# utility functions for ctdb event scripts
-[ -z "$CTDB_VARDIR" ] && {
- if [ -d "/var/lib/ctdb" ] ; then
- export CTDB_VARDIR="/var/lib/ctdb"
- else
- export CTDB_VARDIR="/var/ctdb"
- fi
-}
-[ -z "$CTDB_ETCDIR" ] && {
- export CTDB_ETCDIR="/etc"
-}
+if [ -z "$CTDB_BASE" ] ; then
+ echo 'CTDB_BASE unset in CTDB functions file'
+ exit 1
+fi
+
+CTDB_VARDIR="/usr/local/var/lib/ctdb"
+ctdb_rundir="/usr/local/var/run/ctdb"
+
+# Only (and always) override these variables in test code
+
+if [ -z "$CTDB_SCRIPT_VARDIR" ] ; then
+ CTDB_SCRIPT_VARDIR="/usr/local/var/lib/ctdb/state"
+fi
+
+if [ -z "$CTDB_SYS_ETCDIR" ] ; then
+ CTDB_SYS_ETCDIR="/etc"
+fi
#######################################
# pull in a system config file, if any
+
+rewrite_ctdb_options ()
+{
+ case "$CTDB_DBDIR" in
+ tmpfs|tmpfs:*)
+ _opts_defaults="mode=700"
+ # Get any extra options specified after colon
+ if [ "$CTDB_DBDIR" = "tmpfs" ] ; then
+ _opts=""
+ else
+ _opts="${CTDB_DBDIR#tmpfs:}"
+ fi
+ # This is an internal variable, only used by ctdbd_wrapper.
+ # It is OK to repeat mount options - last value wins
+ CTDB_DBDIR_TMPFS_OPTIONS="${_opts_defaults}${_opts:+,}${_opts}"
+
+ CTDB_DBDIR="${ctdb_rundir}/CTDB_DBDIR"
+ ;;
+ *)
+ CTDB_DBDIR_TMPFS_OPTIONS=""
+ esac
+}
+
_loadconfig() {
if [ -z "$1" ] ; then
return
fi
- if [ -f $CTDB_ETCDIR/sysconfig/$1 ]; then
- . $CTDB_ETCDIR/sysconfig/$1
- elif [ -f $CTDB_ETCDIR/default/$1 ]; then
- . $CTDB_ETCDIR/default/$1
+ if [ -f $CTDB_SYS_ETCDIR/sysconfig/$1 ]; then
+ . $CTDB_SYS_ETCDIR/sysconfig/$1
+ elif [ -f $CTDB_SYS_ETCDIR/default/$1 ]; then
+ . $CTDB_SYS_ETCDIR/default/$1
elif [ -f $CTDB_BASE/sysconfig/$1 ]; then
. $CTDB_BASE/sysconfig/$1
fi
if [ -r "$_config" ] ; then
. "$_config"
fi
+ rewrite_ctdb_options
fi
}
if [ -n "$CTDB_LOGGING" ] ; then
_file="${CTDB_LOGGING#file:}"
else
- _file="/var/log/log.ctdb"
+ _file="/usr/local/var/log/log.ctdb"
fi
{
if [ -n "$*" ] ; then
$_nice /sbin/service "$_service_name" "$_op"
elif [ -x /usr/sbin/service ]; then
$_nice /usr/sbin/service "$_service_name" "$_op"
- elif [ -x $CTDB_ETCDIR/init.d/$_service_name ]; then
- $_nice $CTDB_ETCDIR/init.d/$_service_name "$_op"
- elif [ -x $CTDB_ETCDIR/rc.d/init.d/$_service_name ]; then
- $_nice $CTDB_ETCDIR/rc.d/init.d/$_service_name "$_op"
+ elif [ -x $CTDB_SYS_ETCDIR/init.d/$_service_name ]; then
+ $_nice $CTDB_SYS_ETCDIR/init.d/$_service_name "$_op"
+ elif [ -x $CTDB_SYS_ETCDIR/rc.d/init.d/$_service_name ]; then
+ $_nice $CTDB_SYS_ETCDIR/rc.d/init.d/$_service_name "$_op"
fi
}
_service "$@"
}
+######################################################
+# Cached retrieval of PNN from local node. This never changes so why
+# open a client connection to the server each time this is needed?
