source: gpfs_3.1_ker2.6.20/lpp/mmfs/samples/nfscluster/nfsfuncs

Last change on this file was 16, checked in by rock, 17 years ago
  • Property svn:executable set to *
File size: 37.2 KB
Line 
1#!/bin/ksh
2# @(#)64  1.17.1.18  src/avs/fs/mmfs/samples/nfscluster/nfsfuncs, mmfs, avs_rgpfs24, rgpfs24s011a 3/7/07 20:57:04
3#
4
5if [ ! -f /var/mmfs/etc/nfsdefs ]; then
6    echo "$0: Can't find NFS defines(nfsdefs) in /var/mmfs/etc"
7    exit 0
8fi
9
10. /var/mmfs/etc/nfsdefs
11
12# Configure path used for programs
13PATH=$PATH:/sbin:/usr/sbin:/usr/bin:/bin                # standard programs
14PATH=$PATH:/usr/lpp/mmfs/bin:/var/mmfs/etc    # for mmfs programs
15PATH=$PATH:/etc/ha.d/resource.d:/usr/lib/heartbeat  # for HA stuff
16export PATH
17
18# Shared storage for HA-NFS
19# The directory structure is as follows:
20# shared/.ha
21#     nfs (mirrors /var/lib/nfs)
22#         rmtab
23#         node1
24#       statd
25#           sm
26#           sm.bak
27#   ...
28#     recovery
29#   node1
30#       nodeX
31#   ...
32#
33# Local storage for node1
34# /var/lib/nfs
35# rmtab  -> shared/.ha/nfs/rmtab
36# (RHEL):
37# statd  -> shared/.ha/nfs/node1/statd
38# (SLES):
39# sm     -> shared/.ha/nfs/node1/statd/sm
40# sm.bak -> shared/.ha/nfs/node1/statd/sm.bak
41#
42SHARED_HA=${SHARED_ROOT}/.ha
43SHARED_NFS=${SHARED_HA}/nfs
44SHARED_RECOVERY=${SHARED_HA}/recovery
45
46# For failover of locks to happen correctly, the lockmgrs on all GPFS nodes
47# need to listen on different port numbers so clients are forced to
48# re-establish socket connection with the takeover node for reclaims
49# Default port number to use for NLM (NFS lock manager)
50# For node i, nlmport = NLM_PORT + i
51[ -z "$NLM_PORT" ] && NLM_PORT=10000
52
53# Dependencies for distribution - RHEL, SLES
54if [ -d /etc/sysconfig/network ]; then
55    IFPATH="/etc/sysconfig/network"
56else
57    IFPATH="/etc/sysconfig/network-scripts"
58fi
59
60typeset -i  iptakeover=0  # does GPFS perform IP failover?
61    debug=0   # debug level for messages to be logged?
62    notifyfix=0 # SM_NOTIFY fix required for SLES?
63    monitor=1 # monitoring of daemons required?
64    customLog=1 # Log file specified
65    hardMount=1 # Clients use "hard" NFS mounts
66
67[ -f $NODELIST ]     && iptakeover=1
68[ -n "$DEBUG" ]      && debug=$DEBUG
69[ -n "$NOTIFYFIX" ]  && notifyfix=1
70[ -n "$MONITOR" ]    && monitor=$MONITOR
71[ -z "$LOGFILE" ]    && LOGFILE=/var/mmfs/gen/mmfslog && customLog=0
72[ -z "$NFSD_PROCS" ] && NFSD_PROCS=32
73
74if [ -n "$GPFS_RSH" ]; then
75    GPFS_rshPath=$(which $GPFS_RSH)
76else
77    GPFS_rshPath=$(which rsh)
78fi
79export GPFS_rshPath
80
81################################################################################
82#       Utility functions            #
83################################################################################
84
85die() {
86    echo "$*"
87    exit 1
88}
89
90_log() {
91    let level=$1; shift
92    if [ $debug -ge $level ]; then
93  echo "`date`: $*" >> $LOGFILE 2>&1
94  $* >> $LOGFILE 2>&1
95    else
96  $* > /dev/null 2>&1
97    fi
98}
99
100log() {
101    _log 0 $*
102}
103
104debuglog() {
105    _log 1 $*
106}
107
108debuglog2() {
109    _log 2 $*
110}
111
112_msg() {
113    level=$1; shift
114    [ $debug -ge $level ] && echo "`date`: $*" >> $LOGFILE 2>&1
115}
116
117msg() {
118    _msg 0 $*
119    logger -t HA-NFS "$*"
120}
121
122debugmsg() {
123    _msg 1 $*
124}
125
126debugmsg2() {
127    _msg 2 $*
128}
129
130err() {
131    msg "Error: $*"
132}
133
134warn() {
135    msg "Warning: $*"
136}
137
138_mkdir() {
139    debuglog2 mkdir -m 0700 -p $*
140}
141
142_rmdir() {
143    debuglog2 rm -rf $*
144}
145
146_unlink() {
147    debuglog2 unlink $1
148}
149
150_cp() {
151    debuglog2 cp -dpf $*
152}
153
154_mv() {
155    debuglog2 mv -f $*
156}
157
158# Skip blank and comment lines (Ugh!)
159invalid() {
160    line=$1
161    line=${line## }
162    [[ "$line" = "" || "${line#\#}" != "$line" ]] && return 0
163    return 1
164}
165
166# Return distribution
167getDistro() {
168    if grep -q "SUSE LINUX Enterprise Server 9" /etc/issue; then
169        echo "SLES_8"
170    elif grep -q "SUSE SLES 8" /etc/issue; then
171        echo "SLES_9"
172    elif grep -q "Fedora\|Red Hat" /etc/issue; then
173        echo "RH"
174    else
175        echo ""
176    fi
177}
178
179# Save old log file - use last modified time
180rotatelog() {
181    if [[ $customLog -eq 1 && -f $LOGFILE ]]; then
182  ext=$(stat -c "%y" $LOGFILE)
183  ext=${ext%.*} # get date and time
184  #ext=${ext// /.}  # replace space with . - doesn't work with pdksh
185  ext=$(echo $ext | sed 's/ /./g')  # replace space with .
