config/events.d/60.nfs

   1 #!/bin/sh
   2 # script to manage nfs in a clustered environment
   3
   4 start_nfs() {
   5         /bin/mkdir -p $CTDB_VARDIR/state/nfs
   6         /bin/mkdir -p $CTDB_VARDIR/state/statd/ip
   7         startstop_nfs stop
   8         startstop_nfs start
   9         echo 1 > /proc/sys/net/ipv4/tcp_tw_recycle
  10 }
  11
  12 . $CTDB_BASE/functions
  13
  14 service_name="nfs"
  15 service_start="start_nfs"
  16 service_stop="startstop_nfs stop"
  17
  18 loadconfig
  19
  20 [ "$NFS_SERVER_MODE" != "GANESHA" ] || exit 0
  21
  22 ctdb_start_stop_service
  23
  24 is_ctdb_managed_service || exit 0
  25
  26 case "$1" in
  27      init)
  28         # read statd from persistent database
  29         ;;
  30      startup)
  31         ctdb_service_start
  32         mkdir -p $CTDB_VARDIR/state/statd
  33         touch $CTDB_VARDIR/state/statd/update-trigger
  34         ;;
  35
  36      shutdown)
  37         ctdb_service_stop
  38         ;;
  39
  40      takeip)
  41         ctdb_service_set_reconfigure
  42         ;;
  43
  44      releaseip)
  45         ctdb_service_set_reconfigure
  46         ;;
  47
  48       monitor)
  49         if ctdb_service_needs_reconfigure ; then
  50             ctdb_service_reconfigure
  51             exit 0
  52         fi
  53
  54         update_tickles 2049
  55
  56         # check that statd responds to rpc requests
  57         # if statd is not running we try to restart it
  58         if ctdb_check_rpc "STATD" 100024 1 >/dev/null ; then
  59                 (service_name="nfs_statd"; ctdb_counter_init)
  60         else
  61                 p="rpc.statd" ; cmd="$p"
  62                 cmd="${cmd}${STATD_HOSTNAME:+ -n }${STATD_HOSTNAME}"
  63                 cmd="${cmd}${STATD_PORT:+ -p }${STATD_PORT}"
  64                 cmd="${cmd}${STATD_OUTGOING_PORT:+ -o }${STATD_OUTGOING_PORT}"
  65                 (
  66                         service_name="nfs_statd"
  67                         ctdb_counter_incr
  68                         ctdb_check_counter_limit 10 quiet >/dev/null
  69                 ) || {
  70                         echo "$ctdb_check_rpc_out"
  71                         echo "Trying to restart STATD [$cmd]"
  72                 }
  73                 $cmd
  74         fi
  75
  76
  77         # check that NFS responds to rpc requests
  78         [ "$CTDB_NFS_SKIP_KNFSD_ALIVE_CHECK" = "yes" ] || {
  79             if ctdb_check_rpc "NFS" 100003 3 >/dev/null ; then
  80                 (service_name="nfs_knfsd"; ctdb_counter_init)
  81             else
  82                 (
  83                         service_name="nfs_knfsd"
  84                         ctdb_counter_incr
  85
  86                         ctdb_check_counter_equal 10 || {
  87                                 echo "Trying to restart NFS service"
  88                                 startstop_nfs restart >/dev/null 2>&1 &
  89                                 exit 0
  90                         }
  91
  92                         ctdb_check_counter_limit 15 quiet >/dev/null
  93                 ) || {
  94                         echo "$ctdb_check_rpc_out"
  95                         echo "Trying to restart NFS service"
  96                         startstop_nfs restart
  97                         exit 1
  98                 }
  99             fi
 100         }
 101
 102         # and that its directories are available
 103         [ "$CTDB_NFS_SKIP_SHARE_CHECK" = "yes" ] || {
 104             exportfs | grep -v '^#' | grep '^/' |
 105             sed -e 's/[[:space:]]\+[^[:space:]]*$//' |
 106             ctdb_check_directories
 107         } || exit $?
 108
 109         # check that lockd responds to rpc requests
 110         ctdb_check_rpc "LOCKD" 100021 1 || {
 111                 echo "Trying to restart lock manager service"
 112                 startstop_nfs restart
 113                 startstop_nfslock restart
 114                 exit 1
 115         }
 116
 117         # mount needs special handling since it is sometimes not started
 118         # correctly on RHEL5
 119         if ctdb_check_rpc "MOUNTD" 100005 1 >/dev/null ; then
 120                 (service_name="nfs_mountd"; ctdb_counter_init)
 121         else
 122         (
 123                 service_name="nfs_mountd"
 124                 ctdb_counter_incr
 125
 126                 ctdb_check_counter_equal 5 || {
 127                         p="rpc.mountd"
 128                         cmd="${p}${MOUNTD_PORT:+ -p }${MOUNTD_PORT}"
 129                         echo "Trying to restart MOUNTD [${cmd}]"
 130                         killall -q -9 $p
 131                         $cmd &
 132                         exit 0
 133                 }
 134
 135                 ctdb_check_counter_limit 10 quiet >/dev/null
 136         ) || {
 137                 echo "$ctdb_check_rpc_out"
 138                 p="rpc.mountd"
 139                 cmd="${p}${MOUNTD_PORT:+ -p }${MOUNTD_PORT}"
 140                 echo "Trying to restart MOUNTD [${cmd}]"
 141                 killall -q -9 $p
 142                 $cmd &
 143                 exit 1
 144         }
 145         fi
 146
 147
 148         # rquotad needs special handling since it is sometimes not started
 149         # correctly on RHEL5
 150         # this is not a critical service so we dont flag the node as unhealthy
 151         ctdb_check_rpc "RQUOTAD" 100011 1 || {
 152                 p="rpc.rquotad"
 153                 cmd="${p}${RQUOTAD_PORT:+ -p }${RQUOTAD_PORT}"
 154                 echo "Trying to restart RQUOTAD [${cmd}]"
 155                 killall -q -9 $p
 156                 $cmd &
 157         }
 158
 159         # once every 60 seconds, update the statd state database for which
 160         # clients need notifications
 161         LAST_UPDATE=`stat --printf="%Y" $CTDB_VARDIR/state/statd/update-trigger 2>/dev/null`
 162         CURRENT_TIME=`date +"%s"`
 163         [ $CURRENT_TIME -ge $(($LAST_UPDATE + 60)) ] && {
 164             mkdir -p $CTDB_VARDIR/state/statd
 165             touch $CTDB_VARDIR/state/statd/update-trigger
 166             $CTDB_BASE/statd-callout updatelocal &
 167             $CTDB_BASE/statd-callout updateremote &
 168         }
 169         ;;
 170
 171     ipreallocated)
 172         # if the ips have been reallocated, we must restart the lockmanager
 173         # across all nodes and ping all statd listeners
 174         [ -x $CTDB_BASE/statd-callout ] && {
 175                 $CTDB_BASE/statd-callout notify &
 176         } >/dev/null 2>&1
 177         ;;
 178     *)
 179         ctdb_standard_event_handler "$@"
 180         ;;
 181 esac
 182
 183 exit 0