I’ve got an interesting situation I can’t seem to solve, as the cursory documentation doesn’t address this use case.
I have an AlmaLinux server with four SAS JBOD shelves attached to it for a total number of 336 Hard Drives (84*4).
I installed the zfs-dkms packages on it per the instructions here: RHEL-based distro — OpenZFS documentation
After that, I can’t seem to boot regularly. In the emergency shell, all my regular LVM partitions like /home, /opt, /var, and / load correctly, but /boot and /boot/efi partitions do not. I have to run partprobe /dev/sda -s to get those two to load. From the logs, I think it is simply timing out waiting for ZFS to process all my disks.
What I want to attempt to do is this: remove the ZFS kernel module from being automatically loaded then create a script and accompanying systemd unit with After=multi-user.target (AKA my system has booted as normal) to then load the ZFS kernel module.
I’m aware of how to blacklist a kernel module, which stops it from being loaded automatically, but it also stops me from being able to load it manually.
How can I go about this? Or would it be better to increase the timeout on the XFS filesystem so that the disks can just wait a little longer?
Thanks. Happy holidays.