setting hostnames for control-planes and workers

[bflance]

Hi,

Maybe i am asking somthing that was already answered and is probably already in documentation, but i still cant get my head around it.

all i need, is to set hostnames of nodes on newly created talos cluster

so, when i run this:

kubectl get nodes  -o wide                                                                    
NAME            STATUS   ROLES           AGE   VERSION   INTERNAL-IP    EXTERNAL-IP   OS-IMAGE         KERNEL-VERSION   CONTAINER-RUNTIME
talos-6in-fq6   Ready    control-plane   12m   v1.32.0   192.168.0.14   <none>        Talos (v1.9.3)   6.12.11-talos    containerd://2.0.2
talos-6sb-u94   Ready    control-plane   12m   v1.32.0   192.168.0.31   <none>        Talos (v1.9.3)   6.12.11-talos    containerd://2.0.2
talos-bpu-qji   Ready    control-plane   12m   v1.32.0   192.168.0.29   <none>        Talos (v1.9.3)   6.12.11-talos    containerd://2.0.2

i get meaningful hostnames instead of 'talos-xxx-xxx'

So, to test it i did following,
booted a new VM on proxmox from talos-1.9.3.iso into maintenance mode, and tried to set its hostname
when VM booted it, its hostname as seen on proxmox console is '(no hostname)', and it makes sense, VM is currently just booted , its in "maintenance mode" and awaiting for config..

now, i do this, ping the VM to make sure connectivity fro my laptop is OK

ping -c 2 192.168.0.44
PING 192.168.0.44 (192.168.0.44): 56 data bytes
64 bytes from 192.168.0.44: icmp_seq=0 ttl=64 time=2.450 ms
64 bytes from 192.168.0.44: icmp_seq=1 ttl=64 time=2.887 ms

created a config file to set hostname

cat git/proxmox-talos/14-set-hostname.yaml
machine:
  network:
    hostname: talos-node-01

tried to apply it to the VM/node:

talosctl apply-config --insecure  -n 192.168.0.44 --file git/proxmox-talos/14-set-hostname.yaml
error applying new configuration: rpc error: code = InvalidArgument desc = configuration validation failed: 3 errors occurred:
	* cluster instructions are required
	* install instructions are required in "metal" mode
	* issuing CA or some accepted CAs are required (.machine.ca, machine.acceptedCAs)

i really dont understand something here..
commands like "get disks" do work,

talosctl get disks --insecure --nodes 192.168.0.44
NODE   NAMESPACE   TYPE   ID      VERSION   SIZE     READ ONLY   TRANSPORT   ROTATIONAL   WWID   MODEL           SERIAL
       runtime     Disk   loop0   1         4.1 kB   true
       runtime     Disk   loop1   1         57 kB    true
       runtime     Disk   loop2   1         684 kB   true
       runtime     Disk   loop3   1         2.7 MB   true
       runtime     Disk   loop4   1         74 MB    true
       runtime     Disk   sda     1         34 GB    false       virtio      true                QEMU HARDDISK
       runtime     Disk   sr0     1         108 MB   false       ata                             QEMU DVD-ROM

do i need to bootstrap cluster first? and then change hostnames?
seems counter intuitive since by changing hostname AFTER cluster is up, it seems to me, it will make problems with certificates, etc..

thank you for your answers..

[smira]

So this is a good question, there is a short answer and a long answer.

Short answer you need to apply:

machine:
  network:
    hostname: talos-node-01

... as a config patch to the machine configuration when you first apply configuration to the machine in maintenance mode. This way the hostname will be set from the very beginning.

By the way, Talos supports changing machine configuration on the fly, including the hostname. The only thing you would need to do is to clean up old Nodes from Kubernetes.

Long answer is related to the way Talos machine configuration is structured. Most of the config today is part of a huge monolithic document which can't be applied in parts (only full document can be applied). Talos is moving towards splitting up into parts, small documents for a specific purpose. Once we're done with this work, you will be able to change hostname in maintenance mode by submitting just a hostname config document. But we're not there yet.

[bflance]

Thank you for your answer @smira

in the end i found a way which would automate this for me.

in my "big deploy script", i have put a function like so:

    ## apply config for each node  + setup dynamic meaningful hostnames (comment --config-patch if you want defaults)
    for node in ${NODE_IPS[*]}; do
        hostname="node-${node//./-}"  # Replace dots with dashes
        talosctl apply-config \
            --insecure --nodes $node --file controlplane.yaml \
            --config-patch "[{\"op\": \"replace\", \"path\": \"/machine/network/hostname\", \"value\": \"$hostname\"}]"
    done

what it does is basically, applies a dynamically made hostname created from IP address of that node which was fetched from VM (provided by proxmox).
so the end result for me was like this:

kubectl get nodes
NAME                STATUS   ROLES           AGE     VERSION
node-192-168-0-14   Ready    control-plane   4m45s   v1.32.0
node-192-168-0-29   Ready    control-plane   4m22s   v1.32.0
node-192-168-0-31   Ready    control-plane   4m18s   v1.32.0

yes, i know, i replaced one "automatic hostname settings system" with another,
the reason i am doing this, is to first of all learn Talos and get used to all of the features and tricks, to be more proficient with it.

also a quick question:

does setting custom hostnames, in any way, shape or form, can introduce problems with high availablity?
what i mean is, are hostnames are local to each node or they are used for inter-node communication? (i think/hope, the communication is IP based and not hostnames based..)

[smira]

does setting custom hostnames, in any way, shape or form, can introduce problems with high availablity?
what i mean is, are hostnames are local to each node or they are used for inter-node communication? (i think/hope, the communication is IP based and not hostnames based..)

No, hostnames are not used in Kubernetes to resolve (like DNS resolving), unless you want to do that yourself.

Talos has an option to make hostnames resolvable within the cluster (in hostDNS settings), but it's not required, it just adds some handy stuff like you can do talosctl -n <hostname> version.

[bflance]

thanx!!

you mentioned:

Short answer you need to apply:

machine:
  network:
    hostname: talos-node-01

sorry for asking,
how you do that?

what i am doing in my function, is simply talosctl apply-config ....
which in fact applies whole config including controlplane.yaml + my patch > reboots node.
is this something that you have meant?
just double checking..

let me clarify, is it possible to change hostname of just booted VM in maintenance mode without applying whole configuration + reboot,etc..

and if yes, how to do it 🙏

[smira]

let me clarify, is it possible to change hostname of just booted VM in maintenance mode without applying whole configuration + reboot,etc..

no, without other ways - e.g. you can make DHCP supply the hostname to the machine.

the reboot is caused by the installation process in your case, it's not related to config changes.

[itoffshore]

I've just started experimenting with talos on a remote cluster without DHCP

• If you have VNC access you can also press F3 to enter network settings (also useful when there's no DHCP)
• the hostname can be set at the top / optionally set a static ip & gateway
• having controlplane.yaml & worker.yaml commented in detail is very nice - apply customizations here

I should have spun up talos much sooner - thanks guys ;o)