | 02:37 |
jtv | Ah! Don't know the config for that then. | 02:40 |
bigjools | Baribal: in the maas section in environments.yaml, add this: | 02:41 |
bigjools | placement: local | 02:41 |
bigjools | and charms get dispatched to the bootstrap node | 02:41 |
Baribal | Not exactly what I'd prefer, but definitely what I need right now. Thanks. :) | 02:42 |
Baribal | So... Is installation time the only one I can make a machine a prospective MAAS-node? | 02:43 |
bigjools | yes, maas is designed to provision bare metal from scratch | 02:44 |
jtv | That's what it amounts to... You make it a node first, and installation follows. | 02:44 |
Baribal | Okay... Too bad, I have no control over the machine at installation time. | 02:44 |
bigjools | maas is responsible for installing the OS that the API client requests | 02:44 |
Baribal | So I guess I'm back to using Openstack for the moment. | 02:45 |
bigjools | yeah you don't want MAAS then | 02:45 |
bigjools | maas can do VMs but it's more of a testing feature | 02:45 |
jtv | No creation or destruction — you create them and then basically have maas believe that they're physical machines. | 02:46 |
Baribal | I'll try again when I'm building my cluster... But if I understand correctly, my choices are to deploy all units to the bootstrap node or I need one machine per unit? Not quite what I expected. | 02:46 |
bigjools | there may be more to it, but you should ask on the juju channel about that | 02:47 |
Baribal | Will do. | 02:47 |
bigjools | I believe there's a plan | 02:47 |
Baribal | "Going away and buggering the OpenStack people"? :) | 02:48 |
jtv | "bugging" | 02:50 |
jtv | Because "buggering" means something completely different. :) | 02:50 |
bigjools | well, maybe they ... | 02:51 |
Baribal | Oops... | 02:51 |
* jtv laughs | 02:51 | |
lifeless | Baribal: being one of those openstack people... | 02:59 |
lifeless | Baribal: I'd rather you nag us than bugger us. | 03:00 |
Baribal | Don't worry, I'm great at nagging. :) | 03:00 |
Baribal | But, right now, I'll go sleep, I think. I've tried the whole day to get my cloud running... | 03:01 |
Baribal | lifeless, just one quick question: Is Essex still the current OpenStack release? | 03:03 |
Baribal | 'Cause I'm hunting for setup tutorials. | 03:03 |
lifeless | folsom | 03:05 |
lifeless | grizzly coming up | 03:05 |
Baribal | Thanks; so http://docs.openstack.org/folsom/basic-install/content/ is probably what I'll want to read. Except if there's a good shortcut? | 03:07 |
lifeless | no shortcuts :) | 03:12 |
lifeless | not if you want to understand what you're turning on | 03:12 |
Baribal | I do, but not right now. | 03:14 |
lifeless | well for that case you can just install the openstack packages on both nodes | 03:18 |
lifeless | one controller + compute, one compute | 03:19 |
lifeless | I recommend reading a bit though, its useful to know | 03:19 |
lifeless | ... | 03:19 |
=== jtv1 is now known as jtv | ||
bitwiz | has this question been resolved? https://answers.launchpad.net/ubuntu/+source/ubiquity/+question/207347 I am having the same results on my new install of 12.10 maas | 15:16 |
=== matsubara is now known as matsubara-lunch | ||
=== matsubara-lunch is now known as matsubara | ||
roaksoax | allenap: around by any chance? | 18:29 |
roaksoax | matsubara: around? | 20:17 |
matsubara | roaksoax, yes | 20:40 |
roaksoax | matsubara: had a question for you which you miht be able to help me with | 20:40 |
matsubara | shoot | 20:41 |
roaksoax | matsubara: in a multinode maas cluster, who holds the pxe images, the region or the cluster? | 20:41 |
roaksoax | s/holds/stores | 20:41 |
matsubara | I think the region is the first one to get but then when the cluster requests the images from the region, it then stores them itself | 20:42 |
roaksoax | matsubara: ack! thanks for the info | 20:43 |
matsubara | this is my understanding of it, not sure if it changed since the EOY break. didn't touch the cluster controller stuff | 20:43 |
matsubara | I'll know for sure next week when I resume working on the jenkins job to test the cluster controller | 20:43 |
roaksoax | cool | 20:46 |
roaksoax | thanks for the info though | 20:46 |
matsubara | np | 20:47 |
=== matsubara is now known as matsubara-afk | ||
=== cmagina is now known as cmagina_away | ||
=== cmagina_away is now known as cmagina | ||
=== cmagina is now known as cmagina_away | ||
=== cmagina_away is now known as cmagina | ||
=== cmagina is now known as cmagina_away | ||
dpb_ | Hi -- if a node is claimed by a user or allocated. How do I release it, forcefully is fine. | 22:06 |
=== cmagina_away is now known as cmagina | ||
dpb_ | nm, figured it out maas cli -- node release | 22:22 |
=== cmagina is now known as cmagina_away | ||
=== cmagina_away is now known as cmagina | ||
=== cmagina is now known as cmagina_away |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!