186  mv $LOGFILE $LOGFILE.$ext
187    fi
188}
189
190################################################################################
191#       Network functions            #
192################################################################################
193
194# Get IP address from hostname: use /etc/hosts first
195ipaddr() {
196    ip=$(grep -w "${1}" /etc/hosts | grep -v ^# | awk '{print $1}')
197    if [ -z "$ip" ]; then
198  line=$(host -n $1 | grep 'has address' | awk '{print $4}')
199    fi
200    if [ -n "$ip" ]; then
201  echo $ip
202    else
203  echo $1
204    fi
205}
206
207# Get host name from IP address
208ipname() {
209    name=$(grep -w $1 /etc/hosts | grep -v ^# | awk '{print $2}')
210    if [ -z "$name" ]; then
211  name=$(host -n $1 | grep -v 'not found:')
212        name=${name##* }  # Last word is the host name
213        name=${name%%.}   # Strip trailing dot
214    fi
215    echo $name
216}
217
218shortipname() {
219    name=$(ipname $1)
220    echo $name | awk -F. '{print $1}'
221}
222
223# Get matching subnet given two IP addresses
224getsubnet() {
225    ip1=$1
226    ip2=$2
227    mask=$3
228
229    typeset -i i1 i2 m1
230    IFS=.
231    set $ip1
232    i1=$((($1<<24)+($2<<16)+($3<<8)+$4)) # comment to fix hilit
233
234    set $ip2
235    i2=$((($1<<24)+($2<<16)+($3<<8)+$4)) # comment to fix hilit
236
237    set $mask
238    m1=$((($1<<24)+($2<<16)+($3<<8)+$4)) # comment to fix hilit
239
240    if [[ $((i1&m1)) == $((i2&m1)) ]]; then
241  echo $((i1&m1))
242    fi
243}
244
245# Get configuration file for a given IP address from IFPATH
246# Note: Only the first file that matches the given IP is returned
247getifcfg() {
248    echo $(grep -lw "^IPADDR.*='$1'" ${IFPATH}/ifcfg-* 2> /dev/null | head -n1)
249}
250
251# Get interface name given its IP address
252getifname() {
253    iface=$(getifcfg $1)
254    iface=${iface##*/}  # Strip path
255    iface=${iface#*-} # Strip ifcfg-
256    [ -n "$iface" ] && iface=$(getcfg-interface -- $iface) # FIX: SLES only
257    echo $iface
258}
259
260# Bring up interface corresponding to a given IP address
261ifUp() {
262    # Check if it is already configured and up
263    #if [ -n "$(ifconfig | grep -wo $1)" ]; then
264        iface=$(getifname $1)
265        debuglog ifup $iface
266    #fi
267
268    # Send an arp to the default gateway just in case...
269    gwIP=$(route -n | awk '/UG/ {print $2}')
270    iface=$(mmgetifconf | grep $1 | awk '{print $1}')
271    if [ -n "$iface" ]; then
272        for ip in $gwIP; do
273            arping -q -c 5 -s $1 -I $iface $ip
274        done
275    fi
276}
277
278ifDown() {
279    eth=$(mmgetifconf | grep -w $1 | awk '{print $1}')
280    debugmsg "Invoking ifdown on $eth for ip $1"
281
282    if [ -n "$eth" ]; then
283  # FIX: SuSE only;
284  # ifdown won't work on RedHat for an interface like eth0:0
285  debuglog ifdown $eth
286    fi
287}
288
289# Bring up "bond" interface
290ifBondUp() {
291    iface=$1
292    debuglog modprobe bonding $BONDING_MODULE_OPTS
293    debuglog ifconfig $iface up
294
295    # Get all slave interfaces from hardware descriptions
296    BSINTERFACES=""
297    for i in $(set | egrep "^BONDING_SLAVE") ; do
298  BONDING_SLAVE=${i##*=}
299  [ -z "$BONDING_SLAVE" ] && continue
300  BSIFACE=$(getcfg-interface -- $BONDING_SLAVE) # FIX: SLES only
301  if [ $? != 0 ] ; then
302      debugmsg "Could not get an interface for slave"
303      continue
304  fi
305        # prepare only available slave devices
306  if [ -d /sys/class/net/$BSIFACE ] ; then
307      BSINTERFACES="$BSINTERFACES $BSIFACE"
308  else
309      debugmsg "Bonding Slave $BSIFACE is not available. Skipped"
310  fi
311    done
312    # enslave the slave ifaces only once
313    if [ -n "$BSINTERFACES" ]; then
314  debuglog ifenslave $iface $BSINTERFACES
315    fi
316}
317
318# Check if a given IP address is an alias (virtual)
319isVirtualIP() {
320    grep -qlw "^IPADDR..*='$1'" ${IFPATH}/ifcfg-* 2> /dev/null
321    return $?
322}
323
324################################################################################
325#     Nodes list functions               #
326################################################################################
327
328# Extract GPFS IP, iface and netmask from nodes file with the format:
329# GPFS_IP[:eth:mask] NFS_IP1 NFS_IP2 ...
