[Linux-ha-dev] Software RAID HA Resource Script

[Eric Z. Ayers]

Hello Alan, et al.

I've been playing with Software RAID in a high availability setup and
thought I would contribute this example RAID script back to the
heartbeat project.  I'm currently working with kimberlite,
(http://oss.missioncriticallinux.com) but I mocked up this script as a
proof of concept.  

I've got some STONITH hardware that works with DB-9 serial ports
coming in "Real Soon Now" they tell me that I will probably contribute
patches for as well:  


Western Telematics RPC10 (do you support that already? http://www.wti.com/)
Nightware RPC100S (http://www.nightware.com/) 
An 'EMB port' on an industrial PC platform (don't have a URL)

-Eric.

#!/bin/bash
#
# Example Linux RAID high availability RAID resource script
#
# Author:  Eric Z. Ayers <eric.ayers@compgen.com>
#
# License: Copyright 2000, Computer Generation Incorporated.
#          You are free to use and modify this file.  You are also free 
#          to redistribute this file and any changes you make.  
#
#          You are encouraged to submit modifications/updates to  
#          linux-ha-dev@lists.tummy.com 
#
# This script is an example start/stop script for a 
# high availability RAID resource. 
# 
# Note, I hardcoded config variables at the top of the script. It might
# be more "elegant" to pass these in as parameters, but that was too much
# like hard work for me.
#
# This script assumes you are running linux kernel 2.2.X with
# the RAID MD driver version 0.90 or better for help with linux Software 
# RAID, see the Software RAID howto or try the Linux RAID mailing list: 
# linux-raid@vger.kernel.org
#
# You might need to pass the command line parameter: raid=noautodetect 
# in an HA environment so that the kernel doesn't automatically start
# up your raid partitions when you boot the node.  This means that it isn't
# going to work to use RAID for the system disks and the shared disks.
#
# 0) partition the disks to use for RAID
# 1) Create /etc/raidtab.md?  on both systems (see example file below)
# 2) Initialize your raid partition with 
#     /sbin/mkraid --configfile /etc/raidtab.md? /dev/md?
# 3) Format your filesystem
#     mke2fs /dev/md0
# 3) Create the mount point on both systems.
#     DO NOT add your raid filesystem to /etc/fstab
# 4) copy this script (to /etc/rc.d/init.d if you wish) and edit it to
#    reflect your desired settings.
# 5) Modify the heartbeat 'haresources' setup file
# 6) unmount the filesystem and stop the raid device with 'raidstop'
# 7) fire up heartbeat!
#
# 
# EXAMPLE config file /etc/raidtab.md0
# This file must exist on both machines!
#
#  raiddev		    /dev/md0
#  raid-level		    1
#  nr-raid-disks		    2
#  chunk-size		    64k
#  persistent-superblock	    1
#  #nr-spare-disks	    0
#    device	    /dev/sda1
#    raid-disk     0
#    device	    /dev/sdb1
#    raid-disk     1
#


MD=md0
MOUNTPOINT=/data1 # where the filesystem should be mounted
SCSI_MODULE=0     # set to 1 if there is a SCSI module to be loaded
RAID0_MODULE=0    # set to 1 if there is a RAID0 module to be loaded
RAID1_MODULE=1    # set to 1 if there is a RAID1 module to be loaded
MDDEV=/dev/$MD    # Device name for RAID block device


# Utilities used by this script
MODPROBE=/sbin/modprobe
FSCK=/sbin/fsck
FUSER=/sbin/fuser
RAIDSTART=/sbin/raidstart
MOUNT=/bin/mount
UMOUNT=/bin/umount
RAIDSTOP=/sbin/raidstop


function check_util {
    if [ ! -x "$1" ] ; then
	echo "ERROR: setup problem: Couldn't find utility $1"
	exit 1
    fi
}

check_util $MODPROBE
check_util $FSCK
check_util $FUSER
check_util $RAIDSTART
check_util $MOUNT
check_util $UMOUNT
check_util $RAIDSTOP

# Look for the 'start' or 'stop' argument
case "$1" in

#
# Start up the RAID device and mount the filesystem
#
start)

$MOUNT | grep -e "^$MDDEV" >/dev/null
if [ $? -ne 1 ] ; then
    echo "ERROR: Device $MDDEV is already mounted!"
    exit 1;
fi


# Insert SCSI module
if [ "$SCSI_MODULE" -gt 0 ] ; then
    $MODPROBE scsi_hostadapter
    if [ $? -ne 0 ] ; then
	    echo "ERROR: Couldn't insert SCSI module."
	    exit 1;
    fi
fi

# Insert raid personality modules
if [ "$RAID0_MODULE" -gt 0 ] ; then
    $MODPROBE raid0
    if [ $? -ne 0 ] ; then
	    echo "ERROR: Couldn't insert RAID0 module"
	    exit 1
    fi
fi
if [ "$RAID1_MODULE" -gt 0 ] ; then
    $MODPROBE raid1
    if [ $? -ne 0 ] ; then
	    echo "ERROR: Couldn't insert RAID1 module"
	    exit 1
    fi
fi

# Run raidstart to start up the RAID array
$RAIDSTART --configfile /etc/raidtab.$MD $MDDEV
if [ $? -ne 0 ] ; then
	echo "ERROR: Couldn't start RAID for $MDDEV"
	exit 1
fi

# Check the filesystem & auto repair
$FSCK -a $MDDEV

# NOTE: if any errors at all are detected, it returns non-zero
# if the error is >4 then there is a big problem
if [ $? -ge 4 ] ; then
	echo "ERROR: Couldn't sucessfully fsck filesystem for $MDDEV"
	exit 1
fi

# Mount the filesystem
$MOUNT $MDDEV $MOUNTPOINT
if [ $? -ne 0 ] ; then
	echo "ERROR: Couldn't mount filesystem for $MDDEV on $MOUNTPOINT"
	exit 1
fi

;;

stop)

# See if the MD device is mounted
$MOUNT | grep -e "^$MDDEV" >/dev/null
if [ $? -ne 1 ] ; then

	# Kill all processes open on filesystem
	$FUSER -mk $MOUNTPOINT

	# the return from fuser doesn't tell us much
	#if [ $? -ne 0 ] ; then
	#	echo "ERROR: Couldn't kill processes on $MOUNTPOINT"
	#	exit 1;
	#fi

	# Unmount the filesystem
	$UMOUNT $MDDEV
	if [ $? -ne 0 ] ; then
		echo "ERROR: Couldn't unmount filesystem for $MDDEV"
		exit 1
	fi
else
	echo "WARNING: Filesystem $MOUNTPOINT not mounted?"
fi


# Turn off raid
$RAIDSTOP --configfile /etc/raidtab.$MD $MDDEV
if [ $? -ne 0 ] ; then
	echo "ERROR: Couldn't stop RAID for $MDDEV"
	exit 1
fi

;;

*)
    echo "This script should be run with the argument 'start' or 'stop'"
    exit 1
;;
esac

# If you got to this point, chances are everything is O.K.
exit 0;


# Revision History
# $Log: ha_raid1.sh,v $
# Revision 1.1  2000/08/24 16:25:09  eric
# Initial revision
#
#