Cannot deploy from the cli when "Allow DNS resolution" is set on minimal subnet

[Liam Young]

I'm using snap version/build 3.3.2-13177-g.a73a6e2bd

In my deployment I could not deploy servers from the cli, but using
the web interface was fine.

I tracked the issue down to the call to get_curtin_merged_config
*1. It which in turn includes a call to reduce_routable_address_map which
contains an assert statement which is the source of the error *2.

I eventually found that disabling allow_dns on one of the subnets
attached to guest (and the rack controller) fixed the issue.

I have a raised a separate bug to cover the fact that assertion
errors generally are not being surfaced to the user properly. I think this
bug should cover the fact that a guest cannot be booted from the
CLI in this circumstance.

# maas admin machine deploy kgr4kk distro_series=ubuntu/jammy --debug
400 Bad Request

       Connection: keep-alive
   Content-Length: 26
     Content-Type: text/plain; charset=utf-8
             Date: Mon, 24 Apr 2023 08:59:43 GMT
           Server: nginx/1.18.0 (Ubuntu)
           Status: 400
             Vary: Cookie
  X-Frame-Options: DENY

Failed to render preseed:
#

NOTE: Nothing follows the "Failed to render preseed"

# maas admin subnet update "fd42:5eda:f578:7bba::/64" allow_dns=false
Success.
Machine-readable output follows:
{
    "name": "fd42:5eda:f578:7bba::/64",
    "description": "",
    "vlan": {
        "vid": 0,
        "mtu": 1500,
        "dhcp_on": false,
        "external_dhcp": null,
        "relay_vlan": null,
        "secondary_rack": null,
        "id": 5003,
        "fabric": "fabric-2",
        "fabric_id": 2,
        "primary_rack": null,
        "space": "undefined",
        "name": "untagged",
        "resource_uri": "/MAAS/api/2.0/vlans/5003/"
    },
    "cidr": "fd42:5eda:f578:7bba::/64",
    "rdns_mode": 2,
    "gateway_ip": "",
    "dns_servers": [],
    "allow_dns": false,
    "allow_proxy": true,
    "active_discovery": false,
    "managed": true,
    "disabled_boot_architectures": [],
    "id": 5,
    "space": "undefined",
    "resource_uri": "/MAAS/api/2.0/subnets/5/"
}
# maas admin machine deploy kgr4kk distro_series=ubuntu/jammy --debug
200 OK

Logs from immeditaly after the initial failure:

The log messages are:

==> maas.log <==
2023-04-24T08:45:23.221516+00:00 maas maas.rpc.rackcontrollers: message repeated 2 times: [ [info] Existing rack controller 'maas' running version 3.3.2-13177-g.a73a6e2bd has connected to region 'maas'.]
2023-04-24T08:46:41.122016+00:00 maas maas.api: [info] Request from user admin to acquire machine: microk8s07.maas (kgr4kk)
2023-04-24T08:46:41.140294+00:00 maas maas.node: [info] microk8s07: Status transition from READY to ALLOCATED
2023-04-24T08:46:41.155743+00:00 maas maas.node: [info] microk8s07: allocated to user admin

==> regiond.log <==
2023-04-24 08:46:41 maasserver.preseed: [warn] WARNING: '/snap/maas/27109/etc/maas/preseeds/curtin_userdata' contains deprecated preseed variables. Please remove: main_archive_directory, ports_archive_directory

Note: The symptoms are very similar to https://discourse.maas.io/t/deploy-over-api-fails-to-render-preseed/6134

*1 https://git.launchpad.net/maas/tree/src/maasserver/api/machines.py?h=3.3.2#n878
*2 https://git.launchpad.net/maas/tree/src/maasserver/routablepairs.py?h=3.3.2#n116

[Liam Young (gnuoy)]

fwiw I raised https://bugs.launchpad.net/maas/+bug/2017503 to cover improving the handling of the error messages