330getip() {
331    echo $1 | awk -F: '{print $1}'
332}
333
334getiface() {
335    echo $1 | awk -F: '{print $2}'
336}
337
338getnetmask() {
339    echo $1 | awk -F: '{print $3}'
340}
341
342# Get all NFS IP addresses from nodes file
343getAllNfsIPs() {
344    exec 3< $NODELIST
345    while read -u3 gpfs_if nfs_list; do
346        # Skip empty and comment lines
347        if invalid $gpfs_if; then 
348            continue
349        fi
350  echo $nfs_list
351    done
352}
353
354# Get NFS IP addresses for a given GPFS IP address from nodes file
355getNfsIPs() {
356    exec 3< $NODELIST
357    while read -u3 gpfs_if nfs_list; do
358        # Skip empty and comment lines
359        if invalid $gpfs_if; then 
360            continue
361        fi
362  gpfs_ip=$(getip $gpfs_if)
363  if [ "$gpfs_ip" == "$1" ]; then
364      debugmsg2 "getNfsIPs: $gpfs_ip $nfs_list"
365      echo $nfs_list
366      break
367  fi
368    done
369}
370
371# Get interface for a given NFS+GPFS address
372getEth() {
373    gpfs_ip=$1
374    nfs_ip=$2
375    eth=""
376
377    debugmsg "getEth: gpfs_ip $gpfs_ip nfs_ip $nfs_ip"
378
379    # First try to get eth from node list
380    exec 3< $NODELIST
381    while read -u3 gpfs_if nfs_list; do
382        # Skip empty and comment lines
383        if invalid $gpfs_if; then 
384            continue
385        fi
386  if [ "$(getip $gpfs_if)" == "$gpfs_ip" ]; then
387      iface=$(getiface $gpfs_if)
388      if [ -n "$iface" ]; then
389    debugmsg "getEth: from $NODELIST $gpfs_ip interface $iface"
390    echo $iface
391    return
392      fi
393  fi
394    done
395
396    # Now try to get eth from list of interfaces
397    tmp=/tmp/mmgetifconf.$$
398    mmgetifconf > $tmp
399    exec 3< $tmp
400    while read -u3 iface ip mask; do
401  subnet=$(getsubnet $ip $nfs_ip $mask)
402  if [ -n "$subnet" ]; then
403      debugmsg "getEth: from mmgetifconf $nfs_ip interface $iface"
404      echo $iface
405      return
406  fi
407    done
408    unlink $tmp
409
410    msg "getEth: not found $gpfs_ip interface"
411    echo ""
412}
413
414# Get netmask for a given NFS+GPFS IP address
415getmask() {
416    gpfs_ip=$1
417    nfs_ip=$2
418
419    # First try to get mask from node list
420    exec 3< $NODELIST
421    while read -u3 gpfs_if nfs_list; do
422        # Skip empty and comment lines
423        if invalid $gpfs_if; then 
424            continue
425        fi
426  if [ "$(getip $gpfs_if)" == "$gpfs_ip" ]; then
427      mask=$(getnetmask $gpfs_if)
428      if [ -n "$mask" ]; then
429    debugmsg "getmask: from $NODELIST $gpfs_ip netmask $mask"
430    echo $mask
431    return
432      fi
433  fi
434    done
435
436    # Now try from list of interfaces
437    #sub_ip=$(echo $nfs_ip | cut -d . -f1,2,3)
438    sub_ip=${nfs_ip%.*}
439    mask=$(mmgetifconf | grep -m1 $sub_ip | awk '{print $3}')
440    if [ -n "$mask" ]; then
441  debugmsg "getmask: from get_ifconf $gpfs_ip netmask $mask"
442  echo $mask
443  return
444    fi
445
446    msg "getmask: not found $gpfs_ip netmask default"
447    echo "255.255.255.0"
448}
449
450# return the next node for a given node and a start point.
451getNextNode() {
452  given=$1
453  start=$2
454  typeset -i next n i=0
455  set -A gpfsIPs $(getAllGPFSIPs)
456  n=${#gpfsIPs[@]}
457 
458  while [ i -lt $n ]; do
459    [ "${gpfsIPs[$i]}" == "$given" ] && break
460    i=$i+1
461  done
462  [ $i -eq $n ] && return # not found
463  next=$i+1
464  [ $next -eq $n ] && next=0
465  [ "${gpfsIPs[$next]}" == "$start" ] && return
466  echo ${gpfsIPs[$next]}
467}
468
469
470################################################################################
471#       GPFS functions               #
472################################################################################
473
474# Get GPFS IP addresses from nodes file
475getAllGPFSIPs() {
476    exec 3< $NODELIST
477    while read -u3 gpfs_if nfs_list; do
478        # Skip empty and comment lines
479        if invalid $gpfs_if; then 
480            continue
481        fi
482  debugmsg2 "getAllGPFSIPs: $gpfs_if"
483  echo $(getip $gpfs_if)
484    done
485}
486
487# Get current node's GPFS IP address
488myGPFSIP() {
489    for ip in $(getAllGPFSIPs); do
490        my_ip=$(mmgetifconf | grep -w $ip)
491        if [ "$my_ip" != "" ]; then
492            echo $ip
493            break
494        fi
495    done
496}
497
498isGpfsFS() {
499    # Get list of GPFS filesystems from /etc/fstab
500    fses1="$(grep -w gpfs /etc/fstab | awk '{print $2}')"
501    exp1=$1
502    for fs1 in $fses1; do
503        debugmsg isGpfsFS: exp=$exp1 fs=$fs1
504        if [ "${exp1##$fs1}" != "$exp1" ]; then
505            debugmsg isGpfsFS: return 0
506            return 0
507        fi
508    done
509    debugmsg isGpfsFS: $1 return 1
510    return 1
511}
512
513# Mount GPFS filesystems that are to be NFS-exported
514mountExportedFS() {
515    # Get list of GPFS filesystems from /etc/fstab
516    fses="$(grep -w gpfs /etc/fstab | awk '{print $2}')"
517    # Get list of GPFS exports from /etc/exports
518    exports="$(awk '{print $1}' /etc/exports | grep ^/ | sort | uniq)"
519    for exp in $exports; do
520        for fs in $fses; do
521            if [ "${exp##$fs}" != "$exp" ]; then
522                debuglog mount $fs
523            fi
524        done
525    done
526}
527
528isExported() {
529    # Get list of GPFS filesystems from /etc/mtab
530    fs="$(grep -w "gpfs .*dev=$1" /etc/mtab | awk '{print $2}')"
531    debugmsg isExported: $fs
532    # Get list of GPFS exports from /etc/exports
533    exports="$(awk '{print $1}' /etc/exports | grep ^/ | sort | uniq)"
534    for exp in $exports; do
535        if [ "${exp##$fs}" != "$exp" ]; then
536            debugmsg isExported: $fs return 0
537            return 0
538        fi
539    done
540    debugmsg isExported: $fs return 1
541    return 1
542}
543
544isAnyExported() {
545    mountExportedFS
546    # Get list of GPFS filesystems from /etc/fstab
547    fses="$(grep " gpfs .*dev=" /etc/mtab | awk '{print $2}')"
548    debugmsg isAnyExported: $fses
549    # Get list of GPFS exports from /etc/exports
550    exports="$(awk '{print $1}' /etc/exports | grep ^/ | sort | uniq)"
551    for exp in $exports; do
552        for fs in $fses; do
553            debugmsg isAnyExported: exp=$exp fs=$fs
554            if [ "${exp##$fs}" != "$exp" ]; then
555                debugmsg isAnyExported: return 0
556                return 0
557            fi
558        done
559        isGpfsFS $exp
560        rc=$?
