pnfs-block-server.rst (1802B)
1=================================== 2pNFS block layout server user guide 3=================================== 4 5The Linux NFS server now supports the pNFS block layout extension. In this 6case the NFS server acts as Metadata Server (MDS) for pNFS, which in addition 7to handling all the metadata access to the NFS export also hands out layouts 8to the clients to directly access the underlying block devices that are 9shared with the client. 10 11To use pNFS block layouts with the Linux NFS server the exported file 12system needs to support the pNFS block layouts (currently just XFS), and the 13file system must sit on shared storage (typically iSCSI) that is accessible 14to the clients in addition to the MDS. As of now the file system needs to 15sit directly on the exported volume, striping or concatenation of 16volumes on the MDS and clients is not supported yet. 17 18On the server, pNFS block volume support is automatically if the file system 19support it. On the client make sure the kernel has the CONFIG_PNFS_BLOCK 20option enabled, the blkmapd daemon from nfs-utils is running, and the 21file system is mounted using the NFSv4.1 protocol version (mount -o vers=4.1). 22 23If the nfsd server needs to fence a non-responding client it calls 24/sbin/nfsd-recall-failed with the first argument set to the IP address of 25the client, and the second argument set to the device node without the /dev 26prefix for the file system to be fenced. Below is an example file that shows 27how to translate the device into a serial number from SCSI EVPD 0x80:: 28 29 cat > /sbin/nfsd-recall-failed << EOF 30 31.. code-block:: sh 32 33 #!/bin/sh 34 35 CLIENT="$1" 36 DEV="/dev/$2" 37 EVPD=`sg_inq --page=0x80 ${DEV} | \ 38 grep "Unit serial number:" | \ 39 awk -F ': ' '{print $2}'` 40 41 echo "fencing client ${CLIENT} serial ${EVPD}" >> /var/log/pnfsd-fence.log 42 EOF