Part of a series: Install · Configure · DHCP · Commission · Deploy · jq · SSH · More jq

Now that we have running machines, what if we set ourselves up to SSH into one? This also unlocks scp for file transfers — a practical starting point for provisioning.

Creating a VM host

Start by creating a vm-host. A useful lesson in CLI discovery:

maas admin vm-host create

Wrong subcommand — create isn’t valid for vm-host (singular). The collection form is what adds new entries:

maas admin vm-hosts create

Better, but we need to specify a type:

{"type": ["This field is required."]}

For an LXD vm-host:

maas admin vm-hosts create type=lxd

Still needs one more thing:

{"power_address": ["This field is required."]}

The power_address for LXD is https://<gateway-ip>:8443. The 8443 is the default port from lxd init. In my case:

maas admin vm-hosts create type=lxd power_address=https://10.38.31.1:8443

Success. The key line in the JSON response:

"resource_uri": "/MAAS/api/2.0/pods/7/"

The vm-host ID is 7 — needed for the next step.

A note on terminology: vm-hosts were originally called “pods” (hence the URI), then “KVM hosts,” then “VM hosts” as MAAS expanded beyond libvirt to support LXD. The MAAS CLI reflects all three eras in various places.

Composing a machine

With a VM host, we need a virtual machine. The compose command is on vm-host (singular), not machines:

maas admin vm-host compose --help

This reveals a rich set of options — cores, memory, storage, interfaces, NUMA pinning, and more. For now, all we need is the vm-host ID:

maas admin vm-host compose 7

Success:

{
    "system_id": "xttpfx",
    "resource_uri": "/MAAS/api/2.0/machines/xttpfx/"
}

Check its status with our jq table command:

HOSTNAME    SYSID   POWER  STATUS   OWNER  TAGS     POOL     VLAN      FABRIC    SUBNET
--------    -----   -----  ------   -----  ----     ----     ----      ------    ------
native-cub  xttpfx  on     Testing  admin  virtual  default  untagged  fabric-1  10.38.31.0/24

By the time I finished typing the command, commissioning had nearly completed. Running it again:

native-cub  xttpfx  off    Ready   -      virtual  default  untagged  fabric-1  10.38.31.0/24

Commissioning with SSH enabled

The machine is Ready but powered off, and SSH keys weren’t loaded during commissioning. Fix that by recommissioning with SSH enabled — and leaving the machine powered on afterward:

maas admin machine commission xttpfx enable_ssh=1

After it finishes:

native-cub  xttpfx  on     Ready   -      virtual  default  untagged  fabric-1  10.38.31.0/24

Ready and powered on.

Finding the IP address

Several ways to get the IP. The easiest is lxc list:

lxc list

Output:

+------------+---------+---------------------+-----------------------------------------------+-----------------+-----------+
|    NAME    |  STATE  |        IPV4         |                     IPV6                      |      TYPE       | SNAPSHOTS |
+------------+---------+---------------------+-----------------------------------------------+-----------------+-----------+
| first-one  | RUNNING | 10.38.31.193 (eth0) |                                               | CONTAINER       | 0         |
+------------+---------+---------------------+-----------------------------------------------+-----------------+-----------+
| native-cub | RUNNING | 10.38.31.202 (eth0) | fd42:fd4c:6ab9:19bc:216:3eff:fe9e:bc7b (eth0) | VIRTUAL-MACHINE | 0         |
+------------+---------+---------------------+-----------------------------------------------+-----------------+-----------+

Our machine is 10.38.31.202. Note the distinction: CONTAINER vs VIRTUAL-MACHINE. Both show up here.

Logging in via SSH

ssh ubuntu@10.38.31.202

First-login fingerprint prompt — say yes:

Warning: Permanently added '10.38.31.202' (ECDSA) to the list of known hosts.
Welcome to Ubuntu 20.04.1 LTS (GNU/Linux 5.4.0-64-generic x86_64)
...
ubuntu@native-cub:~$

Always use the ubuntu user for MAAS-deployed machines.

Copying files with SCP

The home directory on the new machine is empty. Exit, create a test file locally, and copy it over:

ubuntu@native-cub:~$ exit
stormrider@wintermute:~$ touch zork
stormrider@wintermute:~$ scp ./zork ubuntu@10.38.31.202:
zork     100%    0     0.0KB/s   00:00

Log back in and verify:

ubuntu@native-cub:~$ ls
zork

Copying files to a deployed machine

Files copied to a commissioned machine don’t survive deployment — the machine gets wiped and reloaded. Acquire and deploy first:

maas admin machines allocate system_id=xttpfx
maas admin machine deploy xttpfx

When it reaches Deployed state, the SSH key has changed. You’ll see:

@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@
@    WARNING: REMOTE HOST IDENTIFICATION HAS CHANGED!     @
@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@

Expected — the machine was reloaded. Just do what it says:

ssh-keygen -f "/home/stormrider/.ssh/known_hosts" -R "10.38.31.2"
ssh ubuntu@10.38.31.2

You’re in.

Running a script on a deployed machine

Install a few packages on the deployed machine:

sudo apt install fortune-mod ddate cowsay

Then exit, create a local script called motd.sh:

#!/bin/bash
ddate > /tmp/foo
echo '   ' >> /tmp/foo
fortune -s >> /tmp/foo
cat /tmp/foo | cowsay

Make it executable and copy it over:

chmod 777 motd.sh
scp ./motd.sh ubuntu@10.38.31.2:

Log in and run it:

ubuntu@native-cub:~$ ./motd.sh

Output:

 ________________________________________
/ Today is Boomtime, the 32nd day of     \
| Chaos in the YOLD 3187                 |
|                                        |
| Water, taken in moderation cannot hurt |
| anybody.                               |
|                                        |
\ -- Mark Twain                          /
 ----------------------------------------
        \   ^__^
         \  (oo)\_______
            (__)\       )\/\
                ||----w |
                ||     ||

This is one method MAAS users rely on to configure deployed machines — scripts scped over and kicked off. There are more sophisticated approaches, but this demonstrates the basic workflow.

Next up: more jq tricks.