You probably deploy your OCP/OKD cluster through PXE/DHCP or static IPs with manual interactions. DHCP is the most convenient way to deploy nodes in an automated Way. But what about having to init a new environment without DHCP at all remotely? Suppose you have an OCP4/OKD cluster that you have to deploy, but there is no DHCP environment due to security constraints.
I want to show you a smart way to run pre-checks and install through lightweight iso image.
- BIOS - 1MB
- EFI - 2MB
Based on the iPXE network bootloader: https://github.com/ipxe/ipxe
This (iPXE) is not a supported by Red Hat. Use it at your own please and risk!
Why not ?
This project's primary goal is to automate Openshift/OKD cluster installation and run some pre-checks even before the OS deployment.
Available checks:
- Network connectivity - ping gateway
- Check DNS records such as api, api-int, *.apps, hostnames
- Check if the IP corresponds to the correct hostname.
- Network device detection based on busdev (Good for labs, so far just VMware and libvirt)
1 - Ping Gateway
2 - Check dns - hostname: hostname.cluster.domainname
3 - Check dns - api : api.cluster.domainname
4 - Check dns - api-int : api-int.cluster.domainname
5 - Check dns - *.apps : *apps.cluster.domainname
6 - Try to detect net_interface based on busdev
- ens192 0000:0b:00.0 ( vSphere nic 1)
- ens224 0000:13:00.0 ( vSphere nic 2)
- ens33 0000:02:01.0 (Vmware workstation nic 1)
- ens38 0000:02:06.0 (Vmware workstation nic 2)
- enp1s0 0000:01:00.0 (libvirt)
7 - Menu Selector based on Node Role.
- Role (master,worker,bootstrap)
8 - Fetch OCP Images - kernel, initrd and rootfs
9 - Install CoreOS.
The SmartPXE offers the following configuration parameters.
| Parameter | Description |
|---|---|
| cluster_name | OCP Cluster Name |
| domain_name | Domain Name |
| fileserver | Webserver:port that holds the master.ign,bootstrap.ign,worker.ig,kernel,initrd and rootfs images. |
| kernel | The RHCOS live-kernel |
| initrd | The RHCOS live-initrd |
| rootfs | The RHCOS live-rootfs |
| install_drive | Drive to install RHCOS on |
| net_interface | NIC device name |
| gateway | Default router IP |
| dns_server | DNS server - For now, it supports just one. |
| netmask | Default netmask |
| bootstrap_hostname | Bootstrap Node hostname |
| bootstrap_ip | Bootstrap Node IP Address |
| bootstrap_role | Bootstrap Node Role(Mandatory) It is obvious - bootstrap |
| master_hostname | Master Node hostname |
| master_ip | Master Node IP Address |
| master_role | Master Role(Mandatory) It is obvious - master |
| worker_hostname | Worker Node hostname |
| worker_ip | Worker IP Address |
| worker_role | Worker Role(Mandatory) It is obvious - worker |
X is the number of the Master Node: From 1 to 3
Y is the number of the Worker Node: From 2 to Up to you
cluster_name ocp4
domain_name example.com
fileserver 172.16.160.1:8080
kernel rhcos-live-kernel-x86_64
initrd rhcos-live-initramfs.x86_64.img
rootfs rhcos-live-rootfs.x86_64.img
install_drive /dev/sda
net_interface ens33
gateway 172.16.160.1
dns_server 172.16.160.1
netmask 255.255.255.0
bootstrap_hostname bootstrap
bootstrap_ip 172.16.160.20
bootstrap_role bootstrap
master1_role master
master1_hostname master001
master1_ip 172.16.160.21
master2_role master
master2_hostname master002
master2_ip 172.16.160.22
master3_hostname master003
master3_ip 172.16.160.23
master3_role master
worker1_hostname worker001
worker1_ip 172.16.160.30
worker1_role worker
worker2_hostname worker002
worker2_ip 172.16.160.31
worker2_role worker
.
.
.
worker<Y>_hostname worker002
worker<Y>_ip 172.16.160.31
worker<Y>_role worker
git clone https://github.com/ralvares/smartpxe
cd smartpxe
vi inventory
podman build -t smartpxe -f Dockerfile .
podman create --name smartpxe smartpxe --entrypoint /
podman cp smartpxe:/ipxe/src/bin/ipxe.iso ipxe_bios.iso
podman cp smartpxe:/ipxe/src/bin-x86_64-efi/ipxe.iso ipxe_efi.iso
podman rm smartpxe
From now, you just need to boot up all the nodes using the iso image, select the respective node you want to install and go grab a Coffee :)
SmartPXE menu selector preview:
- Multiple DNS Servers
- VLAN Tagging
- Bonding interface
- Second Network Interface