Chapter 4 PFS Operations

This chapter applies only if you are implementing Sun PFS file systems.

A PFS I/O daemon must be running on each PFS I/O server, and a PFS proxy daemon must be running on each node that will access PFS file systems. The procedures for starting and stopping these daemons are described in the sections "Starting PFS I/O Daemons" through "Stopping PFS Proxy Daemons". The section "PFS Node or I/O Daemon Failures" discusses PFS I/O failures.

Starting PFS I/O Daemons

PFS requires an I/O daemon to be running on each I/O server node. These daemons start automatically when node(s) on which they reside boot. However, if you want to start an I/O daemon on a newly created I/O server without rebooting, start the daemon manually on that node.

# /etc/init.d/sunhpc.pfs_server start

Once the daemons are running, execute the mkfs and mount commands in each parallel file system. For example, if cities is to be a 512-Mbyte file system.

# mkfs cities 512M# /opt/SUNhpc/bin/pfsmount cities

Alternatively, after executing the mkfs command, do the following on every node in the cluster.

# mount cities

The parallel file system cities is now ready to use.

Once mkfs has been run on a file system, You should not attempt to change its configuration attributes in the CRE database. Doing so could result in the loss of file system contents.

Starting PFS Proxy Daemons

PFS also requires a proxy daemon to be running on each client node. Like the I/O daemons running on the I/O servers, PFS proxy daemons start automatically. If you want to start them manually (to avoid rebooting I/O client nodes), do the following on each node that requires it.

# /etc/init.d/sunhpc.pfs_client start

Alternatively, you can manually start both the PFS client and server daemons on every node in the cluster by executing the following command on any one node in the cluster.

# /opt/SUNWhpc/etc/pfs/pfsstart

Stopping PFS I/O Daemons

Stopping PFS is a two-step process.

Unmount each PFS file system. For example, to unmount the PFS file system cities, enter

# /opt/SUNWhpc/bin/pfsumount cities

on any node in the cluster. Alternatively, enter

# umount cities

on each client node.

Once you've unmounted the PFS file systems of interest, stop the I/O daemons on each I/O server. You do this by running the sunhpc.pfs_server stop command on each I/O server node. For example:

# /etc/init.d/sunhpc.pfs_server stop

PFS file systems are automatically unmounted and PFS I/O daemons are automatically stopped when a node is shut down or rebooted.

Stopping PFS Proxy Daemons

Shutting down client proxy daemons is similar to the proxy daemon startup process described in "Starting PFS Proxy Daemons". For example, to manually stop a PFS proxy daemon a individual nodes, enter

# /etc/init.d/sunhpc.pfs_client stop

Alternatively, you can manually stop both the PFS client and I/O server daemons on every node in the cluster by executing the following command on any one node in the cluster.

# /opt/SUNWhpc/etc/pfs/pfsstop

PFS Node or I/O Daemon Failures

If a PFS I/O daemon or PFS I/O node crashes, data stored in local memory buffers will be lost. This lost data may leave any mounted file systems in an inconsistent state. In such cases, use the fsck utility to restore file system consistency so they are safe to use.