When mount.lustre is doing DNS name resolution (e.g. "mount -t lustre mgsnode@tcp:/testfs /mnt/testfs") the mgsnode hostname is looked up via DNS (or other service) and may return multiple NIDs to the caller. However, mount.lustre only uses the first IP address returned.

It would be useful to use all of the NIDs returned from the DNS hostname lookup to locate the MGS, as if they were all listed on the command-line. Then, a new mount option like "--mgs-display=MGSNAME" would be passed to the kernel for display in the /proc/mounts output, instead of listing all of the NIDs explicitly.

This will simplify mounting with many MGS replicas, while also making the "mount" and "df" output shorter and less ugly.

See LU-16738 for details.