We already have a DNS server ... can we still use MAAS?

[djh-quantifind]

Hello,

We have been building systems by PXE boot and USB keys for years. We run our own DNS, NTP &c. Can we get MAAS running if it doesn’t control DNS? We are presently using dnsmasq.

Thanks,
-danny

[lloyd]

I ran a non-MAAS DNS for several years, but it never quite worked for me. I find that the MAAS cli lets me do the DNS changes that I used to do by hand, so I’m okay with that.

Giving MAAS a subdomain to manage might work for you.

[djh-quantifind]

Thank you, lloyd.

We’re setting up a new datacenter. Many machines will be rebuilt, and other machines we’re looking to just move over, with a new name/IP. I’ll experiment today with trying MAAS for the first time for new machine builds, and whether it will be appropriate to our needs for “importing” machines that we do not wish to rebuild.

Do folks typically let MAAS run the whole show and let MAAS manage DNS for non-MAAS hardware, or just set up a MAAS subdomain over to the side? My feeling is if MAAS is going to run a DHCP server then we shuold consider going all in.

[georcon]

Hi,

If it isn’t too late, I stumbled upon a (similar) issue. I am also running a DNS server, on the same machine as maas rack+region.

I use CoreDNS.

It seems that the packaged BIND server will listen on all interfaces on port 53, as long as the port is available.

There is an option within the fabric-x configuration, “Provide DNS on this fabric”. This doesn’t seem to affect whether BIND listens on that interface.

I needed to modify the BIND server included in MAAS to only listen on the cluster-local interface.

There are possibilities in modifying the generated BIND configs (under/var). I did this by modifying named.conf.options (or you could do the same in named.conf, or any of the parsed configuration files).

I added thelisten-on {<MAAS DNS IP>;}; directive, to make (MAAS) BIND listen only on the interface that. It then gets complicated - as a very complicated.

The solution I am currently using:

Stop the MAAS service via:

sudo snap stop maas.supervisor

Start your custom DNS server

restart MAAS

sudo snap start maas.supervisor

[nateybobo]

I gave MAAS a subdomain to be authoritative over for my bare metal hosts.

ie. maas.example.com

I then told MAAS to point to my upstream primary/secondary DNS servers at this setting page:
http://<ip>:5240/MAAS/r/settings/network/dns

Finally, on my primary/secondary DNS, I added forward and reverse zones for that subdomain:

zone "maas.example.com" {
	type forward;
	forwarders {
		<maas-ip>;
		};
	};
zone "<arpa-ip>.in-addr.arpa" {
	type forward;
	forwarders {
		<maas-ip>;
		};
	};

Your mileage my vary, depending on your prod DNS vendor. I am using BIND9.

When I deploy virtual machines, I provision them to my production DNS zones and VLANs.

[jaylavine]

if you are still around.
can you share more specifics of this config?

why do you do the forwarding?

do nodes pxe boot when not on the same net as the MAAS net?
I have had problems with nodes pxe booting with multiple nics/networks and found that it needs to pxe of the primary net, the maas net.
but of course I could be doing something wrong.

[nateybobo]

For me, PXE boot issues are usually related to broken DHCP-relay across my VLANs. I would check the network settings after you commission and before you deploy your hosts.

The reason I carved out a subdomain for MAAS is because I have my own internal authoritative DNS servers. The rub comes when MAAS wants to be authoritative on a DNS zone for provisioning. It doesn’t have the functionality to update another DNS server’s zone, so it just run its own.

Therefore, when systems on my internal network do a DNS query for a MAAS host, my DNS server will see the request is for a MAAS domain, and FORWARD it to MAAS to get the answer.

I hope that makes sense.

[jaylavine]

yep, makes sense.
thanks for the reply.