Rbd module for ceph

Hey
I am trying to add the rbd module to the default kernel, as I am at the beginning of the whole adventure with this task, can you advise if the public version of os’es Debian, Ubuntu or Armbian for ROCK5B have any known issues that prevent me from manually compile and adding rbd module to the kernel? Main intention of this task is to get this module for installation ceph (under the rook project).
Thank you

Main issue is that main market for rockchip are android devices and thanks to that supported kernel is merged from that branch, where there is not much development It’s not the same thing as the one You see in x86 ubuntu or debian, even if version is almost same.

Just try to build it with favourite build system and linux flavour. Some modules can be just included, others require additional work or causes kernel panic, some instability. You will need to test that.

Each build system has it’s own advantages and complexity. None of them is perfect or universal for all use cases. Good point to start is https://wiki.radxa.com/Rock5/guide/build-debian-from-debos-radxa
Armbian needs to be universal across many other boards and version. Whatever you choose - good luck :slight_smile:

1 Like

@dominik

Thanks for showing the way, after compiling the kernel along with the rbd module then propagated the new kernel version to the all worker nodes in the kubernetes cluster, this allowed to complete the rest of the steps related to deploy rook ceph on k8, now remains me to check the stability !

This particular module should not be a problem to run, did You choose legacy or mainline/edge kernel version?
Looking forward for Your insights about whole process and results on rock5, for now I’m using longhorn with rock5 (as well as other boards) but also looking to switch to openebs and ceph (and comparing both with commercial data fabric).

In my case to compile kernel 5.10.110-X-rockchip I went through the
steps from the below web page
https://wiki.radxa.com/Rock5/guide/build-kernel-on-5b

but I will have to revisits the steps because new kernel has writable clk DebugFS enabled , what cause that my system is exposed on security issues , so will have to rebuild with disabled flag CONFIG_IOMMU_DEBUGFS

K3s has nice script to check system health, network and most of needed and optional kernel features, most of them are easy to configure on kernel config, just type:

curl https://raw.githubusercontent.com/k3s-io/k3s/master/contrib/util/check-config.sh | bash

If You are using debian also remember to switch iptables to legacy.

Thanks for tip, generally in my case Kubernetes env based on k8s (with container.io), and callico as CNI,
hardware layer consists three workers node (arm64) and one master (x86), so far env seems to be stable ,the intention of this lab is self study so we will see where this takes me :slight_smile: