ctdb/tools/onnode

   1 #!/bin/bash
   2
   3 # Run commands on CTDB nodes.
   4
   5 # See http://ctdb.samba.org/ for more information about CTDB.
   6
   7 # Copyright (C) Martin Schwenke  2008
   8
   9 # Based on an earlier script by Andrew Tridgell and Ronnie Sahlberg.
  10
  11 # Copyright (C) Andrew Tridgell  2007
  12
  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.
  17
  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.
  22
  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/>.
  25
  26 prog=$(basename "$0")
  27
  28 usage ()
  29 {
  30     cat >&2 <<EOF
  31 Usage: onnode [OPTION] ... <NODES> <COMMAND> ...
  32   options:
  33     -c          Run in current working directory on specified nodes.
  34     -f          Specify nodes file, overrides CTDB_NODES_FILE.
  35     -i          Keep standard input open - the default is to close it.
  36     -n          Allow nodes to be specified by name.
  37     -o <prefix> Save standard output from each node to file <prefix>.<ip>
  38     -p          Run command in parallel on specified nodes.
  39     -P          Push given files to nodes instead of running commands.
  40     -q          Do not print node addresses (overrides -v).
  41     -v          Print node address even for a single node.
  42   <NODES>       "all", "any", "ok" (or "healthy"), "con" (or "connected") ; or
  43                 a node number (0 base); or
  44                 a hostname (if -n is specified); or
  45                 list (comma separated) of <NODES>; or
  46                 range (hyphen separated) of node numbers.
  47 EOF
  48     exit 1
  49
  50 }
  51
  52 invalid_nodespec ()
  53 {
  54     echo "Invalid <nodespec>" >&2 ; echo >&2
  55     usage
  56 }
  57
  58 # Defaults.
  59 current=false
  60 parallel=false
  61 verbose=false
  62 quiet=false
  63 prefix=""
  64 names_ok=false
  65 push=false
  66 stdin=false
  67
  68 if [ -z "$CTDB_BASE" ] ; then
  69     CTDB_BASE="/usr/local/etc/ctdb"
  70 fi
  71
  72 . "${CTDB_BASE}/functions"
  73 loadconfig "ctdb"
  74
  75 parse_options ()
  76 {
  77     # $POSIXLY_CORRECT means that the command passed to onnode can
  78     # take options and getopt won't reorder things to make them
  79     # options ot onnode.
  80     local temp
  81     # Not on the previous line - local returns 0!
  82     temp=$(POSIXLY_CORRECT=1 getopt -n "$prog" -o "cf:hno:pqvPi" -l help -- "$@")
  83
  84     [ $? != 0 ] && usage
  85
  86     eval set -- "$temp"
  87
  88     while true ; do
  89         case "$1" in
  90             -c) current=true ; shift ;;
  91             -f) CTDB_NODES_FILE="$2" ; shift 2 ;;
  92             -n) names_ok=true ; shift ;;
  93             -o) prefix="$2" ; shift 2 ;;
  94             -p) parallel=true ; shift ;;
  95             -q) quiet=true ; shift ;;
  96             -v) verbose=true ; shift ;;
  97             -P) push=true ; shift ;;
  98             -i) stdin=true ; shift ;;
  99             --) shift ; break ;;
 100             -h|--help|*) usage ;; # Shouldn't happen, so this is reasonable.
 101         esac
 102     done
 103
 104     [ $# -lt 2 ] && usage
 105
 106     nodespec="$1" ; shift
 107     command="$*"
 108 }
 109
 110 echo_nth ()
 111 {
 112     local n="$1" ; shift
 113
 114     shift "$n"
 115     local node="$1"
 116
 117     if [ -n "$node" -a "$node" != "#DEAD" ] ; then
 118         echo "$node"
 119     else
 120         echo "${prog}: \"node ${n}\" does not exist" >&2
 121         exit 1
 122     fi
 123 }
 124
 125 parse_nodespec ()
 126 {
 127     # Subshell avoids hacks to restore $IFS.
