3 # Run commands on CTDB nodes.
5 # See http://ctdb.samba.org/ for more information about CTDB.
7 # Copyright (C) Martin Schwenke 2008
9 # Based on an earlier script by Andrew Tridgell and Ronnie Sahlberg.
11 # Copyright (C) Andrew Tridgell 2007
13 # This program is free software; you can redistribute it and/or modify
14 # it under the terms of the GNU General Public License as published by
15 # the Free Software Foundation; either version 3 of the License, or
16 # (at your option) any later version.
18 # This program is distributed in the hope that it will be useful,
19 # but WITHOUT ANY WARRANTY; without even the implied warranty of
20 # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
21 # GNU General Public License for more details.
23 # You should have received a copy of the GNU General Public License
24 # along with this program; if not, see <http://www.gnu.org/licenses/>.
31 Usage: onnode [OPTION] ... <NODES> <COMMAND> ...
33 -c Run in current working directory on specified nodes.
34 -o <prefix> Save standard output from each node to file <prefix>.<ip>
35 -p Run command in parallel on specified nodes.
36 -q Do not print node addresses (overrides -v).
37 -n Allow nodes to be specified by name.
38 -f Specify nodes file, overrides CTDB_NODES_FILE.
39 -v Print node address even for a single node.
40 <NODES> "all", "any", "ok" (or "healthy"), "con" (or "connected"),
41 "rm" (or "recmaster"), "lvs" (or "lvsmaster"),
42 "natgw" (or "natgwlist"); or
43 a node number (0 base); or
44 a hostname (if -n is specified); or
45 list (comma separated) of <NODES>; or
46 range (hyphen separated) of node numbers.
54 echo "Invalid <nodespec>" >&2 ; echo >&2
68 # $POSIXLY_CORRECT means that the command passed to onnode can
69 # take options and getopt won't reorder things to make them
71 local temp=$(POSIXLY_CORRECT=1 getopt -n "$prog" -o "cf:hno:pqv" -l help -- "$@")
79 -c) current=true ; shift ;;
80 -f) CTDB_NODES_FILE="$2" ; shift 2 ;;
81 -n) names_ok=true ; shift ;;
82 -o) prefix="$2" ; shift 2 ;;
83 -p) parallel=true ; shift ;;
84 -q) quiet=true ; shift ;;
85 -v) verbose=true ; shift ;;
87 -h|--help|*) usage ;; # Shouldn't happen, so this is reasonable.
104 if [ -n "$node" -a "$node" != "#DEAD" ] ; then
107 echo "${prog}: \"node ${n}\" does not exist" >&2
114 # Subshell avoids hacks to restore $IFS.
119 *-*) seq "${i%-*}" "${i#*-}" 2>/dev/null || invalid_nodespec ;;
120 # Separate lines for readability.
121 all|any|ok|healthy|con|connected) echo "$i" ;;
122 rm|recmaster|lvs|lvsmaster|natgw|natgwlist) echo "$i" ;;
124 [ $i -gt -1 ] 2>/dev/null || $names_ok || invalid_nodespec
131 ctdb_status_output="" # cache
132 get_nodes_with_status ()
143 bits="0:[0-1]:[0-1]:[0-1]:[0-1]"
149 if [ -z "$ctdb_status_output" ] ; then
150 # FIXME: need to do something if $CTDB_NODES_SOCKETS is set.
151 ctdb_status_output=$(ctdb -Y status 2>/dev/null)
152 if [ $? -ne 0 ] ; then
153 echo "${prog}: unable to get status of CTDB nodes" >&2
156 ctdb_status_output="${ctdb_status_output#* }"
161 for i in $ctdb_status_output ; do
162 # Try removing bits from end.
163 local t="${i%:${bits}:}"
164 if [ "$t" != "$i" ] ; then
165 # Succeeded. Get address. NOTE: this is an optimisation.
166 # It might be better to get the node number and then get
167 # the nth node to get the address. This would make things
168 # more consistent if /etc/ctdb/nodes actually contained
170 nodes="${nodes} ${t##*:}"
177 ctdb_props="" # cache
178 get_node_with_property ()
184 if [ "${ctdb_props##:${prop}:}" = "$ctdb_props" ] ; then
185 prop_node=$(ctdb "$prop" -Y 2>/dev/null)
186 # We only want the first line.