561        # export if not gpfs
562        if [ $rc -ne 0 ]; then
563                debugmsg isAnyExported: $exp is not GPFS return 0
564                return 0
565        fi
566    done
567    debugmsg isAnyExported: return 1
568    return 1
569}
570
571isMounted() {
572    # Get list of GPFS filesystems from /etc/mtab
573    fses="$(grep " gpfs .*dev=" /etc/mtab | awk '{print $2}')"
574    debugmsg isMounted: $fses
575    # Get list of GPFS exports from /etc/exports
576    exp=$1
577    for fs in $fses; do
578        if [ "${exp##$fs}" != "$exp" ]; then
579            debugmsg isMounted: return 0
580            return 0
581        fi
582    done
583    debugmsg isMounted: return 1
584    return 1
585}
586
587isSharedRoot() {
588    # Get filesystem from /etc/mtab
589    fs=$(grep -w "gpfs .*dev=$1" /etc/mtab | awk '{print $2}')
590    if [ "${SHARED_ROOT##$fs}" != "$SHARED_ROOT" ]; then
591         return 0
592       fi
593    return 1
594}
595
596# Run mmdsh command
597mmdshcmd() {
598    debugmsg "mmdsh -vL $*"
599    mmdsh -vL $*
600}
601
602# Run mmdsh command and return exit code correctly
603mmdshcmdRC() {
604    debugmsg "mmdsh -vL $*"
605
606    typeset -i rc=0
607    myIP=$(myGPFSIP)
608    targets=$1
609    cmd=../../../../$2    # relative to /usr/lpp/mmfs/bin
610    shift 2
611    parms=$*
612
613    remoteVerb=hanfs
614    # FIX: Use the following from mmglobfuncs
615    tmpDir=/var/mmfs/tmp/
616    mmremote=/usr/lpp/mmfs/bin/mmremote
617    MMMODE=LC
618    NO_LINK=_NO_LINK_
619
620    rm -f $tmpDir/$remoteVerb.*
621    mmdsh -vL $targets $mmremote onbehalf2 $myIP $remoteVerb $MMMODE $NO_LINK $cmd $parms
622    rcInfo=$(ls $tmpDir$remoteVerb.* 2> /dev/null | sort -rn | head -1)
623    rm -f $tmpDir$remoteVerb.*
624    if [ -n "$rcInfo" ]; then
625        rc=${rcInfo#$tmpDir$remoteVerb\.}
626    fi
627    return $rc
628}
629
630# Run command on all GPFS nodes
631mmdshAll() {
632    gpfsIPs=$(getAllGPFSIPs)
633    gpfsIPs=$(echo $gpfsIPs | sed 's/ /,/g')
634    mmdshcmd $gpfsIPs $*
635}
636
637# stop another node
638stopNode() {
639    [ -z "$1" ] && return 1
640    cmd="/var/mmfs/etc/nfsmonitor -e && /var/mmfs/etc/stop.nfs"
641    mmdshcmd $1 "$cmd" &
642    return 0
643}
644
645# Return 0 (success) if quorum is "Active" on this node; 1 otherwise
646isNodeUp() {
647    [ -z "$1" ] && return 1
648    #status=$(tsstatus -m | grep -w "$1")
649    #status=$(mmgetstate -k -N $1 | grep -w "active")
650    #status=$(mmdshcmd $1 "/usr/lpp/mmfs/bin/mmremote mmGetState | grep -w active")
651    #debugmsg "mmgetstate $1: $status"
652    #[ -n "$status" ] && return 0
653    mmfsadm dump cfgmgr | grep -q "$1 .* up "
654    return $?
655}
656
657# Exit GPFS for given IP address is "down"
658ifGPFSDownExit() {
659    [ -n "$(pidof mmfsd)" ] && return 0
660    stop.nfs
661    mmfsadm cleanup
662    exit 1
663}
664
665################################################################################
666#     Configuration functions              #
667################################################################################
668
669# Check status of a service
670checkStatus() {
671    if [ -e /sbin/checkproc ]; then
672        opts=""
673        if [[ "$1" == "nfsd" || "$1" == "lockd" ]]; then
674            opts="-n"
675        fi
676        checkproc $opts $1 > /dev/null 2>&1
677        return $?
678    elif [ -f /etc/rc.d/init.d/functions ]; then
679        savedpath=$PATH
680        . /etc/rc.d/init.d/functions
681        PATH=$savedpath
682        status $1 > /dev/null 2>&1
683        return $?
684    else
685        return 3 
686    fi
687}
688
689# Get service for communicating between GPFS nodes
690rshService() {
691    if [ -n "$GPFS_RSHD" ]; then
692        echo "$GPFS_RSHD"
693    elif [ "$GPFS_rshPath" == "/usr/bin/rsh" ]; then
694        echo "xinetd"
695    elif [ "$GPFS_rshPath" == "/usr/bin/ssh" ]; then
696        echo "sshd"
697    else 
698        debugmsg "Unsupported service $GPFS_rshPath"
699        echo ""
700    fi
701}
702
703# Start rsh (or ssh) server for communication between GPFS nodes
704startRshd() {
705    service=$(rshService)
706    checkStatus $service
707    if [ $? -ne 0 ]; then
708        /etc/init.d/$service start
709    fi
710}
711
712# Configure NLM ports
713configNLMPorts() {
714    # Determine which port to use for NLM from the node id
715    # and ensure it is set
716    typeset -i nlmport curport nodeid=0
717    while [ $nodeid -eq 0 ]; do
718        #nodeid=$(mmgetstate -k | awk -F: '{print $2}')
719        #nodeid=$(mmlscluster | grep -w $1 | awk '{print $1}')
720        nodeid=$(mmdsm dsmGetNodeNumber)
721        [ $nodeid -eq 0 ] && warn "Cannot get nodeid for $1 from mmgetstate, retrying..."
722    done
723    nlmport=$NLM_PORT+$nodeid
724
725    # Ensure that nfsd is loaded
726    debuglog modprobe nfsd
727
728    curport=$(rpcinfo -p 2> /dev/null | grep -m1 nlockmgr | awk '{print $4}')
729    [ -z "$curport" ] && curport=$(sysctl -n fs.nfs.nlm_tcpport)
730    if [ $curport -ne $nlmport ]; then
731  debugmsg "Current NLM port used is $curport, should be $nlmport"
732  debuglog sysctl -w fs.nfs.nlm_tcpport=$nlmport
733  debuglog sysctl -w fs.nfs.nlm_udpport=$nlmport
734  nfsService stop
735  nfsService start
736  # Check if we the port got assigned correctly
737  curport=$(rpcinfo -p 2> /dev/null | grep -m1 nlockmgr | awk '{print $4}')
738  [ -z "$curport" ] && curport=$(sysctl -n fs.nfs.nlm_tcpport)
739  if [ $curport -ne $nlmport ]; then
740      err "Cannot change existing port $curport to $nlmport for HA-NFS. Terminating..."
