#!/bin/sh
# script to manage nfs in a clustered environment
-start_nfs() {
- mkdir -p $CTDB_VARDIR/state/nfs
- mkdir -p $CTDB_VARDIR/state/statd/ip
- ctdb_service_stop
- ctdb_service_start
- echo 1 > /proc/sys/net/ipv4/tcp_tw_recycle
+. $CTDB_BASE/functions
+
+service_name="nfs-ganesha-gpfs"
+
+
+service_start ()
+{
+ startstop_ganesha stop
+ startstop_ganesha start
+ set_proc "sys/net/ipv4/tcp_tw_recycle" 1
}
-. $CTDB_BASE/functions
+service_stop ()
+{
+ startstop_ganesha stop
+}
+
+service_reconfigure ()
+{
+ # if the ips have been reallocated, we must restart ganesha
+ # across all nodes and ping all statd listeners
+ [ -x $CTDB_BASE/statd-callout ] && {
+ $CTDB_BASE/statd-callout notify &
+ } >/dev/null 2>&1
+}
+loadconfig "nfs"
-loadconfig nfs
+[ "$NFS_SERVER_MODE" == "ganesha" ] || exit 0
-[ "$NFS_SERVER_MODE" = "GANESHA" ] || exit 0
+ctdb_setup_service_state_dir
+
+statd_update_trigger="$service_state_dir/update-trigger"
+# We want this file to always exist. The corner case is when
+# auto-start/stop is switched off, NFS is added as a managed service
+# some time after ctdbd is started and someone else starts the NFS
+# service for us. In this case this file might not otherwise exist
+# when we get to a monitor event.
+touch "$statd_update_trigger"
-service_name="nfs-ganesha-gpfs"
ctdb_start_stop_service
is_ctdb_managed_service || exit 0
+ctdb_service_check_reconfigure
+
case "$1" in
init)
# read statd from persistent database
;;
startup)
ctdb_service_start
- mkdir -p $CTDB_VARDIR/state/statd
- touch $CTDB_VARDIR/state/statd/update-trigger
;;
shutdown)
;;
monitor)
- if ctdb_service_needs_reconfigure ; then
- ctdb_service_reconfigure
- exit 0
- fi
update_tickles 2049
-
- # check that statd responds to rpc requests
- # if statd is not running we try to restart it
- if ctdb_check_rpc "STATD" status 1 >/dev/null ; then
- (service_name="nfs_statd"; ctdb_counter_init)
- else
- p="rpc.statd" ; cmd="$p"
- cmd="${cmd}${STATD_HOSTNAME:+ -n }${STATD_HOSTNAME}"
- cmd="${cmd}${STATD_PORT:+ -p }${STATD_PORT}"
- cmd="${cmd}${STATD_OUTGOING_PORT:+ -o }${STATD_OUTGOING_PORT}"
- (
- service_name="nfs_statd"
- ctdb_counter_incr
- ctdb_check_counter_limit 10 quiet >/dev/null
- ) || {
- echo "$ctdb_check_rpc_out"
- echo "Trying to restart STATD [$cmd]"
- }
- $cmd
- fi
-
- # check that NFS responds to rpc requests
- [ "$CTDB_NFS_SKIP_KNFSD_ALIVE_CHECK" = "yes" ] || {
- if ctdb_check_rpc "NFS" nfs 3 >/dev/null ; then
- (service_name="nfs_knfsd"; ctdb_counter_init)
- else
- (
- service_name="nfs_knfsd"
- ctdb_counter_incr
-
- ctdb_check_counter_equal 10 || {
- echo "Trying to restart NFS service"
- ctdb_service_stop
- ctdb_service_start
- exit 0
- }
-
- ctdb_check_counter_limit 15 quiet >/dev/null
- ) || {
- echo "$ctdb_check_rpc_out"
- echo "Trying to restart NFS service"
- ctdb_service_stop
- ctdb_service_start
- exit 1
- }
- fi
- }
-
- # and that its directories are available
- [ "$CTDB_NFS_SKIP_SHARE_CHECK" = "yes" ] || {
- grep Path /etc/ganesha/gpfs.ganesha.exports.conf |
- cut -f2 -d\" | ctdb_check_directories
- } || exit $?
