| 1 | #! @BASH@ |
| 2 | ### |
| 3 | ### Backup script |
| 4 | ### |
| 5 | ### (c) 2012 Mark Wooding |
| 6 | ### |
| 7 | |
| 8 | ###----- Licensing notice --------------------------------------------------- |
| 9 | ### |
| 10 | ### This file is part of the `rsync-backup' program. |
| 11 | ### |
| 12 | ### rsync-backup is free software; you can redistribute it and/or modify |
| 13 | ### it under the terms of the GNU General Public License as published by |
| 14 | ### the Free Software Foundation; either version 2 of the License, or |
| 15 | ### (at your option) any later version. |
| 16 | ### |
| 17 | ### rsync-backup is distributed in the hope that it will be useful, |
| 18 | ### but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 19 | ### MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
| 20 | ### GNU General Public License for more details. |
| 21 | ### |
| 22 | ### You should have received a copy of the GNU General Public License |
| 23 | ### along with rsync-backup; if not, write to the Free Software Foundation, |
| 24 | ### Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA. |
| 25 | |
| 26 | set -e |
| 27 | |
| 28 | thishost=$(hostname -s) |
| 29 | quis=${0##*/} |
| 30 | . @pkgdatadir@/lib.sh |
| 31 | |
| 32 | verbose=: |
| 33 | dryrun=nil |
| 34 | |
| 35 | ###-------------------------------------------------------------------------- |
| 36 | ### Utility functions. |
| 37 | |
| 38 | RSYNCOPTS="--verbose" |
| 39 | |
| 40 | do_rsync () { |
| 41 | ## Run rsync(1) in an appropriate manner. Configuration should ovrride |
| 42 | ## this or set $RSYNCOPTS if it wants to do something weirder. Arguments |
| 43 | ## to this function are passed on to rsync. |
| 44 | |
| 45 | rsync \ |
| 46 | --archive --hard-links --numeric-ids --del \ |
| 47 | --sparse --compress \ |
| 48 | --one-file-system \ |
| 49 | --partial \ |
| 50 | $RSYNCOPTS \ |
| 51 | --filter="dir-merge .rsync-backup" \ |
| 52 | "$@" |
| 53 | } |
| 54 | |
| 55 | log () { |
| 56 | case $dryrun in |
| 57 | t) |
| 58 | echo >&2 " *** $*" |
| 59 | ;; |
| 60 | nil) |
| 61 | now=$(date +"%Y-%m-%d %H:%M:%S %z") |
| 62 | echo >&9 "$now $*" |
| 63 | ;; |
| 64 | esac |
| 65 | } |
| 66 | |
| 67 | maybe () { |
| 68 | ## Run CMD, if this isn't a dry run. |
| 69 | |
| 70 | case $dryrun in |
| 71 | t) echo >&2 " +++ $*" ;; |
| 72 | nil) "$@" ;; |
| 73 | esac |
| 74 | } |
| 75 | |
| 76 | copy () { |
| 77 | prefix=$1 |
| 78 | ## Copy lines from stdin to stdout, adding PREFIX. |
| 79 | |
| 80 | while IFS= read -r line; do |
| 81 | printf "%s %s\n" "$prefix" "$line" |
| 82 | done |
| 83 | } |
| 84 | |
| 85 | run () { |
| 86 | tag=$1 cmd=$2; shift 2 |
| 87 | ## Run CMD, logging its output in a pleasing manner. |
| 88 | |
| 89 | case $dryrun in |
| 90 | t) |
