I have an odd problem concerning NFS. I run a JBOD disk array of 4 disks in a tower with ext4 file system. The disks are mounted nicely. I also use them as NFS mounted drives with two disks mounted for my whole home network as read only and all four disks mounted for my desktop pc as rw. I remember, that it worked. Though, after some time it happened that 2 disks broke (it was recertified disks) and i had to change them. I remember that i changed one disk from one slot to another. The second broken disk i replaced by a new disk of same capacity. Now, on my desktop pc, only three of the four disks mount correctly. The third mount point simply reconnects the first disk, leaving the third disk unmounted.

I have found in the internet older posts from 2012 and also later, detailing exactly this problem for NFS.

I also noted an odd thing about the disk tower. The disk tower is JBOD configured and all the management is done by my Raspberry Pi 4 4GB running Raspbian. Though, all four disks are mounted correctly on the server itself, when booting the server or reawakening the tower, the control lights of the tower for the disks switch on in a strange pattern. Disks 1-3 are cycled through several times and the light for Disk 4 is only flickering at the end. I assume that some of the kernel mount tables keep information that messes up the NFS mounts on the client, or the client keeps up some tables that are messed up. The disk tower is operated as a JBOD, so the tower itself should not store any info, being without any controller apart from some switch controlled SCSI controller.

What could be the problem here?

The first thing I would check is the /etc/exports file on the server. Then the /etc/fstab file on your desktop pc.

Done that. The /etc/fstab file of the server lists the correct partition UUIDs for the four disks, also all four are nicely accessible via file explorer under their mount points on the server. /etc/exports looks like this:

/mnt/D01 192.168.1.X(rw,sync,no_root_squash) 192.168.1.0/24(ro,sync,no_root_squash)
/mnt/D02 192.168.1.X(rw,sync,no_root_squash)
/mnt/D03 192.168.1.X(rw,sync,no_root_squash)
/mnt/D04 192.168.1.X(rw,sync,no_root_squash)

exportfs command lists as follows:

/mnt/D01 192.168.1.X
/mnt/D02 192.168.1.X
/mnt/D03 192.168.1.X
/mnt/D04 192.168.1.X
/mnt/D01 192.168.1.0/24

What happens is that on the NFS client side of my desktop pc 192.168.1.X D01, D02, D04 are mounted correctly on their NFS mountpoints, but instead of D03 D01 is mounted on the NFS mount point of D03.

the /etc/fstab of my desktop 192.168.1.X lists the NFS disks as follows:

192.168.1.Y:/mnt/D01 /mnt/D01 nfs rw,auto,nofail,_netdev 0 2
192.168.1.Y:/mnt/D02 /mnt/D02 nfs rw,auto,nofail,_netdev 0 2
192.168.1.Y:/mnt/D03 /mnt/D03 nfs rw,auto,nofail,_netdev 0 2
192.168.1.Y:/mnt/D04 /mnt/D04 nfs rw,auto,nofail,_netdev 0 2

Take a look at /etc/hosts on D03 - do you have hosts entries defined for D01-4?
Check output of "host D01" from D03, is it correct IP for D01?
What is output of "showmount -e" on D03? Is it correct and what you expect? Then, compare with other D0's?

showmount -e on my desktop shows: "clnt_create: RPC: Program not registered"


showmount -e on my server where the disks are connected shows: "clnt_create: RPC: Unknown host"

/etc/hosts on my server shows no entry on D01-D04

/etc/hosts on my desktop also shows no such entries