-
- # check that lockd responds to rpc requests
- ctdb_check_rpc "LOCKD" nlockmgr 4 || {
- echo "Trying to restart lock manager service"
- ctdb_service_stop
- ctdb_service_start
- exit 1
- }
-
- # check mounts responds to rpc requests
- ctdb_check_rpc "MOUNTD" mountd 1 >/dev/null || {
- echo "Trying to restart mountd service"
- ctdb_service_stop
- ctdb_service_start
- exit 1
- }
-
- # rquotad needs special handling since it is sometimes not started
- # correctly on RHEL5
- # this is not a critical service so we dont flag the node as unhealthy
- ctdb_check_rpc "RQUOTAD" rquotad 1 || {
- p="rpc.rquotad"
- cmd="${p}${RQUOTAD_PORT:+ -p }${RQUOTAD_PORT}"
- echo "Trying to restart RQUOTAD [${cmd}]"
- killall -q -9 $p
- $cmd &
- }
-
- # once every 60 seconds, update the statd state database for which
- # clients need notifications
- LAST_UPDATE=`stat --printf="%Y" $CTDB_VARDIR/state/statd/update-trigger 2>/dev/null`
- CURRENT_TIME=`date +"%s"`
- [ $CURRENT_TIME -ge $(($LAST_UPDATE + 60)) ] && {
- mkdir -p $CTDB_VARDIR/state/statd
- touch $CTDB_VARDIR/state/statd/update-trigger
- $CTDB_BASE/statd-callout updatelocal &
- $CTDB_BASE/statd-callout updateremote &
- }
- ;;
+ # check that statd responds to rpc requests
+ # if statd is not running we try to restart it
+ # we only do this IF we have a rpc.statd command.
+ # For platforms where rpc.statd does not exist, we skip
+ # the check completely
+ p="rpc.statd"
+ which $p >/dev/null 2>/dev/null && \
+ nfs_check_rpc_service "statd" 1 \
+ -ge 6 "verbose unhealthy" \
+ -eq 4 "verbose restart" \
+ -eq 2 "restart:bs"
+
+ PIDFILE="/var/run/ganesha.pid"
+ RUNNING=0
+ if [ -e $PIDFILE ]
+ then
+ PID=`cat $PIDFILE`
+ GANESHA="/usr/bin/gpfs.ganesha.nfsd"
+ RUNNING=`cat /proc/$PID/cmdline | grep $GANESHA | wc -l`
+ fi
+ if [ $RUNNING != 1 ]
+ then
+ echo "Trying fast restart of NFS service"
+ startstop_ganesha restart
+ fi
+
+ # check that NFS responds to rpc requests
+ if [ "$CTDB_NFS_SKIP_KNFSD_ALIVE_CHECK" != "yes" ] ; then
+ nfs_check_rpc_service "ganesha" \
+ -ge 6 "verbose unhealthy" \
+ -eq 4 "verbose restart" \
+ -eq 2 "restart:bs"
+ fi
+
+
+ # rquotad is sometimes not started correctly on RHEL5
+ # not a critical service so we dont flag the node as unhealthy
+ nfs_check_rpc_service "rquotad" 1\
+ -gt 0 "verbose restart:b"
+
+
+ # Check that directories for shares actually exist.
+ [ "$CTDB_NFS_SKIP_SHARE_CHECK" = "yes" ] || {
+ grep Path /etc/ganesha/gpfs.ganesha.exports.conf |
+ cut -f2 -d\" | ctdb_check_directories
+ } || exit $?
+
+ # once every 60 seconds, update the statd state database for which
+ # clients need notifications
+ LAST_UPDATE=`stat --printf="%Y" "$statd_update_trigger" 2>/dev/null`
+ CURRENT_TIME=`date +"%s"`
+ [ $CURRENT_TIME -ge $(($LAST_UPDATE + 60)) ] && {
+ touch "$statd_update_trigger"
+ $CTDB_BASE/statd-callout updatelocal &
+ $CTDB_BASE/statd-callout updateremote &
+ }
+ ;;
ipreallocated)
- # if the ips have been reallocated, we must restart the lockmanager
- # across all nodes and ping all statd listeners
- [ -x $CTDB_BASE/statd-callout ] && {
- $CTDB_BASE/statd-callout notify &
- } >/dev/null 2>&1
+ ctdb_service_set_reconfigure
;;
*)
ctdb_standard_event_handler "$@"