NFS Shares: storage is not online (500)

[arnaud056]

Hello forum,

I've got this issue between 2 proxmox servers (v4 and v5, both up-to-date, both physical machines connected by a switch).
Both machines can ping each-other and their IP and hostname are present into the /etc/hosts of the other machine.

I used this simple method that worked well a few years ago but now I get a "NFS Shares: storage is not online (500)" by configuring the nfs storage of the client machine.

On the server machine:
# ls -l /home/
drwxr-xr-x 6 root root 4096 déc. 27 20:54 nfs <= that is the shared folder

/etc/exports:
/home/nfs 192.168.20.132(rw,sync,no_subtree_check) <= the IP of a "test-client" fedora
/home/nfs 192.168.20.150(rw,sync,no_subtree_check) <= the IP of the proxmox client
/home/nfs 192.168.20.160(rw,sync,no_subtree_check) <= the IP of the proxmox server itself (for tests)

/etc/hosts.allow and /etc/hosts.deny are empty.

On the proxmox client machine:
# cat /etc/pve/storage.cfg
nfs: nfs_proxmox
export /home/nfs
path /mnt/pve/nfs_proxmox
server 192.168.20.160 <= the IP of the proxmox nfs server
content vztmpl,iso,rootdir,images,backup
maxfiles 20
options vers=3

# mount -t nfs 192.168.20.160:/home/nfs /mnt/proxmox-asrock
gets in time-out.

# showmount -e 192.168.20.160
Export list for 192.168.20.160:
/home/nfs 192.168.20.160,192.168.20.132,192.168.20.150
This result is correct, but I'm very surprised that it takes more than 1 min. to get it.

From the fedora test client: it is the same: mounting gets in time out.

From the proxmox server itself: the nfs connection on itself works good and immediately => I assume that /etc/export is correct.

Does somebody has got an idea where the issue could come from?
Thanks in advance.
Regards
Arnaud

 

[arnaud056]

Hi again,
in between I had the glorious idea to connect directly the proxmox nfs server to the test client fedora (= with a hub and not wit the switch) => the nfs share worked immediately.

The issue must therefore come from the switch, but I've got no idea which setting can block the nfs sharing.
What I can say is that the connection between all the machines works over a VLAN:
- by both proxmox machines, the network setting of the servers are without VLAN and the switch is set to put the interfaces where both proxmox are connected to into the VLAN (=connection to the VLAN done by the switch)
- by the test client fedora, it is the contrary: the connection to the VLAN is done by the machine and not by the switch

For the connection over the hub, I set the fedora client not to use the VLAN.

Is VLAN a problem for NFS sharing?
Which parameter of the switch (HP) could make trouble?

Thanks.
Arnaud

 

[arnaud056]

Hi again,

the issue is solved. The HP switch made trouble with NFS: https://community.hpe.com/t5/Web-an...to-DoS-quot-prevents-NIS-and-NFS/td-p/4693460 (old problem year 2012....).
To get NFS working, the "Auto DoS" feature must be disabled.......

Bye
Arnaud