| 91 | echo >&2 " *** RUN $tag" |
| 92 | echo >&2 " +++ $cmd $*" |
| 93 | rc=0 |
| 94 | ;; |
| 95 | nil) |
| 96 | log "BEGIN $tag" |
| 97 | rc=$( |
| 98 | { { { ( set +e |
| 99 | "$cmd" "$@" 3>&- 4>&- 5>&- 9>&- |
| 100 | echo $? >&5; ) | |
| 101 | copy "|" >&4; } 2>&1 | |
| 102 | copy "*" >&4; } 4>&1 | |
| 103 | cat >&9; } 5>&1 </dev/null |
| 104 | ) |
| 105 | case $rc in |
| 106 | 0) log "END $tag" ;; |
| 107 | *) log "FAIL $tag (rc = $rc)" ;; |
| 108 | esac |
| 109 | ;; |
| 110 | esac |
| 111 | return $rc |
| 112 | } |
| 113 | |
| 114 | localp () { |
| 115 | h=$1 |
| 116 | ## Answer whether H is a local host. |
| 117 | |
| 118 | case $h in |
| 119 | "$thishost") return 0 ;; |
| 120 | *) return 1 ;; |
| 121 | esac |
| 122 | } |
| 123 | |
| 124 | hostrun () { |
| 125 | tag=$1 cmd=$2 |
| 126 | ## Run CMD on the current host. If the host seems local then run the |
| 127 | ## command through a local shell; otherwise run it through ssh(1). Either |
| 128 | ## way it will be processed by a shell. |
| 129 | |
| 130 | if localp $host; then run "@$host: $tag" sh -c "$cmd" |
| 131 | else run "@$host: $tag" ssh $userat$host "$cmd" |
| 132 | fi |
| 133 | } |
| 134 | |
| 135 | _hostrun () { |
| 136 | h=$1 cmd=$2 |
| 137 | ## Like hostrun, but without the complicated logging, and targetted at a |
| 138 | ## specific host. |
| 139 | |
| 140 | if localp $h; then sh -c "$cmd" |
| 141 | else ssh $h "$cmd" |
| 142 | fi |
| 143 | } |
| 144 | |
| 145 | hostpath () { |
| 146 | path=$1 |
| 147 | ## Output (to stdout) either PATH or HOST:PATH, choosing the former if the |
| 148 | ## current host is local. |
| 149 | |
| 150 | if localp $host; then echo $path |
| 151 | else echo $userat$host:$path |
| 152 | fi |
| 153 | } |
| 154 | |
| 155 | defhook () { |
| 156 | hook=$1 |
| 157 | ## Define a hook called HOOK. |
| 158 | |
| 159 | eval hk_$hook= |
| 160 | } |
| 161 | |
| 162 | addhook () { |
| 163 | hook=$1 cmd=$2 |
| 164 | ## Add command CMD to the hook HOOK. |
| 165 | |
| 166 | eval old=\$hk_$hook; new="$old $cmd" |
| 167 | eval hk_$hook=\$new |
| 168 | } |
| 169 | |
| 170 | runhook () { |
| 171 | hook=$1; shift 1 |
| 172 | ## Invoke HOOK, passing it the remaining arguments. |
| 173 | |
| 174 | eval cmds=\$hk_$hook |
| 175 | for cmd in $cmds; do |
| 176 | if ! $cmd "$@"; then return $?; fi |
| 177 | done |
| 178 | } |
| 179 | |
| 180 | remove_old_logfiles () { |
| 181 | base=$1 |
| 182 | ## Remove old logfiles with names of the form BASE.DATE#N, so that there |
| 183 | ## are at most $MAXLOG of them. |
| 184 | |
| 185 | ## Count up the logfiles. |
| 186 | nlog=0 |
| 187 | for i in "$base".*; do |
| 188 | if [ ! -f "$i" ]; then continue; fi |
| 189 | nlog=$(( nlog + 1 )) |
| 190 | done |
| 191 | |
| 192 | ## If there are too many, go through and delete some early ones. |
