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