741      return 1
742  fi
743    fi
744    return 0
745}
746
747# Get system boot time
748getBootTime() {
749    grep -w btime /proc/stat | awk '{print $2}'
750}
751
752# Create a place to backup entries (statd/sm) different from /var/lib/nfs
753# We need this because sending SM_NOTIFY messages on failover/failback wipes out
754# client entries from /var/lib/nfs/sm and these do not get created again until
755# next reboot
756# This backup will be cleaned up on next reboot
757backupSmDir() {
758    typeset -i current_btime saved_btime
759
760    if [ -f $1/btime ]; then
761  current_btime=$(getBootTime)
762  saved_btime=$(cat $1/btime)
763  if [ $current_btime -gt $saved_btime ]; then
764      # Erase backup smdir since a reboot has happened
765      debugmsg "Erasing backup statd dirs in $1"
766      _rmdir $1/sm $1/sm.bak
767            echo $current_btime > $1/btime
768  fi
769    else
770  _mkdir $1/sm $1/sm.bak
771        # Save boot time so we can decide when to cleanup $1
772  btime=$(getBootTime)
773  debugmsg "Saving current boot time $btime in $1"
774  echo $btime > $1/btime
775    fi
776}
777
778# Keep the following data from /var/lib/nfs in shared space (GPFS)
779# so all nodes have access to it for failover/failback purposes:
780# rmtab
781# sm
782# sm.bak
783shareSmDir() {
784    myip=$1
785
786    sh_rmtab=$SHARED_NFS/rmtab
787    [ ! -f $sh_rmtab ] && touch $sh_rmtab && chmod 644 $sh_rmtab
788#   No need to share rmtab. Its no longer used to validate NFS requests.
789#   ln -sf $sh_rmtab /var/lib/nfs/rmtab
790
791    sh_statd=$SHARED_NFS/$myip/statd
792    _mkdir $sh_statd/sm $sh_statd/sm.bak
793
794    if [ -e /var/lib/nfs/statd ]; then
795  smdir=/var/lib/nfs/statd/sm
796    else
797  smdir=/var/lib/nfs/sm
798    fi
799    if [[ -d $smdir && ! -L $smdir ]]; then
800  # Move stuff from local smdir to shared
801  _mv $smdir/* $sh_statd/sm
802  _mv ${smdir}.bak/* $sh_statd/sm.bak
803  _rmdir $smdir ${smdir}.bak
804    fi
805    if [ ! -d $smdir ]; then
806  ln -sf $sh_statd/sm $smdir
807  if [ $? -ne 0 ]; then
808            err "Failed to link $smdir to $sh_statd/sm"
809            return 1
810        fi
811  ln -sf $sh_statd/sm.bak ${smdir}.bak
812  if [ $? -ne 0 ]; then
813            err "Failed to link ${smdir}.bak to $sh_statd/sm.bak"
814            return 1
815        fi
816    fi
817    if [ -e /var/lib/nfs/statd ]; then
818  # Redhat requires rpcuser as uid/gid for statd stuff
819  chown -R rpcuser.rpcuser $sh_statd
820    fi
821    return 0
822}
823
824# Remove the symlinks created by shareSmDir
825unshareSmDir() {
826    if [ -e /var/lib/nfs/statd ]; then
827  smdir=/var/lib/nfs/statd/sm
828    else
829  smdir=/var/lib/nfs/sm
830    fi
831    [ -L /var/lib/nfs/rmtab ] && _unlink /var/lib/nfs/rmtab && touch /var/lib/nfs/rmtab
832    [ -L $smdir ] && _unlink $smdir && _mkdir $smdir
833    [ -L ${smdir}.bak ] && _unlink ${smdir}.bak && _mkdir ${smdir}.bak
834}
835
836# Configure GPFS for HA-NFS - first time only
837configHA() {
838    myip=$1
839
840    # Check if this is the first time we are configuring
841    # FIX: check needed here?
842
843    # Configure NLM ports
844    # Note: This is now done by the startup script /etc/init.d/gpfs
845    # configNLMPorts $myip
846
847    # Check the shared directory is available and on GPFS
848    if [ ! -d $SHARED_ROOT ]; then 
849        err "Cannot find shared directory $SHARED_ROOT" 
850        return 1
851    fi
852    df -Tl $SHARED_ROOT | grep -qw gpfs
853    if [ $? -ne 0 ]; then 
854        err "$SHARED_ROOT found but is not on a GPFS filesystem"
855        return 1
856    fi
857    debugmsg "Shared fs is $SHARED_ROOT"
858
859    # Create shared data for HA-NFS (statd, rmtab) and recovery
860
861    _mkdir $SHARED_NFS $SHARED_RECOVERY
862
863    shareSmDir $myip
864    return $?
865}
866
867################################################################################
868#       NFS functions              #
869################################################################################
870
871startMountd() {
872    savedpath=$PATH
873    RPCMOUNTDOPTS=
874    #Unrolling mountd part of /etc/init.d/nfs
875    [ -f /etc/init.d/functions ] && . /etc/init.d/functions
876    [ -f /etc/sysconfig/network ] && . /etc/sysconfig/network
877    [ -f /etc/sysconfig/nfs ] && . /etc/sysconfig/nfs
878    PATH=$savedpath
879    [ -n "$MOUNTD_PORT" ] && RPCMOUNTDOPTS="$RPCMOUNTDOPTS -p $MOUNTD_PORT"
880    case $MOUNTD_NFS_V2 in
881  no|NO)
882      RPCMOUNTDOPTS="$RPCMOUNTDOPTS --no-nfs-version 2" ;;
883    esac
884    case $MOUNTD_NFS_V3 in
885  no|NO)
886      RPCMOUNTDOPTS="$RPCMOUNTDOPTS --no-nfs-version 3" ;;
887    esac
888    if [ -e /sbin/startproc ]; then
889  debuglog startproc /usr/sbin/rpc.mountd $RPCMOUNTDOPTS
890    else
891  daemon rpc.mountd $RPCMOUNTDOPTS
892    fi
893}
894
895nfsService() {
896    nfslock=/etc/init.d/nfslock
897    if [ -f /etc/init.d/nfsserver ]; then
898  nfs=/etc/init.d/nfsserver
899    else
900  nfs=/etc/init.d/nfs
901    fi
902
903    case $1 in
904  start)
905      msg "Starting NFS services"
906            sysctl -e -q -w fs.nfs.use_underlying_lock_ops=1
907      debuglog $nfs start
908      debuglog $nfslock start
909            if [ -n "$MOUNTD_PORT" ]; then
910                # Make sure mountd is bound to the right port if specified
911                curport=$(rpcinfo -p 2> /dev/null | grep -m1 mountd | awk '{print $4}')
912                if [ "$curport" != "$MOUNTD_PORT" ]; then
913                    pid=$(pidof rpc.mountd)
914              debugmsg "Current mountd port is $curport, should be $MOUNTD_PORT. Stopping current rpc.mountd (pid $pid) and restarting with correct port."