| 193 | if [ $dryrun = nil ] && [ $nlog -gt $MAXLOG ]; then |
| 194 | n=$(( nlog - MAXLOG )) |
| 195 | for i in "$base".*; do |
| 196 | if [ ! -f "$i" ]; then continue; fi |
| 197 | rm -f "$i" |
| 198 | n=$(( n - 1 )) |
| 199 | if [ $n -eq 0 ]; then break; fi |
| 200 | done |
| 201 | fi |
| 202 | } |
| 203 | |
| 204 | ###-------------------------------------------------------------------------- |
| 205 | ### Database operations. |
| 206 | |
| 207 | insert_index () { |
| 208 | host=$1 fs=$2 date=$3 vol=$4 |
| 209 | |
| 210 | if [ -f "$INDEXDB" ]; then |
| 211 | sqlite3 "$INDEXDB" <<EOF |
| 212 | INSERT INTO idx (host, fs, date, vol) |
| 213 | VALUES ('$host', '$fs', '$date', '$vol'); |
| 214 | EOF |
| 215 | fi |
| 216 | } |
| 217 | |
| 218 | delete_index () { |
| 219 | host=$1 fs=$2 date=$3 |
| 220 | |
| 221 | if [ -f "$INDEXDB" ]; then |
| 222 | sqlite3 "$INDEXDB" <<EOF |
| 223 | DELETE FROM idx WHERE |
| 224 | host = '$host' AND fs = '$fs' AND date = '$date'; |
| 225 | EOF |
| 226 | fi |
| 227 | } |
| 228 | |
| 229 | ###-------------------------------------------------------------------------- |
| 230 | ### Snapshot handling. |
| 231 | |
| 232 | ## Snapshot protocol. Each snapshot type has a pair of functions snap_TYPE |
| 233 | ## and unsnap_TYPE. Each is given the current snapshot arguments and the |
| 234 | ## filesystem name to back up. The snap_TYPE function should create and |
| 235 | ## mount the snapshot and output an rsync(1) path to where the filesystem can |
| 236 | ## be copied; the unsnap_TYPE function should unmount and tear down the |
| 237 | ## snapshot. |
| 238 | |
| 239 | ## Fake snapshot by not doing anything. Use only if you have no choice. |
| 240 | snap_live () { hostpath "$2"; } |
| 241 | unsnap_live () { :; } |
| 242 | |
| 243 | ## Fake snapshot by remounting a live filesystem read-only. Useful if the |
| 244 | ## underlying storage isn't in LVM. |
| 245 | |
| 246 | snap_ro () { |
| 247 | fs=$1 mnt=$2 |
| 248 | |
| 249 | ## Place a marker in the filesystem so we know why it was made readonly. |
| 250 | ## (Also this serves to ensure that the filesystem was writable before.) |
| 251 | hostrun "snap-ro $mnt" " |
| 252 | echo rsync-backup >$mnt/.lock |
| 253 | mount -oremount,ro $mnt" || return $? |
| 254 | |
| 255 | ## Done. |
| 256 | hostpath $mnt |
| 257 | } |
| 258 | |
| 259 | unsnap_ro () { |
| 260 | fs=$1 mnt=$2 |
| 261 | |
| 262 | ## Check that the filesystem still has our lock marker. |
| 263 | hostrun "unsnap-ro $mnt" " |
| 264 | case \$(cat $mnt/.lock) in |
| 265 | rsync-backup) ;; |
| 266 | *) echo unlocked by someone else; exit 31 ;; |
| 267 | esac |
| 268 | mount -oremount,rw $mnt |
| 269 | rm $mnt/.lock" || return $? |
| 270 | } |
| 271 | |