+# This sets $pnn - this avoid an unnecessary subprocess.
+ctdb_get_pnn ()
+{
+ _pnn_file="${CTDB_SCRIPT_VARDIR}/my-pnn"
+ if [ ! -f "$_pnn_file" ] ; then
+ ctdb pnn | sed -e 's@.*:@@' >"$_pnn_file"
+ fi
+
+ read pnn <"$_pnn_file"
+}
+
######################################################
# wrapper around /proc/ settings to allow them to be hooked
# for testing
echo "$2" >"/proc/$1"
}
+set_proc_maybe ()
+{
+ if [ -w "/proc/$1" ] ; then
+ set_proc "$1" "$2"
+ fi
+}
+
######################################################
# wrapper around getting file contents from /proc/ to allow
# this to be hooked for testing
done
}
-######################################################
-# Check that an RPC service is healthy -
-# this includes allowing a certain number of failures
-# before marking the NFS service unhealthy.
-#
-# usage: nfs_check_rpc_service SERVICE_NAME [ triple ...]
-#
-# each triple is a set of 3 arguments: an operator, a
-# fail count limit and an action string.
-#
-# For example:
-#
-# nfs_check_rpc_service "lockd" \
-# -ge 15 "verbose restart unhealthy" \
-# -eq 10 "restart:bs"
-#
-# says that if lockd is down for 15 iterations then do
-# a verbose restart of lockd and mark the node unhealthy.
-# Before this, after 10 iterations of failure, the
-# service is restarted silently in the background.
-# Order is important: the number of failures need to be
-# specified in reverse order because processing stops
-# after the first condition that is true.
-######################################################
-nfs_check_rpc_service ()
-{
- _prog_name="$1" ; shift
-
- if _nfs_check_rpc_common "$_prog_name" ; then
- return
- fi
-
- while [ -n "$3" ] ; do
- if _nfs_check_rpc_action "$1" "$2" "$3" ; then
- break
- fi
- shift 3
- done
-}
-
-# The new way of doing things...
-nfs_check_rpc_services ()
-{
- # Files must end with .check - avoids editor backups, RPM fu, ...
- for _f in "${CTDB_BASE}/nfs-rpc-checks.d/"[0-9][0-9].*.check ; do
- _t="${_f%.check}"
- _prog_name="${_t##*/[0-9][0-9].}"
-
- if _nfs_check_rpc_common "$_prog_name" ; then
- # This RPC service is up, check next service...
- continue
- fi
-
- # Check each line in the file in turn until one of the limit
- # checks is hit...
- while read _cmp _lim _rest ; do
- # Skip comments
- case "$_cmp" in
- \#*) continue ;;
- esac
-
- if _nfs_check_rpc_action "$_cmp" "$_lim" "$_rest" ; then
- # Limit was hit on this line, no further checking...
- break
- fi
- done <"$_f"
- done
-}
-
-_nfs_check_rpc_common ()
-{
- _prog_name="$1"
-
- # Some platforms don't have separate programs for all services.
- case "$_prog_name" in
- statd)
- which "rpc.${_prog_name}" >/dev/null 2>&1 || return 0
- esac
-
- case "$_prog_name" in
- nfsd)
- _rpc_prog=nfs
- _version=3
- ;;
- mountd)
- _rpc_prog=mountd
- _version=1
- ;;
- rquotad)
- _rpc_prog=rquotad
- _version=1
- ;;
- lockd)
- _rpc_prog=nlockmgr
- _version=4
- ;;
- statd)
- _rpc_prog=status
- _version=1
- ;;
- *)
- echo "Internal error: unknown RPC program \"$_prog_name\"."