189 prop_node="${prop_node%%${nl}*}"
190 if [ $? -eq 0 ] ; then
191 ctdb_props="${ctdb_props}${ctdb_props:+ }:${prop}:${prop_node}"
196 prop_node="${ctdb_props##:${prop}:}"
197 prop_node="${prop_node%% *}"
199 if [ -n "$prop_node" ] ; then
200 echo_nth "$prop_node" $all_nodes
202 echo "${prog}: No ${prop} available" >&2
207 get_any_available_node ()
211 # We do a recursive onnode to find which nodes are up and running.
212 local out=$($0 -pq all ctdb pnn 2>&1)
215 local pnn="${line#PNN:}"
216 if [ "$pnn" != "$line" ] ; then
217 echo_nth "$pnn" $all_nodes
220 # Else must be an error message from a down node.
229 if [ -n "$CTDB_NODES_SOCKETS" ] ; then
230 all_nodes="$CTDB_NODES_SOCKETS"
232 [ -e "$CTDB_NODES_FILE" ] || CTDB_NODES_FILE=/etc/ctdb/nodes
233 all_nodes=$(sed -e 's@#.*@@g' -e 's@ *@@g' -e 's@^$@#DEAD@' $CTDB_NODES_FILE)
238 for n in $(parse_nodespec "$1") ; do
239 [ $? != 0 ] && exit 1 # Required to catch exit in above subshell.
242 echo "${all_nodes//#DEAD/}"
245 get_any_available_node "$all_nodes" || exit 1
248 get_nodes_with_status "$all_nodes" "healthy" || exit 1
251 get_nodes_with_status "$all_nodes" "connected" || exit 1
254 get_node_with_property "$all_nodes" "recmaster" || exit 1
257 get_node_with_property "$all_nodes" "lvsmaster" || exit 1
260 get_node_with_property "$all_nodes" "natgwlist" || exit 1
262 [0-9]|[0-9][0-9]|[0-9][0-9][0-9])
263 echo_nth $n $all_nodes
266 $names_ok || invalid_nodespec
274 CTDB_SOCKET="$1" sh -c "$2" 3>/dev/null
279 if [ -n "$prefix" ] ; then
280 cat >"${prefix}.${n//\//_}"
281 elif $verbose && $parallel ; then
290 if $verbose && $parallel ; then
297 ######################################################################
301 $current && command="cd $PWD && $command"
304 if [ -n "$CTDB_NODES_SOCKETS" ] ; then
307 # Could "2>/dev/null || true" but want to see errors from typos in file.
308 [ -r /etc/ctdb/onnode.conf ] && . /etc/ctdb/onnode.conf
309 [ -n "$SSH" ] || SSH=ssh
310 if [ "$SSH" = "ssh" ] ; then
313 : # rsh? All bets are off!
317 ######################################################################
319 nodes=$(get_nodes "$nodespec")
320 [ $? != 0 ] && exit 1 # Required to catch exit in above subshell.
325 # If $nodes contains a space or a newline then assume multiple nodes.
328 [ "$nodes" != "${nodes%[ ${nl}]*}" ] && verbose=true
332 trap 'kill -TERM $pids 2>/dev/null' INT TERM
333 # There's a small race here where the kill can fail if no processes
334 # have been added to $pids and the script is interrupted. However,
335 # the part of the window where it matter is very small.
338 set -o pipefail 2>/dev/null
340 { exec 3>&1 ; { $SSH $ssh_opts $EXTRA_SSH_OPTS $n "$command" | stdout_filter >&3 ; } 2>&1 | stderr_filter ; } &
344 echo >&2 ; echo ">> NODE: $n <<" >&2
347 { exec 3>&1 ; { $SSH $ssh_opts $EXTRA_SSH_OPTS $n "$command" | stdout_filter >&3 ; } 2>&1 | stderr_filter ; }
348 [ $? = 0 ] || retcode=$?
355 [ $? = 0 ] || retcode=$?