| 272 | ## Snapshot using LVM. |
| 273 | |
| 274 | SNAPSIZE="-l10%ORIGIN" |
| 275 | |
| 276 | snap_lvm () { |
| 277 | vg=$1 lv=$2 |
| 278 | |
| 279 | ## Make the snapshot. |
| 280 | hostrun "snap-lvm $vg/$lv" " |
| 281 | lvcreate --snapshot -n$lv.bkp $SNAPSIZE $vg/$lv |
| 282 | mkdir -p $SNAPDIR/$lv |
| 283 | mount -oro /dev/$vg/$lv.bkp $SNAPDIR/$lv" || return $? |
| 284 | |
| 285 | ## Done. |
| 286 | hostpath $SNAPDIR/$lv |
| 287 | } |
| 288 | |
| 289 | unsnap_lvm () { |
| 290 | vg=$1 lv=$2 |
| 291 | |
| 292 | ## Remove the snapshot. Sometimes LVM doesn't notice that the snapshot is |
| 293 | ## no longer in open immdiately, so try several times. |
| 294 | hostrun "unsnap-lvm $vg/$lv" " |
| 295 | umount $SNAPDIR/$lv |
| 296 | rc=1 |
| 297 | for i in 1 2 3 4; do |
| 298 | if lvremove -f $vg/$lv.bkp; then rc=0; break; fi |
| 299 | sleep 2 |
| 300 | done |
| 301 | exit $rc" || return $? |
| 302 | } |
| 303 | |
| 304 | ## Complicated snapshot using LVM, where the volume group and filesystem are |
| 305 | ## owned by different machines, so they need to be synchronized during the |
| 306 | ## snapshot. |
| 307 | |
| 308 | do_rfreezefs () { |
| 309 | lvhost=$1 vg=$2 lv=$3 fshost=$4 fsdir=$5 |
| 310 | |
| 311 | ## Engage in the rfreezefs protocol with the filesystem host. This |
| 312 | ## involves some hairy plumbing. We want to get exit statuses out of both |
| 313 | ## halves. |
| 314 | set +e |
| 315 | ssh $fshost rfreezefs $fsdir | { |
| 316 | set -e |
| 317 | |
| 318 | ## Read the codebook from the remote end. |
| 319 | ready=nil |
| 320 | while read line; do |
| 321 | set -- $line |
| 322 | case "$1" in |
| 323 | PORT) port=$2 ;; |
| 324 | TOKEN) eval tok_$2=$3 ;; |
| 325 | READY) ready=t; break ;; |
| 326 | *) |
| 327 | echo >&2 "$quis: unexpected keyword $1 (rfreezefs to $rhost)" |
| 328 | exit 1 |
| 329 | ;; |
| 330 | esac |
| 331 | done |
| 332 | case $ready in |
| 333 | nil) |
| 334 | echo >&2 "$quis: unexpected eof (rfreezefs to $rhost)" |
| 335 | exit 1 |
| 336 | ;; |
| 337 | esac |
| 338 | |
| 339 | ## Connect to the filesystem host's TCP port and get it to freeze its |
| 340 | ## filesystem. |
| 341 | exec 3<>/dev/tcp/$fshost/$port |
| 342 | echo $tok_FREEZE >&3 |
| 343 | read tok <&3 |
| 344 | case $tok in |
| 345 | "$tok_FROZEN") ;; |
| 346 | *) |
| 347 | echo >&2 "$quis: unexpected token $tok (rfreezefs $fsdir on $fshost)" |
| 348 | exit 1 |
| 349 | ;; |
| 350 | esac |
| 351 | |
| 352 | ## Get the volume host to create the snapshot. |
| 353 | set +e |
| 354 | _hostrun >&2 3>&- $userat$lvhost \ |
| 355 | "lvcreate --snapshot -n$lv.bkp $SNAPSIZE $vg/$lv" |
| 356 | snaprc=$? |
| 357 | set -e |
| 358 | |
| 359 | ## The filesystem can thaw now. |
| 360 | echo $tok_THAW >&3 |
| 361 | read tok <&3 |
| 362 | case $tok in |
