How to debug nfs client connection timeout?

Question

I have 1 nfs-server(10.192.244.109) & 2 nfs-client there.

On one client, both ok to mount the nfs-server directory with v3 & v4, while on another client, it's only ok to mount the nfs-server directory with v4, the client reports connection timeout if use v3.

• Client 1 ok for v3 and v4:

root@shubuntu1:~# mount -t nfs -o nfsvers=3 10.192.244.109:/var/lib/lava/dispatcher/tmp delete
root@shubuntu1:~# umount delete
root@shubuntu1:~# mount -t nfs -o nfsvers=4 10.192.244.109:/var/lib/lava/dispatcher/tmp delete
root@shubuntu1:~# umount delete

• Client 2 just ok for v4 but not ok for v3:

root@LAVA-Debug:~# mount -t nfs -o nfsvers=4 10.192.244.109:/var/lib/lava/dispatcher/tmp delete
root@LAVA-Debug:~# umount delete
root@LAVA-Debug:~# time mount -t nfs -o nfsvers=3 10.192.244.109:/var/lib/lava/dispatcher/tmp delete
mount.nfs: Connection timed out

real    2m5.713s
user    0m0.079s
sys     0m0.061s

Additional info if helps:

shubuntu1@shubuntu1:~$ mount.nfs -V
mount.nfs: (linux nfs-utils 1.3.3)
root@LAVA-Debug:~# mount.nfs -V
mount.nfs: (linux nfs-utils 1.3.3)

What could be the possible reason here? Where I can go to take a look?

Answer 1

Where I can go to take a look?

Use a packet capture tool (e.g. tcpdump/tshark, Wireshark/termshark) to see the connection attempts, i.e. network packets being sent from the client to the server. Look for those that are re-sent several times but never get replied to.

You might also try enabling verbose RPC call logging on the client machine's kernel (the messages will go to dmesg) using the sysctl sunrpc.rpc_debug=0xFFFF command.

What could be the possible reason here?

For NFSv3, it is likely that this will be a connection attempt either to the portmapper (rpcbind service), or to one of the dynamic ports (for NFS itself or its auxiliary mountd/statd services). NFSv4 does not have this problem because it uses only a single TCP connection.