915                    debuglog kill -9 $pid
916                    startMountd
917          fi
918            fi
919            # Update number of nfsd processes
920            debuglog rpc.nfsd $NFSD_PROCS
921
922            # Reload exportfs anyway since starting nfs server may not do this
923      # if it was already running
924      exportfs -r
925      ;;
926
927  stop)
928      msg "Stopping NFS services"
929      debuglog $nfs stop
930      ;;
931
932  terminate)
933      msg "Cleaning NFS services"
934      debuglog $nfs stop
935      debuglog $nfslock stop
936      ;;
937
938  soft-restart)
939      debuglog rpc.nfsd 0
940            debuglog sleep 1  # FIX: required?
941            debuglog rpc.nfsd $NFSD_PROCS
942      ;;
943
944  *)
945      echo "Usage: $0 start|stop|restart"
946    esac
947}
948
949################################################################################
950#       NLM functions              #
951################################################################################
952
953# Restart rpc.statd
954# Note that we don't want to issue a "nfslock restart" directly since this would
955# involve restarting lockd (which results in lock recovery) which we don't want
956restartStatd() {
957    if [ -f /sbin/rpc.statd ]; then
958  # Kill the statd process if exists (important for registering with portmap)
959  debuglog kill -9 $(/sbin/pidof -x /sbin/rpc.statd)
960  # Start the statd process
961  debuglog /etc/init.d/nfslock start
962    fi
963}
964
965# Release all locks by sending a KILL signal to kernel lockd thread
966resetLockd() {
967    if [ -f /etc/init.d/nfsserver ]; then
968  # SuSE
969  debuglog /etc/init.d/nfslock stop
970    else
971        # Redhat fails to kill lockd to start grace period, so do it explicitly
972  pid=$(ps -aef | grep -w "\[lockd\]" | awk '{print $2}')
973  [ -n "$pid" ] && kill -9 $pid
974    fi
975}
976
977# Check grace period support in kernel
978checkDynamicGrace() {
979    debuglog2 mount -t nfsd nfsd /proc/fs/nfsd
980    [ -f /proc/fs/nfsd/grace ] && return 1
981    debugmsg2 "Cannot find /proc/fs/nfsd/grace, will restart lockd (and reclaim all locks) for failover/failback."
982    return 0
983}
984
985# Start grace period
986startGrace() {
987    checkDynamicGrace
988    if [ $? -eq 1 ]; then
989  echo 1 > /proc/fs/nfsd/grace
990    else
991  # Kernel does not support starting grace period through /proc
992  # Only thing to do is kill lockd
993  msg "Dynamic enabling of grace period not supported in this kernel. Restarting lockd"
994  resetLockd
995    fi
996}
997
998# Send SM_NOTIFY message to client on server restart
999# Usage: notify statd_dir [server IP]
1000notify() {
1001    opts=""
1002    if [ -e /sbin/sm-notify ]; then # SLES - kernel-space statd
1003  _cp $1/sm/* /var/lib/nfs/sm
1004  [ -n "$2" ] && opts="-m 1 -v $2"
1005  debugmsg2 "Notify clients: " && log ls $1/sm
1006  debuglog2 sm-notify $opts
1007    elif [ -e /sbin/rpc.statd ]; then # RHEL - user-space statd
1008  [ -n "$2" ] && opts="-P $2"
1009  debugmsg2 "Notify clients: " && log ls $1/sm
1010  debuglog2 rpc.statd -N -n $1 $opts
1011    fi
1012}
1013
1014# Send SM_NOTIFY message to client on all available interfaces
1015# SLES9 has a bug wherein the NFS client compares the hostname in the notify
1016# message against the hostname it registered during mount to determine whether
1017# to handle SM_NOTIFY requests coming from the server. We try to work around
1018# the problem by (optionally) sending notify messages on combinations of
1019# hostnames and IP address
1020notifyClient() {
1021    debugmsg2 "notify host:$1, vip:$VIP, statddir:$2"
1022
1023    #debugmsg "Notify on local name"
1024    #notify $2
1025
1026    # If VIP is specified, always send notify on the VIP
1027    if [ -n "$VIP" ]; then
1028  debugmsg "SM_NOTIFY clients for VIP $VIP"
1029  notify $2 $VIP
1030  [ $notifyfix -eq 0 ] && return
1031    fi
1032
1033    shortname=$(shortipname $1)
1034    debugmsg "Notify for host $shortname"
1035    notify $2 $shortname
1036    [ $notifyfix -eq 0 ] && return
1037
1038    host=$(ipname $1)
1039    if [[ -n "$host" && "$host" != "$shortname" ]]; then
1040  debugmsg "Notify for host.domain $host"
1041  notify $2 $host
1042    fi
1043
1044    ip=$(ipaddr $1)
1045    if [[ -n "$ip" && "$ip" != "$host" && "$ip" != "$shortname" ]]; then
1046  debugmsg "Notify for IP $ip"
1047  notify $2 $ip
1048    fi
1049}
1050
1051################################################################################
1052#       Failover functions             #
1053################################################################################
1054
1055startReclaim() {
1056    gpfs_ip=$1
1057    smdir=${SHARED_NFS}/$gpfs_ip/statd/sm
1058    statedir=/tmp/statd
1059
1060    msg "Reclaim of NLM locks initiated for node $gpfs_ip"
1061
1062    _mkdir $statedir/sm $statedir/sm.bak
1063    _cp $smdir/* $statedir/sm.bak
1064
1065    if [ $iptakeover -eq 1 ]; then
1066  for nfsip in "$(getNfsIPs $gpfs_ip)"; do
1067      _cp $statedir/sm.bak/* $statedir/sm
1068      notifyClient $nfsip $statedir
1069  done
1070    else
1071  # get VIP from loopback
1072  ip=$(mmgetifconf | grep -w 'lo' | awk '{print $2}')
1073  _cp $statedir/sm.bak/* $statedir/sm
1074  if [[ -n "$ip"  && "$ip" != "$gpfs_ip" ]]; then
1075      notifyClient $ip $statedir
1076  fi
1077    fi
1078    # restore list of lock users
1079    _cp $statedir/sm.bak/* $smdir
1080}
1081
1082# Find the next entry after one with node that failed and use it as the
1083# takeover node. If no more line wrap around to the top. Check that the node is
1084# up, if not use the next entry. There can be few NFS external IP address for
1085# each GPFS IP.