 128     (
 129         IFS=","
 130         for i in $1 ; do
 131             case "$i" in
 132                 *-*) seq "${i%-*}" "${i#*-}" 2>/dev/null || invalid_nodespec ;;
 133                 all|any|ok|healthy|con|connected) echo "$i" ;;
 134                 *)
 135                     [ "$i" -gt -1 ] 2>/dev/null || $names_ok || invalid_nodespec
 136                     echo "$i"
 137             esac
 138         done
 139     )
 140 }
 141
 142 ctdb_status_output="" # cache
 143 get_nodes_with_status ()
 144 {
 145     local all_nodes="$1"
 146     local status="$2"
 147
 148     if [ -z "$ctdb_status_output" ] ; then
 149         ctdb_status_output=$(ctdb -X status 2>&1)
 150         if [ $? -ne 0 ] ; then
 151             echo "${prog}: unable to get status of CTDB nodes" >&2
 152             echo "$ctdb_status_output" >&2
 153             exit 1
 154         fi
 155         local nl="
 156 "
 157         ctdb_status_output="${ctdb_status_output#*${nl}}"
 158     fi
 159
 160     (
 161         local i
 162         IFS="${IFS}|"
 163         while IFS="" read i ; do
 164
 165             # Intentional word splitting
 166             # shellcheck disable=SC2086
 167             set -- $i # split line on colons
 168             shift     # line starts with : so 1st field is empty
 169             local pnn="$1" ; shift
 170             shift # ignore IP address but need status bits below
 171
 172             case "$status" in
 173                 healthy)
 174                     # If any bit is 1, don't match this address.
 175                     local s
 176                     for s ; do
 177                         [ "$s" != "1" ] || continue 2
 178                     done
 179                     ;;
 180                 connected)
 181                     # If disconnected bit is not 0, don't match this address.
 182                     [ "$1" = "0" ] || continue
 183                     ;;
 184                 *)
 185                     invalid_nodespec
 186             esac
 187
 188             # Intentional multi-word expansion
 189             # shellcheck disable=SC2086
 190             echo_nth "$pnn" $all_nodes
 191         done <<<"$ctdb_status_output"
 192     )
 193 }
 194
 195 get_any_available_node ()
 196 {
 197     local all_nodes="$1"
 198
 199     # We do a recursive onnode to find which nodes are up and running.
 200     local out line
 201     out=$("$0" -pq all ctdb pnn 2>&1)
 202     while read line ; do
 203         if [[ "$line" =~ ^[0-9]+$ ]] ; then
 204             local pnn="$line"
 205             # Intentional multi-word expansion
 206             # shellcheck disable=SC2086
 207             echo_nth "$pnn" $all_nodes
 208             return 0
 209         fi
 210         # Else must be an error message from a down node.
 211     done <<<"$out"
 212     return 1
 213 }
 214
 215 get_nodes ()
 216 {
 217     local all_nodes
 218
 219     if [ -n "$CTDB_NODES_SOCKETS" ] ; then
 220         all_nodes="$CTDB_NODES_SOCKETS"
 221     else
 222         local f="${CTDB_BASE}/nodes"
 223         if [ -n "$CTDB_NODES_FILE" ] ; then
 224             f="$CTDB_NODES_FILE"
 225             if [ ! -e "$f" -a "${f#/}" = "$f" ] ; then
 226                 # $f is relative, try in $CTDB_BASE
 227                 f="${CTDB_BASE}/${f}"
 228             fi
 229         elif [ -n "$CTDB_NODES" ] ; then
 230             f="$CTDB_NODES"
 231         fi
 232
 233         if [ ! -r "$f" ] ; then
 234             echo "${prog}: unable to open nodes file  \"${f}\"" >&2
 235             exit 1
 236         fi
 237
 238         all_nodes=$(sed -e 's@#.*@@g' -e 's@ *@@g' -e 's@^$@#DEAD@' "$f")
 239     fi
 240
 241     local nodes=""
 242     local n
 243     for n in $(parse_nodespec "$1") ; do
 244         [ $? != 0 ] && exit 1  # Required to catch exit in above subshell.
