config/events.d/60.ganesha

   1 #!/bin/sh
   2 # script to manage nfs in a clustered environment
   3
   4 [ -n "$CTDB_BASE" ] || \
   5     export CTDB_BASE=$(cd -P $(dirname "$0") ; dirname "$PWD")
   6
   7 . $CTDB_BASE/functions
   8
   9 GANRECDIR="/var/lib/nfs/ganesha"
  10 GANRECDIR2="/var/lib/nfs/ganesha/recevents"
  11 GPFS_STATE="/usr/lpp/mmfs/bin/mmgetstate"
  12 GANRECDIR3="/var/lib/nfs/ganesha_local"
  13
  14
  15 service_start ()
  16 {
  17     startstop_ganesha stop
  18     startstop_ganesha start
  19     set_proc "sys/net/ipv4/tcp_tw_recycle" 1
  20 }
  21
  22 service_stop ()
  23 {
  24     startstop_ganesha stop
  25 }
  26
  27 service_reconfigure ()
  28 {
  29     # if the ips have been reallocated, we must restart ganesha
  30     # across all nodes and ping all statd listeners
  31     [ -x $CTDB_BASE/statd-callout ] && {
  32         $CTDB_BASE/statd-callout notify &
  33     } >/dev/null 2>&1
  34 }
  35
  36 loadconfig "nfs"
  37
  38
  39 [ -n "$CTDB_CLUSTER_FILESYSTEM_TYPE" ] || CTDB_CLUSTER_FILESYSTEM_TYPE="gpfs"
  40
  41 service_name="nfs-ganesha-$CTDB_CLUSTER_FILESYSTEM_TYPE"
  42
  43 [ "${CTDB_NFS_SERVER_MODE:-${NFS_SERVER_MODE}}" = "ganesha" ] || exit 0
  44
  45 ctdb_setup_service_state_dir
  46
  47 ctdb_start_stop_service
  48
  49 is_ctdb_managed_service || exit 0
  50
  51 ctdb_service_check_reconfigure
  52
  53 get_cluster_fs_state  ()
  54 {
  55     case $CTDB_CLUSTER_FILESYSTEM_TYPE in
  56         gpfs)
  57             STATE=`$GPFS_STATE | awk 'NR <= 3 {next} {printf "%-6s", $3}'`
  58             echo $STATE
  59             ;;
  60         *)
  61             die "File system $CTDB_CLUSTER_FILESYSTEM_TYPE not supported"
  62             ;;
  63    esac
  64 }
  65
  66 create_ganesha_recdirs ()
  67 {
  68     MOUNTS=$(mount -t $CTDB_CLUSTER_FILESYSTEM_TYPE)
  69     if [ -z "$MOUNTS" ]; then
  70       echo "startup $CTDB_CLUSTER_FILESYSTEM_TYPE not ready"
  71       exit 0
  72     fi
  73     MNTPT=$(echo "$MOUNTS" | sort | awk 'NR == 1 {print $3}')
  74     mkdir -p $MNTPT/.ganesha
  75     if [ -e $GANRECDIR ]; then
  76         if [ ! -L $GANRECDIR ] ; then
  77             rm -rf $GANRECDIR
  78             if ! ln -s $MNTPT/.ganesha  $GANRECDIR ; then
  79                 echo "ln failed"
  80             fi
  81         fi
  82     else
  83         if ! ln -sf $MNTPT/.ganesha  $GANRECDIR ; then
  84             echo "ln failed"
  85         fi
  86     fi
  87
  88     mkdir -p $GANRECDIR2
  89     mkdir -p $GANRECDIR3
  90 }
  91
  92 monitor_ganesha_nfsd ()
  93 {
  94         create_ganesha_recdirs
  95         service_name=${service_name}_process
  96
  97         PIDFILE="/var/run/ganesha.pid"
  98         CUR_STATE=`get_cluster_fs_state`
  99         GANESHA="/usr/bin/$CTDB_CLUSTER_FILESYSTEM_TYPE.ganesha.nfsd"
 100         if { read PID < $PIDFILE && \
 101             grep "$GANESHA" "/proc/$PID/cmdline" ; } >/dev/null 2>&1 ; then
 102                 ctdb_counter_init "$service_name"
 103         else
 104             if [ $CUR_STATE = "active" ]; then
 105                 echo "Trying fast restart of NFS service"
 106                 startstop_ganesha restart
 107                 ctdb_counter_incr "$service_name"
 108                 ctdb_check_counter "error" "-ge" "6" "$service_name"
 109             fi
 110         fi
 111
 112         service_name="nfs-ganesha-$CTDB_CLUSTER_FILESYSTEM_TYPE"_service
 113         # check that NFS is posting forward progress
 114         if [ $CUR_STATE = "active" -a "$CTDB_NFS_SKIP_KNFSD_ALIVE_CHECK" != "yes" ] ; then
 115             MAXREDS=2
 116             MAXSTALL=120
 117             RESTART=0
 118
 119             NUMREDS=`ls $GANRECDIR3 | grep "red" | wc -l`
 120             LASTONE=`ls -t $GANRECDIR3 | sed 's/_/ /' | awk 'NR > 1 {next} {printf $1} '`
