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