ctdb/config/events.d/60.nfs

   1 #!/bin/sh
   2 # script to manage nfs in a clustered environment
   3
   4 . $CTDB_BASE/functions
   5
   6 service_name="nfs"
   7 service_start ()
   8 {
   9     startstop_nfs stop
  10     startstop_nfs start
  11     set_proc "sys/net/ipv4/tcp_tw_recycle" 1
  12     touch "$service_state_dir/update-trigger"
  13 }
  14 service_stop ()
  15 {
  16     startstop_nfs stop
  17 }
  18 service_reconfigure="startstop_nfs restart"
  19
  20 loadconfig
  21
  22 [ "$NFS_SERVER_MODE" != "GANESHA" ] || exit 0
  23
  24 ctdb_setup_service_state_dir
  25
  26 ctdb_start_stop_service
  27
  28 is_ctdb_managed_service || exit 0
  29
  30 ctdb_service_check_reconfigure
  31
  32 case "$1" in
  33      init)
  34         # read statd from persistent database
  35         ;;
  36      startup)
  37         ctdb_service_start
  38         ;;
  39
  40      shutdown)
  41         ctdb_service_stop
  42         ;;
  43
  44      takeip)
  45         ctdb_service_set_reconfigure
  46         ;;
  47
  48      releaseip)
  49         ctdb_service_set_reconfigure
  50         ;;
  51
  52       monitor)
  53         # and that its directories are available
  54         [ "$CTDB_NFS_SKIP_SHARE_CHECK" = "yes" ] || {
  55             exportfs | grep -v '^#' | grep '^/' |
  56             sed -e 's/[[:space:]]\+[^[:space:]]*$//' |
  57             ctdb_check_directories
  58         } || exit $?
  59
  60         update_tickles 2049
  61
  62         # check that statd responds to rpc requests
  63         # if statd is not running we try to restart it
  64         # we only do this IF we have a rpc.statd command.
  65         # For platforms where rpc.statd does not exist, we skip
  66         # the check completely
  67         p="rpc.statd"
  68         which $p >/dev/null 2>/dev/null && {
  69                 if ctdb_check_rpc "STATD" 100024 1 >/dev/null ; then
  70                         (service_name="nfs_statd"; ctdb_counter_init)
  71                 else
  72                         cmd="$p"
  73                         cmd="${cmd}${STATD_HOSTNAME:+ -n }${STATD_HOSTNAME}"
  74                         cmd="${cmd}${STATD_PORT:+ -p }${STATD_PORT}"
  75                         cmd="${cmd}${STATD_OUTGOING_PORT:+ -o }${STATD_OUTGOING_PORT}"
  76                         (
  77                                 service_name="nfs_statd"
  78                                 ctdb_counter_incr
  79                                 ctdb_check_counter_limit 10 quiet >/dev/null
  80                         ) || {
  81                                 echo "$ctdb_check_rpc_out"
  82                                 echo "Trying to restart STATD [$cmd]"
  83                                 $cmd
  84                         }
  85                 fi
  86         }
  87
  88         # check that NFS responds to rpc requests
  89         [ "$CTDB_NFS_SKIP_KNFSD_ALIVE_CHECK" = "yes" ] || {
  90             if ctdb_check_rpc "NFS" 100003 3 >/dev/null ; then
  91                 (service_name="nfs_knfsd"; ctdb_counter_init)
  92             else
  93                 (
  94                         service_name="nfs_knfsd"
  95                         ctdb_counter_incr
  96
  97                         ctdb_check_counter_equal 2 || {
  98                                 echo "Trying to restart NFS service"
  99                                 startstop_nfs restart >/dev/null 2>&1 &
 100                                 exit 0
 101                         }
 102
 103                         ctdb_check_counter_limit 5 quiet >/dev/null
 104                 ) || {
 105                         echo "$ctdb_check_rpc_out"
 106                         echo "Trying to restart NFS service"
 107                         startstop_nfs restart
 108                         exit 1
 109                 }
 110             fi
 111         }
 112
 113         # check that lockd responds to rpc requests
 114         if ctdb_check_rpc "LOCKD" 100021 1 >/dev/null ; then
 115                 (service_name="lockd"; ctdb_counter_init)
 116         else
 117                 (
 118                         service_name="lockd"
 119                         ctdb_counter_incr
 120
 121                         ctdb_check_counter_equal 10 || {
 122                                 echo "Trying to restart NFS lock service"
 123                                 startstop_nfs restart >/dev/null 2>&1 &
 124                                 startstop_nfslock restart  >/dev/null 2>&1 &
 125                                 exit 0
 126                         }
 127
 128                         ctdb_check_counter_limit 15 quiet >/dev/null
 129         ) || {
 130                         echo "$ctdb_check_rpc_out"
 131                         echo "Trying to restart NFS lock service"
 132                         startstop_nfs restart
 133                         startstop_nfslock restart
 134                         exit 1
 135                 }
 136         fi
 137
 138         # mount needs special handling since it is sometimes not started
 139         # correctly on RHEL5
 140         if ctdb_check_rpc "MOUNTD" 100005 1 >/dev/null ; then
 141                 (service_name="nfs_mountd"; ctdb_counter_init)
 142         else
 143         (
 144                 service_name="nfs_mountd"
 145                 ctdb_counter_incr
 146
 147                 ctdb_check_counter_equal 5 || {
 148                         p="rpc.mountd"
 149                         cmd="${p}${MOUNTD_PORT:+ -p }${MOUNTD_PORT}"
 150                         echo "Trying to restart MOUNTD [${cmd}]"
 151                         killall -q -9 $p
 152                         $cmd &
 153                         exit 0
 154                 }
 155
 156                 ctdb_check_counter_limit 10 quiet >/dev/null
 157         ) || {
 158                 echo "$ctdb_check_rpc_out"
 159                 p="rpc.mountd"
 160                 cmd="${p}${MOUNTD_PORT:+ -p }${MOUNTD_PORT}"
 161                 echo "Trying to restart MOUNTD [${cmd}]"
 162                 killall -q -9 $p
 163                 $cmd &
 164                 exit 1
 165         }
 166         fi
 167
 168
 169         # rquotad needs special handling since it is sometimes not started
 170         # correctly on RHEL5
 171         # this is not a critical service so we dont flag the node as unhealthy
 172         ctdb_check_rpc "RQUOTAD" 100011 1 || {
 173                 p="rpc.rquotad"
 174                 cmd="${p}${RQUOTAD_PORT:+ -p }${RQUOTAD_PORT}"
 175                 echo "Trying to restart RQUOTAD [${cmd}]"
 176                 killall -q -9 $p
 177                 $cmd &
 178         }
 179
 180         # once every 600 seconds, update the statd state database for which
 181         # clients need notifications
 182         LAST_UPDATE=`stat --printf="%Y" "$service_state_dir/update-trigger"`
 183         CURRENT_TIME=`date +"%s"`
 184         [ $CURRENT_TIME -ge $(($LAST_UPDATE + 600)) ] && {
 185             touch "$service_state_dir/update-trigger"
 186             $CTDB_BASE/statd-callout updatelocal &
 187             $CTDB_BASE/statd-callout updateremote &
 188         }
 189         ;;
 190
 191     ipreallocated)
 192         # if the ips have been reallocated, we must restart the lockmanager
 193         # across all nodes and ping all statd listeners
 194         [ -x $CTDB_BASE/statd-callout ] && {
 195                 $CTDB_BASE/statd-callout notify &
 196         } >/dev/null 2>&1
 197         ;;
 198     *)
 199         ctdb_standard_event_handler "$@"
 200         ;;
 201 esac
 202
 203 exit 0