- exit 1
- esac
-
- _service_name="nfs_${_prog_name}"
-
- if ctdb_check_rpc "$_rpc_prog" $_version >/dev/null ; then
- ctdb_counter_init "$_service_name"
- return 0
- fi
-
- ctdb_counter_incr "$_service_name"
-
- return 1
-}
-
-_nfs_check_rpc_action ()
-{
- _cmp="$1"
- _limit="$2"
- _actions="$3"
-
- if ctdb_check_counter "quiet" "$_cmp" "$_limit" "$_service_name" ; then
- return 1
- fi
-
- for _action in $_actions ; do
- case "$_action" in
- verbose)
- echo "$ctdb_check_rpc_out"
- ;;
- restart)
- _nfs_restart_rpc_service "$_prog_name"
- ;;
- restart:b)
- _nfs_restart_rpc_service "$_prog_name" true
- ;;
- unhealthy)
- exit 1
- ;;
- *)
- echo "Internal error: unknown action \"$_action\"."
- exit 1
- esac
- done
-
- return 0
-}
-
-_nfs_restart_rpc_service ()
-{
- _prog_name="$1"
- _background="${2:-false}"
-
- if $_background ; then
- _maybe_background="background_with_logging"
- else
- _maybe_background=""
- fi
-
- _p="rpc.${_prog_name}"
-
- case "$_prog_name" in
- nfsd)
- echo "Trying to restart NFS service"
- $_maybe_background startstop_nfs restart
- ;;
- mountd)
- echo "Trying to restart $_prog_name [${_p}]"
- killall -q -9 "$_p"
- nfs_dump_some_threads "$_p"
- $_maybe_background $_p ${MOUNTD_PORT:+-p} $MOUNTD_PORT
- ;;
- rquotad)
- echo "Trying to restart $_prog_name [${_p}]"
- killall -q -9 "$_p"
- nfs_dump_some_threads "$_p"
- $_maybe_background $_p ${RQUOTAD_PORT:+-p} $RQUOTAD_PORT
- ;;
- lockd)
- echo "Trying to restart lock manager service"
- $_maybe_background startstop_nfslock restart
- ;;
- statd)
- echo "Trying to restart $_prog_name [${_p}]"
- killall -q -9 "$_p"
- nfs_dump_some_threads "$_p"
- $_maybe_background $_p \
- ${STATD_HOSTNAME:+-n} $STATD_HOSTNAME \
- ${STATD_PORT:+-p} $STATD_PORT \
- ${STATD_OUTGOING_PORT:+-o} $STATD_OUTGOING_PORT
- ;;
- *)
- echo "Internal error: unknown RPC program \"$_prog_name\"."
