Hi Guys!
I'm not super familiar with Linux kernel so sorry in advance for the long post. I'm currently trying to install BBRV3 on an Azure VM, and I'm using a Ubuntu 22.04 LTS image which has kernel version 6.2.0. I use SSH to communicate with this VM with no issue.
At first I simply tried to use the gce-install.sh script and it actually ran without any issues. But after rebooting I can not SSH into the VM anymore.
After that I tried to manually compile and install the kernel. If I remember correctly there were some manual installation steps included sometimes back in BBRV1. Specifically what I did:
1. install all the necessary dependencies
2. clone BBRV3 branch
3. make oldconfig (Answering many questions relating to the Kernel and enable BBR)
4. Generate a X.509 certificate to sign the modules while compiling
5. Modify .config file to include the certificate that just generated
6. make prepare
7. make -j`nproc`
8. make -j`nproc` modules
9. sudo make -j`nproc` modules_install install
10. Select the correct kernel to boot into (At this step, now I have a new kernel Ubuntu Kernel 6.4.0+, which is the one that I choose as the default kernel to boot into)
11. Reboot the VM
All of the above manual compilation steps also runs fine for me but I still can not SSH into the VM after the reboot. However, a while ago I tried these manual compilation steps on another Azure VM and successfully installed V2alpha branch which resulted in a new 5.13.12 kernel, and I was able to successfully boot and SSH into this kernel.
So some questions I have:
1. Is the gce-install.sh script only works for gce instances? What are some of the modifications I might need to do for it to work on Azure?
2. Are the manual installation steps that I described above seem to be correct? Am I missing anything?
3. What is the stable kernel version that BBRV3 is based on? Could it be that the built kernel version is incorrect?
4. Could this be an VPC (Azure VM) issue?
4. I just noticed that there is a new branch bbr-el9-rhel9 which is waiting to be merged into master, could this become the recommended way for installation on arbitrarily Linux machines?
Thanks a lot for the help in advance!
Best,
Dunk