Learning AlmaLinux - Kernel, Boot & Kernel Modules

#Introduction

In the previous episode, Episode 9, we set up networking and the firewall. Now we touch the most fundamental — and for many admins, the most intimidating — part of the system: the kernel and boot process. When a system fails to boot, all your other skills become useless — and the ability to understand and fix boot issues is what separates an ordinary sysadmin from a senior one.

This episode breaks down the system's journey from the power button to the login shell, then covers kernel management: choosing versions, tuning parameters, and managing modules.

#The Boot Process: From Firmware to Systemd

Linux booting isn't a single leap — it's a chain of components calling each other:

Firmware (UEFI/BIOS)
    -> GRUB2 bootloader
        -> Kernel + initramfs (dracut)
            -> systemd (PID 1)
                -> targets & services

Let's examine each step.

#Firmware: UEFI and BIOS

When you press the power button, the firmware (UEFI on modern machines, or BIOS on older ones) runs a self-test and looks for the bootloader. In episode 3 we covered Secure Boot — the boot component signature verification that's a flagship UEFI firmware feature.

#GRUB2

GRUB2 is AlmaLinux's default bootloader. It reads its configuration from /boot/grub2/grub.cfg, shows a menu (if more than one kernel exists), then loads the selected kernel. This menu is your lifesaver when a new kernel misbehaves — you can boot into the old kernel.

sudo grubby --info=ALL | grep -E "^kernel|^index"

grubby --info=ALL shows all kernel entries GRUB knows about, complete with their index and boot parameters.

#The Kernel and Initramfs

The kernel is the heart of the system — managing processes, memory, hardware, and networking. But the kernel needs help to boot: the initramfs (initial RAM filesystem) is a small archive loaded into memory, containing the drivers and tools the kernel needs to access the root disk. The initramfs is built by dracut:

sudo dracut -f

This command rebuilds the initramfs for the active kernel — useful after changing modules needed at boot.

#Systemd as the Finisher

Once the kernel is ready, systemd runs as PID 1 and takes over: bringing up targets, starting services, and mounting filesystems. Everything we covered in episode 7 happens at this stage.

#Recovery and Rescue Modes

When a system fails to boot, two rescue paths are available:

Both can be reached from the GRUB menu by adding a boot parameter, or through the following command while the system is running:

sudo systemctl isolate rescue.target

#Kernel Management

#Upgrading the Kernel

The kernel on AlmaLinux is managed as a package — and several versions can coexist. installonly_limit=3 in dnf5.conf (episode 4) ensures old versions remain available as a fallback.

sudo dnf5 upgrade kernel

After installation, GRUB automatically makes the newest kernel the default:

sudo grubby --default-kernel

#Default vs Latest Kernel

Sometimes you want to boot a specific kernel, not the newest one. grubby handles this:

sudo grubby --set-default=/boot/vmlinuz-6.7.0

grubby --default-kernel shows the kernel that will be loaded on the next boot; --set-default selects it explicitly. This is very useful if the newest kernel has issues and you want to lock in a stable version.

#Kernel Tuning with sysctl

Runtime kernel parameters can be changed without rebooting via sysctl. Parameters are written in /proc/sys/ and configured persistently in /etc/sysctl.d/.

sysctl -a | head
sysctl net.ipv4.ip_forward

net.ipv4.ip_forward = 1
net.core.somaxconn = 1024
vm.swappiness = 10

To apply new configuration:

sudo sysctl --system
sysctl net.ipv4.ip_forward

vm.swappiness = 10 reduces the system's tendency to move data into swap — a common tuning for servers that want to preserve application performance. Convention: numbered files in /etc/sysctl.d/ are applied in order, so the 99- name wins over default files.

#Kernel Modules

#Viewing and Loading Modules

The kernel supports modules — components loaded as needed, such as drivers for new hardware:

lsmod | head

sudo modprobe vfio-pci
sudo modprobe -r vfio-pci

lsmod shows active modules; modprobe loads or unloads modules along with their dependencies.

#Module Persistence

To have a module load automatically at boot, register it in /etc/modules-load.d/. To exclude or blacklist modules, use /etc/modprobe.d/:

blacklist nouveau
install nouveau /bin/true

This kind of blacklist is commonly used to prevent specific drivers from loading — for example disabling the open NVIDIA driver so it doesn't conflict with proprietary drivers.

#Common Pitfalls

1. Removing the old kernel before testing. Keep at least one backup kernel — installonly_limit=3 is a good reason.
2. Forgetting grubby when choosing a kernel. The default boot follows the newest kernel; lock in a stable version with grubby --set-default.
3. Editing /proc/sys/ directly without sysctl.d. Direct changes are lost on reboot — write to /etc/sysctl.d/.
4. Running dracut -f without a reason. It's usually only needed after boot module or driver changes.
5. Bootstrap-loading modules with insmod. Use modprobe, which handles dependencies automatically.

#Conclusion

In this episode 10 you've understood the kernel and boot process: the firmware → GRUB2 → kernel + initramfs → systemd chain, the rescue and emergency recovery modes, kernel version management with grubby, runtime tuning with sysctl, and kernel module management with modprobe and lsmod.

Key takeaways:

• Boot chain: UEFI/BIOS → GRUB2 → Kernel + initramfs (dracut) → systemd.
• Always keep a backup kernel; use grubby --set-default to choose a version.
• Tune at runtime via /etc/sysctl.d/ with sysctl --system — not by editing /proc/sys/ directly.
• Modules are managed with modprobe and lsmod; policy persists in /etc/modprobe.d/.
• rescue.target and emergency.target are rescue paths when boot fails.

A well-managed kernel is the foundation of a healthy system. In the next episode, Episode 11, we'll cover Cockpit, Monitoring & Web Management — the web dashboard that lets you manage your system from a browser, plus basic monitoring commands like top, free, df, vmstat, and iostat. See you there!