- exit 1
- esac
-}
-
-######################################################
-# check that a rpc server is registered with portmap
-# and responding to requests
-# usage: ctdb_check_rpc SERVICE_NAME VERSION
-######################################################
-ctdb_check_rpc ()
-{
- progname="$1"
- version="$2"
-
- _localhost="${CTDB_RPCINFO_LOCALHOST:-127.0.0.1}"
-
- if ! ctdb_check_rpc_out=$(rpcinfo -u $_localhost $progname $version 2>&1) ; then
- ctdb_check_rpc_out="ERROR: $progname failed RPC check:
-$ctdb_check_rpc_out"
- echo "$ctdb_check_rpc_out"
- return 1
- fi
-}
-
######################################################
# Ensure $service_name is set
assert_service_name ()
_ctdb_check_tcp_common ()
{
assert_service_name
- _ctdb_service_started_file="$ctdb_fail_dir/$service_name.started"
+ _d="${CTDB_SCRIPT_VARDIR}/failcount"
+ _ctdb_service_started_file="${_d}/${service_name}.started"
}
ctdb_check_tcp_init ()
{print $4" "$5}'
}
-##################################################################
-# use statd-callout to update NFS lock info
-##################################################################
-nfs_update_lock_info ()
-{
- if [ -x "$CTDB_BASE/statd-callout" ] ; then
- "$CTDB_BASE/statd-callout" update
- fi
-}
-
-########################################################
-# start/stop the Ganesha nfs service
-########################################################
-startstop_ganesha()
-{
- _service_name="nfs-ganesha-$CTDB_CLUSTER_FILESYSTEM_TYPE"
- case "$1" in
- start)
- service "$_service_name" start
- ;;
- stop)
- service "$_service_name" stop
- ;;
- restart)
- service "$_service_name" stop
- nfs_dump_some_threads "rpc.statd"
- service "$_service_name" start
- ;;
- esac
-}
-
-########################################################
-# start/stop the nfs service on different platforms
-########################################################
-startstop_nfs() {
- PLATFORM="unknown"
- [ -x $CTDB_ETCDIR/init.d/nfsserver ] && {
- PLATFORM="sles"
- }
- [ -x $CTDB_ETCDIR/init.d/nfslock -o \
- -r /usr/lib/systemd/system/nfs-lock.service ] && {
- PLATFORM="rhel"
- }
-
- case $PLATFORM in
- sles)
- case $1 in
- start)
- service nfsserver start
- ;;
- stop)
- service nfsserver stop > /dev/null 2>&1
- ;;
- restart)
- set_proc "fs/nfsd/threads" 0
- service nfsserver stop > /dev/null 2>&1
- pkill -9 nfsd
- nfs_dump_some_threads
- service nfsserver start
- ;;
- esac
- ;;
- rhel)
- case $1 in
- start)
- service nfslock start
- service nfs start
- ;;
- stop)
- service nfs stop
- service nfslock stop
- ;;
- restart)
- set_proc "fs/nfsd/threads" 0
- service nfs stop > /dev/null 2>&1
- service nfslock stop > /dev/null 2>&1
- pkill -9 nfsd
- nfs_dump_some_threads
- service nfslock start
- service nfs start
- ;;
- esac
- ;;
- *)
- echo "Unknown platform. NFS is not supported with ctdb"
- exit 1
- ;;
- esac
-}
-
-# Dump up to the configured number of nfsd thread backtraces.
-nfs_dump_some_threads ()
-{
- _prog="${1:-nfsd}"
-
- _num="${CTDB_NFS_DUMP_STUCK_THREADS:-5}"
- [ $_num -gt 0 ] || return 0
-
- program_stack_traces "$_prog" $_num
-}
-
-########################################################
-# start/stop the nfs lockmanager service on different platforms
-########################################################
-startstop_nfslock() {
- PLATFORM="unknown"
- [ -x $CTDB_ETCDIR/init.d/nfsserver ] && {
- PLATFORM="sles"
- }
- [ -x $CTDB_ETCDIR/init.d/nfslock -o \
- -r /usr/lib/systemd/system/nfs-lock.service ] && {
- PLATFORM="rhel"
- }
-
- case $PLATFORM in
- sles)
- # for sles there is no service for lockmanager
- # so we instead just shutdown/restart nfs
- case $1 in
- start)
- service nfsserver start
- ;;
- stop)
- service nfsserver stop > /dev/null 2>&1
- ;;
- restart)
- service nfsserver stop > /dev/null 2>&1
- service nfsserver start
- ;;
- esac
- ;;
- rhel)
- case $1 in
- start)
- service nfslock start
- ;;
- stop)
- service nfslock stop > /dev/null 2>&1
- ;;
- restart)
- service nfslock stop > /dev/null 2>&1
- service nfslock start
- ;;
- esac
- ;;
- *)
- echo "Unknown platform. NFS locking is not supported with ctdb"
- exit 1
- ;;
- esac
-}
-
########################################################
add_ip_to_iface ()
ip addr show to "${_addr}/${_bits}" 2>/dev/null | \
awk -v family="${_family}" \
- 'NR == 1 { iface = $2; sub(":$", "", iface) } \
+ 'NR == 1 { iface = $2; sub(":$", "", iface) ; \
+ sub("@.*", "", iface) } \
$1 ~ /inet/ { mask = $2; sub(".*/", "", mask); \
print mask, iface, family }'
}
flush_route_cache ()
{
- set_proc sys/net/ipv4/route/flush 1
- set_proc sys/net/ipv6/route/flush 1
+ set_proc_maybe sys/net/ipv4/route/flush 1
+ set_proc_maybe sys/net/ipv6/route/flush 1
}
########################################################
# Simple counters
_ctdb_counter_common () {
_service_name="${1:-${service_name:-${script_name}}}"
- _counter_file="$ctdb_fail_dir/$_service_name"
+ _counter_file="${CTDB_SCRIPT_VARDIR}/failcount/${_service_name}"
