Why would you use OCFS
======================
OCFS is designed to provide an alternative to using raw devices for Oracle
9iRAC. Managing raw devices is usually a difficult task. Also many DBA's
and system administrators are familiar with filesystems. Another issues
with raw devices on linux, is the fact that you are limited to only
255. There cannot be more than 255 /dev/raw entries.

Type of files supported for OCFS
================================
At this time OCFS only supports Oracle Datafiles.  This includes Redo
Log files, Archive log files, Controlfiles and database Datafiles.
The shared quorum disk file for the cluster manager and the shared init
file (srv) are also supported.

In the later part of 2003 we will also support Shared Oracle Home installs.

OCFS blocksize
==============
Format with 128kb blocksize is optimal size. Sizes between 2kb and 1mb are
supported, the smaller blocksizes will have a performance penalty, but
will be useful for the future when we will support regular files. 128kb
blocksize means that every file created with content uses up a minimum
of 128kb space on disk. Even if there is only 1 byte of data in the file.

Splitting up files on partitions
================================
It is recommended to have the archive log files on a seperate disk
partition and if possible each node should have it's own archive log
partition. You can/should mount all these on every node in the cluster
but only one node will be writing out archive log files to it's assigned
partition. This reduces contention on space allocation if the database
is heavily used and lots of archive logs are created on each node.

All datafiles can reside on the same partition. An Oracle blocksize of
8kb is usually recommended and on OCFS this does not change. Performance
of 8kb oracle blocks will be somewhat better. It is recommended to have
an extent size that is the same on every tablespace in order to prevent
disk fragmentation.  OCFS will require contigious space on disk for
every oracle or for the initial datafile creation. Eg. if you create
a 1gb datafile in 1 command, it will require 1gb contiguous space on
disk. If you then extend this datafile with 100mb, it will need another
100mb chunk on disk. This 100 mb chunk does not have to fall right behind
the 1gb part.

Partition size
==============
OCFS will support partition sizes up to 1TB (tested) configurations. Since
there is no volume management built into OCFS it is recommended to have
hardware raid support and create a logical disk volume that is large
enough. If this is not possible, depending on the linux distribution
you run, lvm or md can be used.

Note that having many mountpoints on a system will create a performance
bottleneck, this means more than 50 or so partitions. (this is not
related to ocfs specifically)

Support for mount by label. If you create your ocfs volumes with a unique
ocfs LABEL, mount will support this. RHAS2.1's mount errata release has
updates for ocfs and it is possible to simply do  mount -t ocfs -L mylabel
/ocfs. This will be of great help if the device naming is variable.

Performance
===========
make sure updatedb is not running on the OCFS partitions.
eg add ocfs to PRUNEFS= in /etc/updatedb.conf

Troubleshooting
===============
OCFS will print debug and error information into the system log,
/var/log/messages or dmesg will show ocfs related errors.  If a problem
is suspect to be related to ocfs, a few things can be checked quickly.
type   'dmesg' and see if there are messages starting with ocfs.  run 'ps
aux' and find out which process is hanging, (D state). if this does not
change, strace -p <pid> will show if there is any process activity. Please
also provide this information to Oracle Support.

Sample configuration and layout
===============================
Install the ocfs-kernel module, depending on your kernel
	uname -a will provide you with the current running kernel

Install the UI tool and the format tools.
ocfs-tools-XX.rpm and ocfs-support-XX.rpm

Sample filesystem layout (same mountpoint on every node in the RAC cluster)
/ocfs-data
/ocfs-index
/ocfs-ctrl-redo
/ocfs-ctrl-redo2
/archive1
/archive2
/archive3
/archive4

----------------

if you want to automatically load ocfs correctly with filesystems in
/etc/fstab, please put the following line in /etc/modules.conf

install ocfs load_ocfs

and in /etc/fstab for every ocfs line add _netdev in the 4th field
(usually says defaults)
eg 
/dev/sda1	/ocfs	ocfs	_netdev	0 0 

updatedb runs on ocfs by default
disable this by putting ocfs in the PRUNEFS line in /etc/updatedb.conf