[Discuss] Dracut and fstab

[Peter Willis]

I made a couple errors to begin with.

I was originally attempting to ‘—add'  things to intramfs using multiple calls to dracut.
The optimal method turned out to be performing a single dracut operation for all operations together.

Second to that, the combination of ‘—fstab’  and ‘—fstab-file’  seemed to cause some conflict with ‘—add-drivers’ and ‘—add’ parameters.
This caused some modules to be negated from the initramfs (as visualized by ‘lsinitrd’). Simply using ‘—fstab-file’ loads the fstab as expected.

While I am continuing to work on loading of the NFS modules prior to nfsboot, the PXE is loading the proper initramfs image and the supplied root file system fstab.

The various documentation seem vague and aged. There seem to be varied methods for setting up the tftp and dhcpd. None seem to match the documentaion provided by Redhat. Likewise the infiniband manufacturers docs are obsolete with respect to using dracut and fully expect an initramfs file system prior to cpio.
Better information appeared at kernel.org, but the documentaion seems to favor local boot with PXE being vaguely covered.

The NFS modules seem to recuire an insmod script in an early run level prior to the system mount.
While th exports are present, the boot does not appear to mount them which proably inicates the nfs and nfsv4 modules are not loaded at mount time.

In retrospect debian, prior to dracut, was easier to do this with. The Redhat intramfs does not cpio into distinct files. The output is a single monolithic binary that can only be listed and examined using ‘lsinitrd’. The old system used files that could be examined and edited if needs be, including any scripts required in the ramfs for boot.

Perhaps I am overthinking this. The DHCP settings should allow a boot. Maybe I don’t require fstab if I supply DHCPD settings for the nfs root.

The old Debian system was files, but at least it was cogent.

P


> On Feb 5, 2020, at 1:28 PM, Jeremy Van Veelen <techgooroo at gmail.com> wrote:
> 
> Hi Peter,
> 
> It sounds like an interesting issue, what did you end up doing to resolve it?
> 
> Curiously,
> Jeremy