| 363 | "$tok_THAWED") ;; |
| 364 | *) |
| 365 | _hostrun >&2 3>&- $userat$lvhost "lvremove -f $vg/$lv.bkp" || : |
| 366 | echo >&2 "$quis: unexpected token $tok (rfreezefs $fsdir on $fshost)" |
| 367 | exit 1 |
| 368 | ;; |
| 369 | esac |
| 370 | |
| 371 | ## Done. |
| 372 | exit $snaprc |
| 373 | } |
| 374 | |
| 375 | ## Sift through the wreckage to find out what happened. |
| 376 | rc_rfreezefs=${PIPESTATUS[0]} rc_snapshot=${PIPESTATUS[1]} |
| 377 | set -e |
| 378 | case $rc_rfreezefs:$rc_snapshot in |
| 379 | 0:0) |
| 380 | ;; |
| 381 | 112:*) |
| 382 | echo >&2 "$quis: EMERGENCY failed to thaw $fsdir on $fshost!" |
| 383 | exit 112 |
| 384 | ;; |
| 385 | *) |
| 386 | echo >&2 "$quis: failed to snapshot $vg/$lv ($fsdir on $fshost)" |
| 387 | exit 1 |
| 388 | ;; |
| 389 | esac |
| 390 | |
| 391 | ## Mount the snapshot on the volume host. |
| 392 | _hostrun >&2 $userat$lvhost " |
| 393 | mkdir -p $SNAPDIR/$lv |
| 394 | mount -oro /dev/$vg/$lv.bkp $SNAPDIR/$lv" |
| 395 | } |
| 396 | |
| 397 | snap_rfreezefs () { |
| 398 | rhost=$1 vg=$2 lv=$3 rfs=$4 |
| 399 | |
| 400 | set -e |
| 401 | run "snap-rfreezefs $host:$vg/$lv $rhost:$rfs" \ |
| 402 | do_rfreezefs $host $vg $lv $rhost $rfs || return $? |
| 403 | hostpath $SNAPDIR/$lv |
| 404 | } |
| 405 | |
| 406 | unsnap_rfreezefs () { |
| 407 | |
| 408 | ## Unshapping is the same as for plain LVM. |
| 409 | rhost=$1 vg=$2 lv=$3 rfs=$4 |
| 410 | unsnap_lvm $vg $lv |
| 411 | } |
| 412 | |
| 413 | ###-------------------------------------------------------------------------- |
| 414 | ### Expiry computations. |
| 415 | |
| 416 | expire () { |
| 417 | ## Read dates on stdin; write to stdout `EXPIRE date' for dates which |
| 418 | ## should be expired and `RETAIN date' for dates which should be retained. |
| 419 | |
| 420 | ## Get the current date and convert it into useful forms. |
| 421 | now=$(date +%Y-%m-%d) |
| 422 | parsedate $now |
| 423 | now_jdn=$(julian $now) now_year=$year now_month=$month now_day=$day |
| 424 | kept=: |
| 425 | |
| 426 | ## Work through each date in the input. |
| 427 | while read date; do |
| 428 | keep=nil |
| 429 | |
| 430 | ## Convert the date into a useful form. |
| 431 | jdn=$(julian $date) |
| 432 | parsedate $date |
| 433 | |
| 434 | ## Work through the policy list. |
| 435 | if [ $jdn -le $now_jdn ]; then |
| 436 | while read ival age; do |
| 437 | |
| 438 | ## Decide whether the policy entry applies to this date. |
| 439 | apply=nil |
| 440 | case $age in |
| 441 | forever) |
| 442 | apply=t |
| 443 | ;; |
| 444 | year) |
| 445 | if [ $year -eq $now_year ] || |
| 446 | ([ $year -eq $(( $now_year - 1 )) ] && |
| 447 | [ $month -ge $now_month ]) |
| 448 | then apply=t; fi |
| 449 | ;; |
| 450 | month) |
| 451 | if ([ $month -eq $now_month ] && [ $year -eq $now_year ]) || |