mkdir -p "${_counter_file%/*}" # dirname
}
ctdb_counter_init () {
# unary counting!
echo -n 1 >> "$_counter_file"
}
+ctdb_counter_get () {
+ _ctdb_counter_common "$1"
+ # unary counting!
+ stat -c "%s" "$_counter_file" 2>/dev/null || echo 0
+}
ctdb_check_counter () {
_msg="${1:-error}" # "error" - anything else is silent on fail
_op="${2:--ge}" # an integer operator supported by test
_limit="${3:-${service_fail_limit}}"
shift 3
- _ctdb_counter_common "$1"
- # unary counting!
- _size=$(stat -c "%s" "$_counter_file" 2>/dev/null || echo 0)
+ _size=$(ctdb_counter_get "$1")
+
_hit=false
if [ "$_op" != "%" ] ; then
if [ $_size $_op $_limit ] ; then
########################################################
-ctdb_status_dir="$CTDB_VARDIR/state/service_status"
-ctdb_fail_dir="$CTDB_VARDIR/state/failcount"
-
ctdb_setup_service_state_dir ()
{
- service_state_dir="$CTDB_VARDIR/state/service_state/${1:-${service_name}}"
+ service_state_dir="${CTDB_SCRIPT_VARDIR}/service_state/${1:-${service_name}}"
mkdir -p "$service_state_dir" || {
echo "Error creating state dir \"$service_state_dir\""
exit 1
########################################################
# Managed status history, for auto-start/stop
-ctdb_managed_dir="$CTDB_VARDIR/state/managed_history"
-
_ctdb_managed_common ()
{
- _ctdb_managed_file="$ctdb_managed_dir/$service_name"
+ _ctdb_managed_file="${CTDB_SCRIPT_VARDIR}/managed_history/${service_name}"
}
ctdb_service_managed ()
{
_ctdb_managed_common
- mkdir -p "$ctdb_managed_dir"
+ mkdir -p "${_ctdb_managed_file%/*}" # dirname
touch "$_ctdb_managed_file"
}
[ -f "$_ctdb_managed_file" ]
}
-########################################################
-# Check and set status
-
-log_status_cat ()
-{
- echo "node is \"$1\", \"${script_name}\" reports problem: $(cat $2)"
-}
-
-ctdb_checkstatus ()
-{
- if [ -r "$ctdb_status_dir/$script_name/unhealthy" ] ; then
- log_status_cat "unhealthy" "$ctdb_status_dir/$script_name/unhealthy"
- return 1
- elif [ -r "$ctdb_status_dir/$script_name/banned" ] ; then
- log_status_cat "banned" "$ctdb_status_dir/$script_name/banned"
- return 2
- else
- return 0
- fi
-}
-
-ctdb_setstatus ()
-{
- d="$ctdb_status_dir/$script_name"
- case "$1" in
- unhealthy|banned)
- mkdir -p "$d"
- cat "$2" >"$d/$1"
- ;;
- *)
- for i in "banned" "unhealthy" ; do
- rm -f "$d/$i"
- done
- ;;
- esac
-}
-
##################################################################
# Reconfigure a service on demand
_ctdb_service_reconfigure_common ()
{
- _d="$ctdb_status_dir/${service_name}"
