#!/bin/ksh
# @(#)59	1.5  src/avs/fs/mmfs/samples/nfscluster/gpfsrecovernode.scr, mmfs, avs_rgpfs24, rgpfs240610b 2/24/06 10:48:25 
#
# This shell script is called by the mmfsd daemon during node failure
# recovery before any distributed locks held by failed nodes are released.
#
# IMPORTANT: This script is invoked synchronously, i.e., NAS recovery will
# not progress until this script finishes.  Therefore, it should not invoke
# any long-running actions, and it MUST NOT attempt to access any files or
# directories in a NAS file system.

# The script is invoked with the following parameters:
#  - The first parameter is the "recovery phase".
#  - The second parameter is the IP address used by GPFS for this node.
#  - The remaining parameters are the IP addresses of all failed nodes
#    for which recovery is being run.
#
# Note: Should additional node failures occur while node failure recovery
# is in progress, this script may be invoked more than once.

if [ ! -f /var/mmfs/etc/nfsrecovernode.scr ]; then
    echo "$0: Can't find nfsrecovernode.scr in /var/mmfs/etc"
    exit 0
fi
if [ ! -f /var/mmfs/etc/nfsfuncs ]; then
    echo "$0: Can't find NFS functions in /var/mmfs/etc"
    exit 0
fi
. /var/mmfs/etc/nfsfuncs

allparms=$*
phase=$1
me=$2
shift 2
failed=$*

debugmsg "phase=$phase; I am $me; failed nodes are $failed"

# Start grace period before GPFS recovery can complete.
if [ $phase -eq 1 ]; then
    startGrace
    msg "NLM grace period started due to failure of $failed"
fi

# Do the rest asynchronously allowing GPFS recovery to complete so we can use
# the shared filesystem to keep state.
nohup /var/mmfs/etc/nfsrecovernode.scr $allparms >> $LOGFILE 2>&1 &