I have 3 identical machines, deployed the same way using maas which have multiple NICs and bridges with VLANs - but the DNS entries generated by maas are inconsistent. Short description what is the setup like:
Each server has 2 10Gb NICS bonded using balance-alb which is a trunk of 3 VLANS:
br-storage
=> vlan.20 (10.20.0.0/24)
br-public
=> vlan.5 (Public IPs for VMs passed through here - no IP on the host )
br-vxlan
=> vlan.1010 (10.10.0.0/24) LAN + vxlans for VMs
Plus
local docker bridge 172.17.0.1
local virt-bridge virbr0 192.168.100.1
all machines should show up with their plain hostname (gin01 - gin02 - gin03) on the vlan.1010 interface and address (10.10.0.0/24) - but they seem to randomly take any IP and bridge so it show up like this:
The network configuration seen by MAAS for each of these machines is structured and named identically but still in DNS I find different records for each machine (the 10.10.0.0/24 IPs should be pointing to the plain hostname)
I would like to be able to control which of the interfaces and IPs should be the A record with the real, default hostname ( e.g. gin01) - instead they are seemingly randomly registered with bridge-names in front or not - how can this be controlled ?
Can this be fixed for existing machines/DNS records? How can I control which interface is the “default” IP to be used for the plain hostname?
By design MAAS will assign the DNS name <hostname>.<domain> to the boot interface. All the other interfaces will get <interface_name>.<hostname>.<domain>
That does not seem to be completely true - at least not for 3.5 the boot interfaces were always exactly those in 10.10.0.0/24 network - actually it PXE boots ffrom one of the bonded eth interfaces, then during deployment it will form a bond interface from both NICs and configure the individual vlan interfaces which are slaves of that bond and get attached to their respective bridges - so this mechanism does nit seem to make sense in a case like that to me …
actually all of the machines originally PXE booted from their first 10Gb NIC (eno49) for the deplyoment but that interface is not there anymore as a dedicated, seperate interface with an IP at all in the end.
If you have a look at the image you can see one of the hosts (gin02) even has the docker bridge IP assigned to its hostname … - dont tell me it might have booted from there
Of course the network configuration changed after initial PXE boot - changes are synced back to the MAAS server and these changes update DNS records inapropriately and cause issues.
This functionality cannot be blocked or controlled in any way as far as I understand this tool. I also cannot edit the built-in DNS records coming from these sync operations…
To me it looks like MAAS needs to redesign or extend this concept regarding DNS …