 121             # Beware of startup
 122             if [ -z $LASTONE ] ; then
 123                 LASTONE=`date +"%s"`
 124             fi
 125             TNOW=$(date +"%s")
 126             TSTALL=$(($TNOW - $LASTONE))
 127             if [ $NUMREDS -ge $MAXREDS ] ; then
 128                 echo restarting because of $NUMREDS red conditions
 129                 RESTART=1
 130                 ctdb_counter_incr "$service_name"
 131                 ctdb_check_counter "error" "-ge" "6" "$service_name"
 132             fi
 133             if [ $TSTALL -ge $MAXSTALL ] ; then
 134                 echo restarting because of $TSTALL second stall
 135                 RESTART=1
 136                 ctdb_counter_incr "$service_name"
 137                 ctdb_check_counter "error" "-ge" "6" "$service_name"
 138             fi
 139             if [ $RESTART -gt 0 ] ; then
 140                 startstop_ganesha restart
 141             else
 142                 ctdb_counter_init "$service_name"
 143             fi
 144         fi
 145 }
 146
 147 ############################################################
 148
 149 case "$1" in
 150      init)
 151         # read statd from persistent database
 152         ;;
 153      startup)
 154         create_ganesha_recdirs
 155         ctdb_service_start
 156         ;;
 157
 158      shutdown)
 159         ctdb_service_stop
 160         ;;
 161
 162      takeip)
 163         if [ -n "$2" ] ; then
 164             case  $CTDB_CLUSTER_FILESYSTEM_TYPE in
 165                 gpfs)
 166                     NNUM=`/usr/lpp/mmfs/bin/mmlsconfig myNodeConfigNumber | awk '{print $2}'`
 167                     TDATE=`date +"%s"`
 168                     TOUCHTGT=$1"_"$TDATE"_"$NNUM"_"$3"_"$4"_"$2
 169                     touch $GANRECDIR2/$TOUCHTGT
 170                     ;;
 171             esac
 172         fi
 173         ctdb_service_set_reconfigure
 174         ;;
 175
 176      releaseip)
 177         if [ -n "$2" ] ; then
 178             case  $CTDB_CLUSTER_FILESYSTEM_TYPE in
 179                 gpfs)
 180                     NNUM=`/usr/lpp/mmfs/bin/mmlsconfig myNodeConfigNumber | awk '{print $2}'`
 181                     TDATE=`date +"%s"`
 182                     TOUCHTGT=$1"_"$TDATE"_"$NNUM"_"$3"_"$4"_"$2
 183                     touch $GANRECDIR2/$TOUCHTGT
 184                 ;;
 185             esac
 186         fi
 187         ctdb_service_set_reconfigure
 188         ;;
 189
 190      monitor)
 191         # Check that directories for shares actually exist.
 192         [ "$CTDB_NFS_SKIP_SHARE_CHECK" = "yes" ] || {
 193             grep Path /etc/ganesha/$CTDB_CLUSTER_FILESYSTEM_TYPE.ganesha.exports.conf |
 194             cut -f2 -d\" | sort -u | ctdb_check_directories
 195         } || exit $?
 196
 197         update_tickles 2049
 198
 199         # check that statd responds to rpc requests
 200         # if statd is not running we try to restart it
 201         # we only do this IF we have a rpc.statd command.
 202         # For platforms where rpc.statd does not exist, we skip
 203         # the check completely
 204         p="rpc.statd"
 205         which $p >/dev/null 2>/dev/null && \
 206             nfs_check_rpc_service "statd" \
 207                 -ge 6 "verbose unhealthy" \
 208                 -eq 4 "verbose restart" \
 209                 -eq 2 "restart:b"
 210
 211         if [ "$CTDB_SKIP_GANESHA_NFSD_CHECK" != "yes" ] ; then
 212             monitor_ganesha_nfsd
 213         fi
 214
 215         # rquotad is sometimes not started correctly on RHEL5
 216         # not a critical service so we dont flag the node as unhealthy
 217         nfs_check_rpc_service "rquotad" \
 218             -gt 0 "verbose restart:b"
 219         ;;
 220
 221      *)
 222         ctdb_standard_event_handler "$@"
 223         ;;
 224 esac
 225
 226 exit 0