2007-06-01 18:10:22 +04:00
#!/bin/sh
# script to manage nfs in a clustered environment
2013-01-03 08:26:12 +04:00
[ -n "$CTDB_BASE" ] || \
export CTDB_BASE=$(cd -P $(dirname "$0") ; dirname "$PWD")
2009-11-19 07:00:17 +03:00
. $CTDB_BASE/functions
2007-06-06 06:08:42 +04:00
2009-11-19 07:00:17 +03:00
service_name="nfs"
2015-07-14 05:11:39 +03:00
loadconfig
ctdb_setup_service_state_dir
######################################################################
2015-06-24 14:36:14 +03:00
if [ -z "$CTDB_NFS_CALLOUT" ] ; then
CTDB_NFS_CALLOUT="${CTDB_BASE}/nfs-linux-kernel-callout"
fi
# Always export, for statd callout
export CTDB_NFS_CALLOUT
2015-09-02 02:27:58 +03:00
# If the callout wants to use this then it must create it
export CTDB_NFS_CALLOUT_STATE_DIR="${service_state_dir}/callout-state"
2015-07-14 05:11:39 +03:00
nfs_callout_cache="${service_state_dir}/nfs_callout_cache"
nfs_callout_cache_callout="${nfs_callout_cache}/CTDB_NFS_CALLOUT"
nfs_callout_cache_ops="${nfs_callout_cache}/ops"
nfs_callout_register ()
{
mkdir -p "$nfs_callout_cache_ops"
rm -f "$nfs_callout_cache_ops"/*
echo "$CTDB_NFS_CALLOUT" >"$nfs_callout_cache_callout"
_t=$(eval "$CTDB_NFS_CALLOUT" "register")
if [ -n "$_t" ] ; then
echo "$_t" |
while IFS="" read _op ; do
touch "${nfs_callout_cache_ops}/${_op}"
done
else
touch "${nfs_callout_cache_ops}/ALL"
fi
}
2015-06-24 14:36:14 +03:00
nfs_callout ()
2011-08-11 03:39:25 +04:00
{
2015-07-14 05:11:39 +03:00
# Re-run registration if $CTDB_NFS_CALLOUT has changed
_prev=""
if [ -r "$nfs_callout_cache_callout" ] ; then
read _prev <"$nfs_callout_cache_callout"
fi
if [ "$CTDB_NFS_CALLOUT" != "$_prev" ] ; then
nfs_callout_register
fi
# Run the operation if it is registered...
if [ -e "${nfs_callout_cache_ops}/${1}" ] || \
[ -e "${nfs_callout_cache_ops}/ALL" ]; then
eval "$CTDB_NFS_CALLOUT" "$@"
fi
2011-08-11 03:39:25 +04:00
}
2015-06-24 14:36:14 +03:00
2011-08-11 07:55:02 +04:00
service_reconfigure ()
{
2015-07-08 07:57:51 +03:00
# Restart lock manager, notify clients
if [ -x "${CTDB_BASE}/statd-callout" ] ; then
"${CTDB_BASE}/statd-callout" notify &
fi >/dev/null 2>&1
2011-08-11 07:55:02 +04:00
}
2008-02-11 01:35:37 +03:00
2015-07-13 08:22:23 +03:00
######################################################################
######################################################
# Check the health of NFS services
#
# Use .check files in given directory.
# Default is "${CTDB_BASE}/nfs-checks.d/"
######################################################
nfs_check_services ()
{
_dir="${1:-${CTDB_NFS_CHECKS_DIR:-${CTDB_BASE}/nfs-checks.d}}"
# Files must end with .check - avoids editor backups, RPM fu, ...
for _f in "$_dir"/[0-9][0-9].*.check ; do
_t="${_f%.check}"
_progname="${_t##*/[0-9][0-9].}"
nfs_check_service "$_progname" <"$_f"
done
}
######################################################
# Check the health of an NFS service
#
# $1 - progname, passed to rpcinfo (looked up in /etc/rpc)
#
# Reads variables from stdin
#
# Variables are:
#
# * family - "tcp" or "udp" or space separated list
# default: tcp, not used with "service_check_cmd"
# * version - optional, RPC service version number
# default is to omit to check for any version,
# not used with "service_check_cmd"
# * unhealthy_after - number of check fails before unhealthy
# default: 1
# * restart_every - number of check fails before restart
# default: 0, meaning no restart
# * service_stop_cmd - command to stop service
# default: no default, must be provided if
# restart_every > 0
# * service_start_cmd - command to start service
# default: no default, must be provided if
# restart_every > 0
# * service_check_cmd - command to check health of service
# default is to check RPC service using rpcinfo
# * service_debug_cmd - command to debug a service after trying to stop it;
# for example, it can be useful to print stack
# traces of threads that have not exited, since
# they may be stuck doing I/O;
# no default, see also function program_stack_traces()
#
# Quoting in values is not preserved
#
######################################################
nfs_check_service ()
{
_progname="$1"
(
# Subshell to restrict scope variables...
# Defaults
family="tcp"
version=""
unhealthy_after=1
restart_every=0
service_stop_cmd=""
service_start_cmd=""
service_check_cmd=""
service_debug_cmd=""