1086selectNode() {
1087    failed_node=$1
1088    typeset -i next=0
1089    gpfs_ip=""
1090
1091    while true; do
1092  exec 3< $NODELIST
1093  while read -u3 gpfs_if nfs_list; do
1094            # Skip empty and comment lines
1095      if invalid $gpfs_if; then 
1096    continue
1097      fi
1098      gpfs_ip=$(getip $gpfs_if)
1099      debugmsg2 "selectNode: GPFS IP: $gpfs_ip, NFS IP:$nfs_list"
1100      if [ $next -eq 1 ]; then
1101    [ "$gpfs_ip" == "$failed_node" ] && continue
1102    isNodeUp $gpfs_ip
1103    if [ $? -eq 0 ]; then
1104        debugmsg "takeover node is $gpfs_ip"
1105        echo $gpfs_ip
1106        return
1107    else
1108        debugmsg "selectNode: takeover_node $gpfs_ip is down"
1109        continue
1110    fi
1111      else
1112    [ "$gpfs_ip" == "$failed_node" ] && next=1
1113      fi 
1114  done
1115
1116        # Didn't find a takeover node, so start from top searching for new node
1117  debugmsg "selectNode: start from top"
1118  exec 3< $NODELIST
1119  while read -u3 gpfs_if nfs_list; do
1120            # Skip empty and comment lines
1121      if invalid $gpfs_if; then 
1122    continue
1123      fi
1124      gpfs_ip=$(getip $gpfs_if)
1125      debugmsg2 "selectNode: GPFS IP: $gpfs_ip, NFS IP: $nfs_list"
1126      if [ "$gpfs_ip" != "$failed_node" ]; then
1127    isNodeUp $gpfs_ip
1128    if [ $? -eq 0 ]; then
1129        debugmsg "takeover node is $gpfs_ip"
1130        echo $gpfs_ip
1131        return
1132    else
1133        debugmsg "selectNode: takeover_node $gpfs_ip is down"
1134        continue
1135    fi
1136      fi
1137  done
1138  sleep 10
1139    done
1140}
1141
1142selectNode2() {
1143  failed=$1
1144        candidate=$(getNextNode $failed $failed)
1145        while true; do
1146    while [ -z "$candidate" ]; do
1147      sleep 10
1148            candidate=$(getNextNode $failed $failed)
1149    done
1150    isNodeUp $candidate
1151    if [ $? -eq 0 ]; then
1152      echo $candidate
1153      return
1154    fi
1155    candidate=$(getNextNode $candidate $failed)
1156  done
1157}
1158
1159recoverNode() {
1160    failed_nfs_ip=$1
1161    failed_gpfs_ip=$2
1162    debugmsg "start recoverNode $failed_nfs_ip"
1163    [ -z "$failed_nfs_ip" ] && return 1
1164
1165    if [ -z "$(ifconfig | grep -wo $failed_nfs_ip)" ]; then
1166       typeset -i numberOfRetries=20 attemptNo=0
1167       while [[ $attemptNo -lt $numberOfRetries &&
1168      "$(IPaddr $failed_nfs_ip monitor)" == "OK" ]]; do
1169      attemptNo=$attemptNo+1
1170      sleep 1
1171       done
1172       # call stonith exit
1173       if [ $attemptNo -eq $numberOfRetries ]; then
1174          if [ -e /var/mmfs/etc/stonith ]; then
1175             debugmsg "call /var/mmfs/etc/stonith with $failed_gpfs_ip $failed_nfs_ip"
1176             /var/mmfs/etc/stonith $failed_gpfs_ip $failed_nfs_ip
1177             debugmsg "back from call to /var/mmfs/etc/stonith"
1178         fi
1179       fi
1180       typeset -i numberOfRetries=15 attemptNo=0
1181       while [[ $attemptNo -lt $numberOfRetries &&
1182      "$(IPaddr $failed_nfs_ip monitor)" == "OK" ]]; do
1183      attemptNo=$attemptNo+1
1184      sleep 1
1185       done
1186       if [ $attemptNo -eq $numberOfRetries ]; then
1187          # Somebody else has failed_nfs_ip - maybe the failed node is not down?
1188          msg "Error: some other host already has address $failed_nfs_ip. Recovery will not happen."
1189          return 1
1190       fi
1191    fi
1192
1193    debugmsg "recoverNode $failed_nfs_ip"
1194    debuglog IPaddr $failed_nfs_ip start
1195
1196    eth=$(mmgetifconf | grep -w $failed_nfs_ip | awk '{print $1}')
1197    debugmsg "Checking if interface for ip $failed_nfs_ip is up"
1198    if [ -n "$eth" ]; then
1199  return 0
1200    fi
1201    debugmsg "No interface for ip $failed_nfs_ip is up"
1202    return 1
1203}
1204
1205IPtakeover() {
1206    me=$1
1207    failed=$2
1208    typeset -i do_reclaim=0
1209
1210    [ "$me" == "$failed" ] && return
1211   
1212    msg "Initiating IP takeover of $failed due to node failure"
1213
1214    _mkdir ${SHARED_RECOVERY}/$me
1215    debuglog touch ${SHARED_RECOVERY}/$me/$failed
1216
1217    stopNode $failed
1218    nfsIPs=$(getNfsIPs $failed)
1219    debugmsg "IPtakeover ips: $nfsIPs"
1220    for ip in $nfsIPs; do
1221        # Takeover IP and issue gratuitous ARP to the clients for the node
1222        # that failed so that clients can reconnect to the new address
1223  recoverNode $ip $failed
1224        [ $? -eq 0 ] && do_reclaim=1 
1225    done
1226
1227    if [ $do_reclaim -ne 0 ]; then
1228  # got the IP, check if we are still the node to do takeover
1229  if [ ! -f $SHARED_RECOVERY/$me/$failed ]; then
1230     # drop the inteface
1231     nfsips=$(getNfsIPs $failed)
1232     debugmsg "Node $failed recovery canceled"
1233     for ip in $nfsips; do
1234        ifDown $ip
1235     done
1236     return
1237  fi
1238    else
1239        # did not get IP, takeover failed, remove the entry
1240  _unlink ${SHARED_RECOVERY}/$me/$failed
1241  return
1242    fi
1243    debugmsg "IPtakeover: File contents:"
1244    debugmsg $(ls -R ${SHARED_RECOVERY}/$me)
1245    _cp ${SHARED_NFS}/$failed/statd/sm/* ${SHARED_NFS}/$me/statd/sm
1246    checkDynamicGrace
1247    [ $? -eq 1 ] && startReclaim $failed
1248}
1249
1250# Use mii-diag, mii-tool or ethtool to detect network link status
1251# Return 0 if link beat detected, 1 if invalid (no device), 2 if no link beat
1252# FIX: If none of the tools exist, return 2 since there is no way to detect
1253#      link status. This means caller is responsible for handling the error
1254#      correctly.