| 452 | ((([ $month -eq $(( $now_month - 1 )) ] && |
| 453 | [ $year -eq $now_year ]) || |
| 454 | ([ $month -eq 12 ] && [ $now_month -eq 1 ] && |
| 455 | [ $year -eq $(( $now_year - 1 )) ])) && |
| 456 | [ $day -ge $now_day ]) |
| 457 | then apply=t; fi |
| 458 | ;; |
| 459 | week) |
| 460 | if [ $jdn -ge $(( $now_jdn - 7 )) ]; then apply=t; fi |
| 461 | ;; |
| 462 | *) |
| 463 | echo >&2 "$quis: unknown age symbol \`$age'" |
| 464 | exit 1 |
| 465 | ;; |
| 466 | esac |
| 467 | case $apply in nil) continue ;; esac |
| 468 | |
| 469 | ## Find the interval marker for this date. |
| 470 | case $ival in |
| 471 | daily) |
| 472 | marker=$date |
| 473 | ;; |
| 474 | weekly) |
| 475 | ydn=$(julian $year-01-01) |
| 476 | wk=$(( ($jdn - $ydn)/7 + 1 )) |
| 477 | marker=$year-w$wk |
| 478 | ;; |
| 479 | monthly) |
| 480 | marker=$year-$month |
| 481 | ;; |
| 482 | annually | yearly) |
| 483 | marker=$year |
| 484 | ;; |
| 485 | *) |
| 486 | echo >&2 "$quis: unknown interval symbol \`$ival'" |
| 487 | exit 1 |
| 488 | ;; |
| 489 | esac |
| 490 | |
| 491 | ## See if we've alredy retained something in this interval. |
| 492 | case $kept in |
| 493 | *:"$marker":*) ;; |
| 494 | *) keep=t kept=$kept$marker: ;; |
| 495 | esac |
| 496 | |
| 497 | done <<EOF |
| 498 | $expire_policy |
| 499 | EOF |
| 500 | fi |
| 501 | |
| 502 | case $keep in |
| 503 | t) echo RETAIN $date ;; |
| 504 | *) echo EXPIRE $date ;; |
| 505 | esac |
| 506 | |
| 507 | done |
| 508 | } |
| 509 | |
| 510 | ###-------------------------------------------------------------------------- |
| 511 | ### Actually taking backups of filesystems. |
| 512 | |
| 513 | MAXLOG=14 |
| 514 | HASH=sha256 |
| 515 | unset VOLUME |
| 516 | |
| 517 | bkprc=0 |
| 518 | |
| 519 | remote_fshash () { |
| 520 | _hostrun $userat$host " |
| 521 | umask 077 |
| 522 | mkdir -p $fshashdir |
| 523 | cd ${snapmnt#*:} |
| 524 | echo \"*** $host $fs $date\"; echo |
| 525 | rsync -rx --filter='dir-merge .rsync-backup' ./ | |
| 526 | fshash -c$fshashdir/$fs.bkp -a -H$HASH -frsync |
| 527 | " >new.fshash |
| 528 | } |
| 529 | |
| 530 | local_fshash () { |
| 531 | { echo "*** $host $fs $date"; echo |
| 532 | fshash -c$STOREDIR/fshash.cache -H$HASH new/ |
| 533 | } >$localmap |
| 534 | } |
| 535 | |
| 536 | expire_backups () { |
| 537 | { seen=: |
| 538 | for i in *-*-*; do |
| 539 | i=${i%%.*} |
| 540 | case $i in *[!-0-9]*) continue ;; esac |
| 541 | case $seen in *:"$i":*) continue ;; esac |
| 542 | seen=$seen$i: |
| 543 | echo $i |
| 544 | done; } | |
| 545 | expire | |
| 546 | while read op date; do |
| 547 | case $op,$dryrun in |
| 548 | RETAIN,t) |
| 549 | echo >&2 " --- keep $date" |
| 550 | ;; |
| 551 | EXPIRE,t) |
| 552 | echo >&2 " --- delete $date" |
