Comment 11 for bug 585657

David McBride (david-mcbride) wrote :

Using "async" is not a viable workaround. From `man exports`:

              This option allows the NFS server to violate the NFS protocol
              and reply to requests before any changes made by that request
              have been committed to stable storage (e.g. disc drive).

              Using this option usually improves performance, but at the cost
              that an unclean server restart (i.e. a crash) can cause data to
              be lost or corrupted.

The fact that using 'async' results in higher performance is not a surprise as it is much more careful with data-handling. The fact that (according to the forum thread) enabling it happens not to trigger this particular bug is perhaps interesting from a debugging perspective, but not an acceptable solution to the problem for most organisations.

If need to make a large file for testing, `dd if=/dev/zero of=my-large-file bs=1M count=$SIZE` will make you an arbitrarily-sized file containing all-zeroes. (Other nodes in /dev may well produce more interesting output..)