1255netdiag() {
1256    # We trust mii-diag works for all interfaces;
1257    # if it exists, return its status
1258    tool=$(which mii-diag 2> /dev/null)
1259    if [ -n "$tool" ]; then
1260        debuglog2 $tool -s $eth
1261        return $?
1262    fi
1263
1264    # mii-diag doesn't exist - try both mii-tool and ethtool
1265    tool=$(which mii-tool 2> /dev/null)
1266    if [ -n "$tool" ]; then
1267        output=$($tool $eth 2> /dev/null)
1268        if [ $? -eq 0 ]; then
1269            status=$(echo $output | awk '{print $NF}')
1270            [ "$status" == "ok" ] && return 0
1271        fi
1272    fi
1273
1274    tool=$(which ethtool 2> /dev/null)
1275    if [ -n "$tool" ]; then
1276        output=$($tool $eth 2> /dev/null)
1277        if [ $? -eq 0 ]; then
1278            status=$(echo $output | grep "Link detected" | awk '{print $NF}')
1279            [ $status == "yes" ] && return 0
1280        fi
1281    fi
1282
1283    # We reach here either if no tool exists or if there is an error
1284    return 2
1285}
1286
1287# Returns 0 if ready, 1 otherwise
1288checkLinkStatus() {
1289    eth=$1
1290
1291    # Check if the interface is down
1292    debuglog2 ifconfig $eth
1293    [ $? -ne 0 ] && return 1
1294
1295    typeset -i tries=3 n=0
1296    while [ $n -lt $tries ]; do
1297        n=$n+1
1298        debugmsg2 "Checking link status for $eth - attempt $n"
1299
1300        netdiag $eth
1301        if [ $? -eq 0 ]; then
1302      debugmsg2 "Tested the link for $eth, and it is connected"
1303      return 0
1304        else
1305          debugmsg2 "Tested the link for $eth, and it is NOT connected"
1306            sleep 5
1307            continue
1308        fi
1309    done
1310    return 1
1311}
1312
1313getRecoveryNodes() {
1314    ls -A $SHARED_RECOVERY 2> /dev/null
1315}
1316
1317getFailedNodes() {
1318    ls -A $SHARED_RECOVERY/$1 2> /dev/null
1319}
1320
1321
1322IPfailback() {
1323    # Find failover IP address
1324    failover_ips=""
1325    for recovery_node in $(getRecoveryNodes); do
1326  debugmsg2 "Checking if $recovery_node is recovery node"
1327  for failed_node in $(getFailedNodes $recovery_node); do
1328      debugmsg "Failed node is $failed_node"
1329      if [ "$failed_node" == "$myip" ]; then
1330    failover_ips="$failover_ips $recovery_node"
1331      fi
1332  done
1333    done
1334
1335    if [ -z "$failover_ips" ]; then
1336  debugmsg "No failback is needed"
1337  return
1338    fi
1339    debugmsg "Failover nodes are $failover_ips"
1340
1341    # if recovery node dosen't hold my external IP then no failback
1342    tmp=""
1343    nfsIPs=$(getNfsIPs $myip)
1344    for nfsip in $nfsIPs; do
1345        for failover_ip in $failover_ips; do
1346      tmp=$(mmdshcmd $failover_ip "/usr/lpp/mmfs/bin/mmgetifconf | grep $nfsip")
1347      debugmsg2 check for ip=$nfsip on $failover_ip tmp=$tmp
1348      [ -n "$tmp" ] && break 2
1349  done
1350    done
1351    if [ -z "$tmp" ]; then
1352       # remove my node from the recovery node list
1353  debugmsg "IP is not in use"
1354    else
1355        # wait for any of the NFS IP to be up, but try all IPs at least once
1356        let up=0 
1357  while [ $up -eq 0 ]; do
1358      for ip in $nfsIPs; do
1359    iface=$(getifname $ip)
1360                # if this is a virtual address, the interface is already up
1361                isVirtualIP $ip
1362                RC=$?
1363                if [ $RC -ne 0 ]; then
1364        debugmsg2 wait for $ip
1365        ifcfg=$(getifcfg $ip)
1366
1367        debugmsg2 "Sourcing $ifcfg"
1368        . $ifcfg
1369
1370        if [ "$BONDING_MASTER" == "yes" ]; then
1371            ifBondUp $iface
1372        else
1373            debuglog ifconfig $iface up
1374        fi
1375                fi
1376    checkLinkStatus $iface
1377    if [ $? -eq 0 ]; then
1378        debugmsg2 "Testing the link for $iface, and it is connected"
1379        if ! arping -q -c 2 -w 3 -D -I $iface $ip ; then
1380      debugmsg2 $ip is ready
1381      up=1
1382        fi
1383    fi
1384                [ $RC -ne 0 ] && ifconfig $iface down
1385      done
1386      [ $up -eq 0 ] && sleep 5
1387  done
1388    fi
1389
1390    # remove my node from the recovery node list
1391    debuglog rm $SHARED_RECOVERY/*/$myip > /dev/null 2>&1
1392
1393    # save list of lock users for
1394    debugmsg "Failover IP is $failover_ip"
1395    smdir=/tmp/statd/sm
1396    _mkdir $smdir ${smdir}.bak
1397    _cp ${SHARED_NFS}/$failover_ip/statd/sm/* ${smdir}.bak
1398
1399    # Get recovery node to free my IP addresses
1400    mmdshAll "/var/mmfs/etc/nfsgrace $myip" >> $LOGFILE 2>&1
1401
1402    checkDynamicGrace
1403    if [ $? -eq 1 ]; then
1404        mmdshcmd $failover_ip "/var/mmfs/etc/nfsnodeback $failover_ip $myip" >> $LOGFILE 2>&1
1405    else
1406        gpfsIPs=$(getAllGPFSIPs)
1407        for ip in $gpfsIPs; do
1408            debugmsg2 "ip $ip, myip $ip"
1409      if [ "$ip" != "$myip" ]; then
1410                mmdshcmd $ip "/var/mmfs/etc/nfsnodeback $ip $failover_ip $myip" >> $LOGFILE 2>&1
1411            fi
1412        done
1413    fi
1414}
1415
Note: See TracBrowser for help on using the repository browser.