AlmaLinux 9: NFS export and mount (tested on one host)

Test scope

On a real AlmaLinux 9.8 host (kernel 5.14.0-687.54.1.el9_8): enabled nfs-server, wrote /etc/exports.d/lab.exports, ran exportfs -ra, exportfs -v, showmount -e 127.0.0.1, mount -t nfs 127.0.0.1:/srv/nfslab /mnt/nfslab, findmnt, df -hT, created a file, unmounted and cleaned up. Log: plans/content-expansion/evidence/logs/alma-nfs-mount.log.

Cross-host mounts, firewall rules for 2049 and autofs were not tested here (this was a loopback test on one host).

What a loopback mount can prove

Mounting the host’s own export through 127.0.0.1 validates the protocol and server configuration: the export is active, showmount sees it, the client mounts it as nfs4, and read/write works. It cannot validate the network path (firewall, routing, cross-host latency) — that needs a second machine.

Put the export in exports.d

sudo systemctl enable --now nfs-server
sudo mkdir -p /srv/nfslab && sudo chmod 777 /srv/nfslab
echo "/srv/nfslab 127.0.0.1(rw,sync,no_root_squash)" | sudo tee /etc/exports.d/lab.exports
sudo exportfs -ra

/etc/exports.d/*.exports is equivalent to /etc/exports and easier to manage per purpose; after editing you must run exportfs -ra.

Check the export and mount it

sudo exportfs -v
showmount -e 127.0.0.1

/srv/nfslab     127.0.0.1(sync,wdelay,hide,no_subtree_check,sec=sys,rw,secure,no_root_squash,no_all_squash)
Export list for 127.0.0.1:
/srv/nfslab 127.0.0.1

Mount and confirm the filesystem type:

sudo mkdir -p /mnt/nfslab
sudo mount -t nfs 127.0.0.1:/srv/nfslab /mnt/nfslab
findmnt -no SOURCE,FSTYPE,TARGET /mnt/nfslab
df -hT /mnt/nfslab | tail -1

127.0.0.1:/srv/nfslab nfs4   /mnt/nfslab
127.0.0.1:/srv/nfslab nfs4   39G  2.4G   37G   6% /mnt/nfslab

FSTYPE is nfs4, not nfs — NFSv4 is what gets negotiated by default. Verify read/write:

sudo touch /mnt/nfslab/hello && ls -l /mnt/nfslab/hello

-rw-r--r-- 1 root root 0 Oct 19 13:30 /mnt/nfslab/hello

Cleanup order

sudo umount /mnt/nfslab
sudo rm -f /etc/exports.d/lab.exports
sudo exportfs -ra
sudo systemctl disable --now nfs-server

Unmount first, then withdraw the export, then stop the service. Reversing the order leaves a mount whose export is gone and unmounting reports device is busy.

Common traps

• Editing exports without exportfs -ra: showmount never shows the new directory;
• Cross-host mounts blocked by the firewall: allow nfs (2049/tcp) and, for v3, the mountd/rpc-bind ports — a loopback test cannot reveal this;
• Treating no_root_squash as default: it gives the client’s root server-side root; keep root squash unless you truly need otherwise.

Practice: export and mount on one host

sudo systemctl enable --now nfs-server
sudo mkdir -p /srv/nfslab && sudo chmod 777 /srv/nfslab
echo "/srv/nfslab 127.0.0.1(rw,sync)" | sudo tee /etc/exports.d/lab.exports
sudo exportfs -ra
sudo mkdir -p /mnt/nfslab && sudo mount -t nfs 127.0.0.1:/srv/nfslab /mnt/nfslab
findmnt -no FSTYPE /mnt/nfslab     # expected nfs4
sudo touch /mnt/nfslab/x && ls /mnt/nfslab
sudo umount /mnt/nfslab

Checks:

1. showmount -e 127.0.0.1 → lists /srv/nfslab
2. findmnt -no FSTYPE /mnt/nfslab → nfs4
3. A file can be created inside the mount
4. After unmounting, findmnt /mnt/nfslab prints nothing

FAQ

常见问题

Why does FSTYPE show nfs4 instead of nfs?

NFSv4 is the default negotiated version, so mount -t nfs uses nfs4. Forcing v3 requires -o vers=3 and also opening the dynamic mountd/rpc-bind ports.

Loopback mount works but a remote host times out?

Loopback bypasses the firewall. A remote mount needs nfs (2049/tcp) allowed, plus mountd/rpc-bind for v3. Add firewall-cmd --add-service=nfs on the server and retry.

What if unmounting reports device is busy?

Find what is using the mount (lsof +D /mnt/nfslab or fuser -vm), leave the directory and retry. umount -l defers the cleanup — know that it delays rather than solves the problem.

Next steps

• Growing storage: LVM growth and XFS
• Post-install checks: 7 checks after install
• Auditing sources: Package source audit