 245         case "$n" in
 246             all)
 247                 echo "${all_nodes//#DEAD/}"
 248                 ;;
 249             any)
 250                 get_any_available_node "$all_nodes" || exit 1
 251                 ;;
 252             ok|healthy)
 253                 get_nodes_with_status "$all_nodes" "healthy" || exit 1
 254                 ;;
 255             con|connected)
 256                 get_nodes_with_status "$all_nodes" "connected" || exit 1
 257                 ;;
 258             [0-9]|[0-9][0-9]|[0-9][0-9][0-9])
 259                 # Intentional multi-word expansion
 260                 # shellcheck disable=SC2086
 261                 echo_nth "$n" $all_nodes
 262                 ;;
 263             *)
 264                 $names_ok || invalid_nodespec
 265                 echo "$n"
 266         esac
 267     done
 268 }
 269
 270 push()
 271 {
 272     local host="$1"
 273     local files="$2"
 274
 275     local f
 276     for f in $files ; do
 277         $verbose && echo "Pushing $f"
 278         case "$f" in
 279             /*) rsync "$f" "[${host}]:${f}" ;;
 280             *)  rsync "${PWD}/${f}" "[${host}]:${PWD}/${f}" ;;
 281         esac
 282     done
 283 }
 284
 285 fakessh ()
 286 {
 287     CTDB_SOCKET="$1" sh -c "$2" 3>/dev/null
 288 }
 289
 290 stdout_filter ()
 291 {
 292     if [ -n "$prefix" ] ; then
 293         cat >"${prefix}.${n//\//_}"
 294     elif $verbose && $parallel ; then
 295         sed -e "s@^@[$n] @"
 296     else
 297         cat
 298     fi
 299 }
 300
 301 stderr_filter ()
 302 {
 303     if $verbose && $parallel ; then
 304         sed -e "s@^@[$n] @"
 305     else
 306         cat
 307     fi
 308 }
 309
 310 ######################################################################
 311
 312 parse_options "$@"
 313
 314 ssh_opts=
 315 if $push ; then
 316     SSH=push
 317     EXTRA_SSH_OPTS=""
 318 else
 319     $current && command="cd $PWD && $command"
 320
 321     if [ -n "$CTDB_NODES_SOCKETS" ] ; then
 322         SSH=fakessh
 323         EXTRA_SSH_OPTS=""
 324     else
 325         # Could "2>/dev/null || true" but want to see errors from typos in file.
 326         [ -r "${CTDB_BASE}/onnode.conf" ] && . "${CTDB_BASE}/onnode.conf"
 327         [ -n "$SSH" ] || SSH=ssh
 328         if [ "$SSH" = "ssh" ] ; then
 329             if $parallel || ! $stdin ; then
 330                 ssh_opts="-n"
 331             fi
 332         else
 333             : # rsh? All bets are off!
 334         fi
 335     fi
 336 fi
 337
 338 ######################################################################
 339
 340 nodes=$(get_nodes "$nodespec")
 341 [ $? != 0 ] && exit 1   # Required to catch exit in above subshell.
 342
 343 if $quiet ; then
 344     verbose=false
 345 else
 346     # If $nodes contains a space or a newline then assume multiple nodes.
 347     nl="
 348 "
 349     [ "$nodes" != "${nodes%[ ${nl}]*}" ] && verbose=true
 350 fi
 351
 352 pids=""
 353 # Intentional multi-word expansion
 354 # shellcheck disable=SC2086
 355 trap 'kill -TERM $pids 2>/dev/null' INT TERM
 356 # There's a small race here where the kill can fail if no processes
 357 # have been added to $pids and the script is interrupted.  However,
 358 # the part of the window where it matter is very small.
 359 retcode=0
 360 for n in $nodes ; do
 361     set -o pipefail 2>/dev/null
 362     if $parallel ; then
 363         { exec 3>&1 ; { $SSH $ssh_opts $EXTRA_SSH_OPTS "$n" "$command" | stdout_filter >&3 ; } 2>&1 | stderr_filter ; } &
 364         pids="${pids} $!"
 365     else
 366         if $verbose ; then
 367             echo >&2 ; echo ">> NODE: $n <<" >&2
 368         fi
 369
 370         { exec 3>&1 ; { $SSH $ssh_opts $EXTRA_SSH_OPTS "$n" "$command" | stdout_filter >&3 ; } 2>&1 | stderr_filter ; }
 371         [ $? = 0 ] || retcode=$?
 372     fi
 373 done
 374
 375 $parallel && {
 376     for p in $pids; do
 377         wait "$p"
 378         [ $? = 0 ] || retcode=$?
 379     done
 380 }
 381
 382 exit $retcode