| 553 | ;; |
| 554 | RETAIN,nil) |
| 555 | echo "keep $date" |
| 556 | ;; |
| 557 | EXPIRE,nil) |
| 558 | echo "delete $date" |
| 559 | $verbose -n " expire $date..." |
| 560 | rm -rf $date $date.* |
| 561 | delete_index $host $fs $date |
| 562 | $verbose " done" |
| 563 | ;; |
| 564 | esac |
| 565 | done |
| 566 | } |
| 567 | |
| 568 | ## Backup hooks. |
| 569 | defhook setup |
| 570 | defhook precommit |
| 571 | defhook postcommit |
| 572 | |
| 573 | backup_precommit_hook () { |
| 574 | host=$1 fs=$2 date=$3 |
| 575 | ## Compatibility: You can override this hook in the configuration file for |
| 576 | ## special effects; but it's better to use `addhook precommit'. |
| 577 | |
| 578 | : |
| 579 | } |
| 580 | addhook precommit backup_precommit_hook |
| 581 | |
| 582 | backup_commit_hook () { |
| 583 | host=$1 fs=$2 date=$3 |
| 584 | ## Compatibility: You can override this hook in the configuration file for |
| 585 | ## special effects; but it's better to use `addhook commit'. |
| 586 | |
| 587 | : |
| 588 | } |
| 589 | addhook commit backup_commit_hook |
| 590 | |
| 591 | do_backup () { |
| 592 | date=$1 fs=$2 fsarg=$3 |
| 593 | ## Back up FS on the current host. |
| 594 | |
| 595 | set -e |
| 596 | attempt=0 |
| 597 | |
| 598 | ## Run a hook beforehand. |
| 599 | set +e; runhook setup $host $fs $date; rc=$?; set -e |
| 600 | case $? in |
| 601 | 0) ;; |
| 602 | 99) log "BACKUP of $host:$fs SKIPPED by hook"; return 0 ;; |
| 603 | *) log "BACKUP of $host:$fs FAILED (hook returns $?)"; return $? ;; |
| 604 | esac |
| 605 | |
| 606 | ## Report the start of this attempt. |
| 607 | log "START BACKUP of $host:$fs" |
| 608 | |
| 609 | ## Maybe we need to retry the backup. |
| 610 | while :; do |
| 611 | |
| 612 | ## Create and mount the remote snapshot. |
| 613 | case $dryrun in |
| 614 | t) |
| 615 | maybe snap_$snap $fs $fsarg |
| 616 | snapmnt="<snapshot>" |
| 617 | ;; |
| 618 | nil) |
| 619 | snapmnt=$(snap_$snap $snapargs $fs $fsarg) || return $? |
| 620 | ;; |
| 621 | esac |
| 622 | $verbose " create snapshot" |
| 623 | |
| 624 | ## Build the list of hardlink sources. |
| 625 | linkdests="" |
| 626 | for i in $host $like; do |
| 627 | d=$STOREDIR/$i/$fs/last/ |
| 628 | if [ -d $d ]; then linkdests="$linkdests --link-dest=$d"; fi |
| 629 | done |
| 630 | |
| 631 | ## Copy files from the remote snapshot. |
| 632 | maybe mkdir -p new/ |
| 633 | case $dryrun in |
| 634 | t) $verbose " running rsync" ;; |
| 635 | nil) $verbose -n " running rsync..." ;; |
| 636 | esac |
| 637 | set +e |
| 638 | run "RSYNC of $host:$fs (snapshot on $snapmnt)" do_rsync \ |
| 639 | $linkdests \ |
| 640 | $rsyncargs \ |
| 641 | $snapmnt/ new/ |
| 642 | rc_rsync=$? |
| 643 | set -e |
| 644 | case $dryrun in nil) $verbose " done" ;; esac |
| 645 | |
