maas-devel team mailing list archive

["Francis J. Lacoste" <francis.lacoste@xxxxxxxxxx>]

Hi Michael,

On 12-05-23 09:52 AM, Michael Hughes wrote:
> Hi Devels. I'm continuing to test MAAS under 12.04 and keep running
> into show-stoppers. Before I explain, I want to understand one thing:
> is MAAS working for anybody following the standard installation steps?
> I've been going over them time and time again and I have never been
> able to get the 'Wordpress backed by MySQL' to install smoothly.
> Perhaps this is still in beta?
> 

Yes, MAAS is still very much "an early version" of the product. But yes,
the existing instructions worked at least in our QA lab :-)

We have not set it up with VirtualBox, but it should be possible to make
it to work.

> On with my current hangup: I'm running six nodes plus the MAASserver
> using Virtualbox and they are all connected in a private network
> ("intnet" is the vbox vernacular.) The MAASserver has a second NIC
> that connects to the Internet. All nodes were manually configured to
> use virsh for their power, although I have never seen a node Wake On
> LAN. Perhaps this is a weakness of Virtualbox? For this reason I've
> been starting the nodes manually when I think they're being called
> upon and it's worked okay as far as getting them all in the "READY"
> state on the MAAS dashboard.

That's probably a configuration problem in Cobbler. I don't think
VirtualBox nodes can support Wake-on-lan (that really only works with
real hardware... and even there with some quirks depending on the chipset).

You probably need to edit /etc/cobbler/power/power_virsh.template to
make it work with VirtualBox. The default virsh driver used is qemu, you
probably want to change that to 'vbox'. So I'd suggest applying the
following patch:

--- power_virsh.template.orig	2012-05-23 10:16:35.000000000 -0400
+++ power_virsh.template	2012-05-23 10:18:56.000000000 -0400
@@ -8,11 +8,11 @@
 ## Build connection URI
 ##
driver[+transport]://[username@][hostname][:port]/[path][?extraparameters]

-## Determine requested driver to use (defaults to 'qemu')
+## Determine requested driver to use (defaults to 'vbox')
 #if $power_address and $power_address.count(':') > 0
     #set (driver, power_address) = $power_address.split('://', 1)
 #else
-    #set driver = "qemu"
+    #set driver = "vbox"
 #end if

 ## Was a username requested (defaults to '')?


END

This wasn't tested! It's possible that the URI created by this template
needs further tweaks to work with the VirtualBox hypervisor.

> 
> But when I bootstrap juju and install Wordpress and MySQL, even
> starting the nodes manually doesn't seem to 'activate' them. It
> appears to me that it wants nodes 1, 2, and 4 to be running, so I
> start them and they boot to a login screen. In the MAAS dashboard they
> are now "Deployed" and "Allocated to Mike," however only node #1 seems
> to be 'seen' by juju :
> ----
> $ juju status
> 2012-05-23 08:34:51,354 INFO Connecting to environment...
> 2012-05-23 08:34:55,380 INFO Connected to environment.
> machines:
>   0:
>     agent-state: running
>     dns-name: node-1-.local
>     instance-id: /MAAS/api/1.0/nodes/node-ddd48178-a444-11e1-be23-08002702f035/
>     instance-state: unknown
>   1:
>     agent-state: not-started
>     dns-name: node-2-.local
>     instance-id: /MAAS/api/1.0/nodes/node-d801263e-a444-11e1-b5d7-08002702f035/
>     instance-state: unknown
>   2:
>     agent-state: not-started
>     dns-name: node-4-.local
>     instance-id: /MAAS/api/1.0/nodes/node-5776a196-a445-11e1-be23-08002702f035/
>     instance-state: unknown
> services:
>   mysql:
>     charm: cs:precise/mysql-2
>     relations:
>       db:
>       - wordpress
>     units:
>       mysql/0:
>         agent-state: pending
>         machine: 1
>         public-address: null
>   wordpress:
>     charm: cs:precise/wordpress-1
>     relations:
>       db:
>       - mysql
>     units:
>       wordpress/0:
>         agent-state: pending
>         machine: 2
>         public-address: null
> 2012-05-23 08:34:58,364 INFO 'status' command finished successfully
> ----
> 
> When I initially ran 'juju status' I was prompted to accept the ssh
> fingerprint of the first node. I never saw a prompt for accepting the
> second and fourth nodes. Could this be why it fails to contact them?
> 


No, I've only ever had to accept the bootstrap node fingerprints. When
I've seen similar problems in the past, it was because the juju agents
were not able to contact the MAAS provider end point. You need to make
sure that the maas-server: configuration in your juju environments.yaml
is reachable by the MAAS node.

Other thing to look at are connecting to the node using juju ssh
$machine_number and looking at the juju agent logs (/var/log/juju/ iirc)

Cheers


-- 
Francis J. Lacoste
francis.lacoste@xxxxxxxxxx

Attachment: signature.asc
Description: OpenPGP digital signature