# Eval line-by-line. Expands variable references in values.
# Also allows variable name checking, which seems useful.
while read _line ; do
case "$_line" in
\#*|"") : ;; # Ignore comments, blank lines
family=*|version=*|\
unhealthy_after=*|restart_every=*|\
service_stop_cmd=*|service_start_cmd=*|\
service_check_cmd=*|service_debug_cmd=*)
eval "$_line"
;;
*)
echo "ERROR: Unknown variable for ${_progname}: ${_line}"
exit 1
esac
done
_service_name="nfs_${_progname}"
_ok=false
if [ -n "$service_check_cmd" ] ; then
# Using eval means variables can contain semicolon separated commands
if eval "$service_check_cmd" ; then
_ok=true
fi
else
if nfs_check_rpcinfo \
"$_progname" "$version" "$family" >/dev/null ; then
_ok=true
fi
fi
if $_ok ; then
if [ $unhealthy_after -ne 1 -o $restart_every -ne 0 ] ; then
ctdb_counter_init "$_service_name"
fi
exit 0
fi
ctdb_counter_incr "$_service_name"
_failcount=$(ctdb_counter_get "$_service_name")
_unhealthy=false
if [ $unhealthy_after -gt 0 ] ; then
if [ $_failcount -ge $unhealthy_after ] ; then
_unhealthy=true
echo "ERROR: $ctdb_check_rpc_out"
fi
fi
if [ $restart_every -gt 0 ] ; then
if [ $(($_failcount % $restart_every)) -eq 0 ] ; then
if ! $_unhealthy ; then
echo "WARNING: $ctdb_check_rpc_out"
fi
nfs_restart_service
fi
fi
if $_unhealthy ; then
exit 1
fi
return 0
) || exit 1
}
# Uses: stop_service, start_service, debug_stuck_threads
nfs_restart_service ()
{
if [ -z "$service_stop_cmd" -o -z "$service_start_cmd" ] ; then
die "ERROR: Can not restart service \"${_progname}\" without corresponding service_start_cmd/service_stop_cmd settings"
fi
echo "Trying to restart service \"${_progname}\"..."
# Using eval means variables can contain semicolon separated commands
eval "$service_stop_cmd"
if [ -n "$service_debug_cmd" ] ; then
eval "$service_debug_cmd"
fi
background_with_logging eval "$service_start_cmd"
}
######################################################
# Check an RPC service with rpcinfo
######################################################
ctdb_check_rpc ()
{
_progname="$1" # passed to rpcinfo (looked up in /etc/rpc)
_version="$2" # optional, not passed if empty/unset
_family="${3:-tcp}" # optional, default is "tcp"
2015-07-15 13:15:46 +03:00
case "$_family" in
tcp6|udp6)
_localhost="${CTDB_RPCINFO_LOCALHOST6:-::1}"
;;
*)
_localhost="${CTDB_RPCINFO_LOCALHOST:-127.0.0.1}"
esac
2015-07-13 08:22:23 +03:00
if ! ctdb_check_rpc_out=$(rpcinfo -T $_family $_localhost \
$_progname $_version 2>&1) ; then
ctdb_check_rpc_out="$_progname failed RPC check:
$ctdb_check_rpc_out"
echo "$ctdb_check_rpc_out"
return 1
fi
}
nfs_check_rpcinfo ()
{
_progname="$1" # passed to rpcinfo (looked up in /etc/rpc)
_versions="$2" # optional, space separated, not passed if empty/unset
_families="${3:-tcp}" # optional, space separated, default is "tcp"
for _family in $_families ; do
if [ -n "$_versions" ] ; then
for _version in $_versions ; do
ctdb_check_rpc $_progname $_version $_family || return $?
done
else
ctdb_check_rpc $_progname "" $_family || return $?
fi
done
}
##################################################################
# use statd-callout to update NFS lock info
##################################################################
nfs_update_lock_info ()
{
if [ -x "$CTDB_BASE/statd-callout" ] ; then
"$CTDB_BASE/statd-callout" update
fi
}
######################################################################
2009-11-19 07:00:17 +03:00
ctdb_start_stop_service
2008-02-11 01:35:37 +03:00
2010-11-18 05:23:40 +03:00
is_ctdb_managed_service || exit 0
2010-12-16 01:45:28 +03:00
ctdb_service_check_reconfigure
2015-06-24 14:36:14 +03:00
case "$1" in
startup)
nfs_callout "$@"
2007-06-02 10:44:15 +04:00
;;
2015-06-24 14:36:14 +03:00
shutdown)
nfs_callout "$@"
2007-07-06 04:54:42 +04:00
;;
2015-06-24 14:36:14 +03:00
takeip)
nfs_callout "$@"
2009-11-19 07:00:17 +03:00
ctdb_service_set_reconfigure
2007-06-01 18:10:22 +04:00
;;
2015-06-24 14:36:14 +03:00
releaseip)
nfs_callout "$@"
2009-11-19 07:00:17 +03:00
ctdb_service_set_reconfigure
2007-06-01 18:10:22 +04:00
;;
2015-06-24 14:36:14 +03:00
monitor)
nfs_callout "monitor-pre" || exit $?
# Check that directories for shares actually exist
if [ "$CTDB_NFS_SKIP_SHARE_CHECK" != "yes" ] ; then
nfs_callout "monitor-list-shares" | ctdb_check_directories || \
exit $?
fi
2011-08-11 01:13:28 +04:00
2010-08-26 08:59:59 +04:00
update_tickles 2049
2015-02-13 12:55:43 +03:00
nfs_update_lock_info
2010-08-26 08:59:59 +04:00
2015-06-19 09:35:12 +03:00
nfs_check_services
2010-08-30 12:13:28 +04:00
2015-06-24 14:36:14 +03:00
nfs_callout "monitor-post" || exit $?
2008-07-10 02:05:34 +04:00
;;
2007-06-06 06:08:42 +04:00
2009-12-01 09:43:47 +03:00
*)
ctdb_standard_event_handler "$@"
2009-11-19 07:00:17 +03:00
;;
2007-06-01 18:10:22 +04:00
esac
exit 0