Description of problem: When we nfs mount the path(in our case it is a volume), we provide the path starting with a slash or without a slash to the mount command, the mount is a success either way. Problem is with present nfs-ganesha implementation we need to specifically provide the path along with slash in the begining of the path, otherwise the mount operation fails. Version-Release number of selected component (if applicable): glusterfs-3.4.0.57rhs-1.el6rhs.x86_64 nfs-ganesha-2.0.0.1-5.el6rhs.x86_64 How reproducible: always Actual results: Failure result, [root@rhsauto002 cthon04]# mount -t nfs -o vers=3,proto=tcp rhsauto004.lab.eng.blr.redhat.com:dist-rep /mnt/nfs-test mount.nfs: access denied by server while mounting rhsauto004.lab.eng.blr.redhat.com:dist-rep Expected results: it should work as this, Pass result, [root@rhsauto002 cthon04]# mount -t nfs -o vers=3,proto=tcp rhsauto004.lab.eng.blr.redhat.com:/dist-rep /mnt/nfs-test [root@rhsauto002 cthon04]# Additional info:
Please test this behaviour using kernel-nfs as well. We suspect that the correct format to be given during mount is server_ip:/volume_path.
The behaviour of nfs-ganesha is same as the kernel-nfs. With kernel-nfs,we see the same error when a similar NFSv3 mount is done. On the other hand, kernel-nfs supports both ip:/volume and ip:volume kind of mounts in NFSv4.So does nfs-ganesha.Please close the bug.
Please add doc text for this known issue.
Please review the edited doc text and sign off.
The configuration file used to start nfs-ganesha has a "Tag" parameter in the export block. If the mount path doesn't start with a '/' , the tag name is used to address the export entry at mount time. To enable version 3 mounts like mount -t nfs -o vers=3 IP:vol /mnt, the tag name must be the name of the volume. For example, if the name of the volume is testvol, the tag name should be specified as, Tag = "testvol";
The 'Tag' parameter mentioned in the above notes is currently being used for a different purpose. Also even with the 'Tag' configured same as the volume name, sub-directory mounts will not work if tried to mount using "IP:volume_name" format. So we plan to adhere to kernel-nfs behavior in this case and mandate clients to mount using "/" leading character while using Ganesha. This needs to be documented. Closing the bug. Please re-open if this needs to be addressed in any of the future release.