Chapter 2. Mount options 29
To list all open files associated with Internet connections (Figure 2-9):
lsof -i
Figure 2-9 An example of Linux utility lsof -i
The umount command accepts filesystem specific options via the -o flag. The NFS client does
not have any special options.
2.5 Mount option examples
We provide the following examples as a basis for beginning your experimentation. Start with
an example that closely matches your scenario, then thoroughly test the performance and
reliability of your application while refining the mount options you have selected.
򐂰 On older Linux systems, if you do not specify any mount options, the Linux mount
command (or the automounter) automatically chooses these defaults:
mount –o rw,fg,vers=2,udp,rsize=4096,wsize=4096,hard,intr,
timeo=7,retrans=5
These default settings are designed to make NFS work right out of the box in most
environments.
Almost every NFS server supports NFS version 2 over UDP.
Rsize and wsize are relatively small because some network environments fragment
large UDP packets, which can hurt performance if there is a chance that fragments can
be lost.
The RPC retransmit timeout is set to 0.7 seconds by default to accommodate slow
servers and networks.
On clean single-speed networks, these settings are unnecessarily conservative. Over some
firewalls, UDP packets larger than 1536 are not supported, so these settings do not work. On
congested networks, UDP may have difficulty recovering from large numbers of dropped
packets. NFS version 2 write performance is usually slower than NFS version 3. As you can
see, there are many opportunities to do better than the default mount options.
Note: The NFS client allows these methods to interrupt (kill) pending RPC requests only if
the mount option intr is set for that filesystem.

Get Using the Linux NFS Client with IBM System Storage N series now with O’Reilly online learning.

O’Reilly members experience live online training, plus books, videos, and digital content from 200+ publishers.