+ _d="${CTDB_SCRIPT_VARDIR}/service_status/${service_name}"
mkdir -p "$_d"
_ctdb_service_reconfigure_flag="$_d/reconfigure"
}
ctdb_compat_managed_service "$CTDB_MANAGES_ISCSI" "iscsi"
ctdb_compat_managed_service "$CTDB_MANAGES_CLAMD" "clamd"
ctdb_compat_managed_service "$CTDB_MANAGES_NFS" "nfs"
- ctdb_compat_managed_service "$CTDB_MANAGES_NFS" "nfs-ganesha-gpfs"
t=" $CTDB_MANAGED_SERVICES "
ctdb_standard_event_handler ()
{
- case "$1" in
- status)
- ctdb_checkstatus
- exit
- ;;
- setstatus)
- shift
- ctdb_setstatus "$@"
- exit
- ;;
- esac
+ :
}
iptables_wrapper ()
fi
# iptables doesn't like being re-entered, so flock-wrap it.
- flock -w 30 "${CTDB_VARDIR}/iptables-ctdb.flock" "$_iptables_cmd" "$@"
+ flock -w 30 "${CTDB_SCRIPT_VARDIR}/iptables.flock" "$_iptables_cmd" "$@"
}
# AIX (and perhaps others?) doesn't have mktemp
-if ! which mktemp >/dev/null 2>&1 ; then
+if ! type mktemp >/dev/null 2>&1 ; then
mktemp ()
{
_dir=false
{
_port="$1"
- tickledir="$CTDB_VARDIR/state/tickles"
+ tickledir="${CTDB_SCRIPT_VARDIR}/tickles"
mkdir -p "$tickledir"
- # Who am I?
- _pnn=$(ctdb pnn) ; _pnn=${_pnn#PNN:}
+ ctdb_get_pnn
# What public IPs do I hold?
- _ips=$(ctdb -X ip | awk -F'|' -v pnn=$_pnn '$3 == pnn {print $2}')
+ _ips=$(ctdb -X ip | awk -F'|' -v pnn=$pnn '$3 == pnn {print $2}')
# IPs as a regexp choice
_ipschoice="($(echo $_ips | sed -e 's/ /|/g' -e 's/\./\\\\./g'))"
- # Record connections to our public IPs in a temporary file
- _my_connections="${tickledir}/${_port}.connections"
- rm -f "$_my_connections"
+ # Record connections to our public IPs in a temporary file.
+ # This temporary file is in CTDB's private state directory and
+ # $$ is used to avoid a very rare race involving CTDB's script
+ # debugging. No security issue, nothing to see here...
+ _my_connections="${tickledir}/${_port}.connections.$$"
netstat -tn |
awk -v destpat="^${_ipschoice}:${_port}\$" \
'$1 == "tcp" && $6 == "ESTABLISHED" && $4 ~ destpat {print $5, $4}' |
sort >"$_my_connections"
# Record our current tickles in a temporary file
- _my_tickles="${tickledir}/${_port}.tickles"
- rm -f "$_my_tickles"
+ _my_tickles="${tickledir}/${_port}.tickles.$$"
for _i in $_ips ; do
ctdb -X gettickles $_i $_port |
awk -F'|' 'NR > 1 { printf "%s:%s %s:%s\n", $2, $3, $4, $5 }'
ctdb deltickle $_src $_dst
done
- rm -f "$_my_connections" "$_my_tickles"
+ rm -f "$_my_connections" "$_my_tickles"
+
+ # Remove stale files from killed scripts
+ find "$tickledir" -type f -mmin +10 | xargs -r rm
}
########################################################