| 646 | ## Collect a map of the snapshot for verification purposes. |
| 647 | set +e |
| 648 | case $dryrun in |
| 649 | t) $verbose " remote fshash" ;; |
| 650 | nil) $verbose -n " remote fshash..." ;; |
| 651 | esac |
| 652 | run "@$host: fshash $fs" remote_fshash |
| 653 | rc_fshash=$? |
| 654 | set -e |
| 655 | case $dryrun in nil) $verbose " done" ;; esac |
| 656 | |
| 657 | ## Remove the snapshot. |
| 658 | maybe unsnap_$snap $snapargs $fs $fsarg |
| 659 | $verbose " remove snapshot" |
| 660 | |
| 661 | ## If we failed to copy, then give up. |
| 662 | case $rc_rsync:$rc_fshash in |
| 663 | 0:0) ;; |
| 664 | 0:*) return $rc_fshash ;; |
| 665 | *) return $rc_rsync ;; |
| 666 | esac |
| 667 | |
| 668 | ## Get a matching map of the files received. |
| 669 | maybe mkdir -m750 -p $STOREDIR/tmp/ |
| 670 | localmap=$STOREDIR/tmp/fshash.$host.$fs.$date |
| 671 | case $dryrun in |
| 672 | t) $verbose " local fshash" ;; |
| 673 | nil) $verbose -n " local fshash..." ;; |
| 674 | esac |
| 675 | run "local fshash $host:$fs" local_fshash || return $? |
| 676 | case $dryrun in nil) $verbose " done" ;; esac |
| 677 | |
| 678 | ## Compare the two maps. |
| 679 | set +e |
| 680 | run "compare fshash maps for $host:$fs" diff -u new.fshash $localmap |
| 681 | rc_diff=$? |
| 682 | set -e |
| 683 | case $rc_diff in |
| 684 | 0) |
| 685 | break |
| 686 | ;; |
| 687 | 1) |
| 688 | if [ $attempt -ge $retry ]; then return $rc; fi |
| 689 | $verbose " fshash mismatch; retrying" |
| 690 | attempt=$(( $attempt + 1 )) |
| 691 | ;; |
| 692 | *) |
| 693 | return $rc_diff |
| 694 | ;; |
| 695 | esac |
| 696 | done |
| 697 | |
| 698 | ## Glorious success. |
| 699 | maybe rm -f $localmap |
| 700 | $verbose " fshash match" |
| 701 | |
| 702 | ## Commit this backup. |
| 703 | case $dryrun in |
| 704 | nil) |
| 705 | runhook precommit $host $fs $date |
| 706 | mv new $date |
| 707 | mv new.fshash $date.fshash |
| 708 | insert_index $host $fs $date $VOLUME |
| 709 | runhook commit $host $fs $date |
| 710 | mkdir hack |
| 711 | ln -s $date hack/last |
| 712 | mv hack/last . |
| 713 | rmdir hack |
| 714 | ;; |
| 715 | esac |
| 716 | $verbose " commit" |
| 717 | |
| 718 | ## Expire old backups. |
| 719 | case "${expire_policy+t},${default_policy+t}" in |
| 720 | ,t) expire_policy=$default_policy ;; |
| 721 | esac |
| 722 | case "${expire_policy+t},$dryrun" in |
| 723 | t,nil) run "expiry for $host:$fs" expire_backups ;; |
| 724 | t,t) expire_backups ;; |
| 725 | esac |
| 726 | clear_policy=t |
| 727 | |
| 728 | ## Report success. |
| 729 | case $dryrun in |
| 730 | t) log "END BACKUP of $host:$fs" ;; |
| 731 | nil) log "SUCCESSFUL BACKUP of $host:$fs" ;; |
| 732 | esac |
| 733 | } |
| 734 | |
| 735 | run_backup_cmd () { |
| 736 | fs=$1 date=$2 cmd=$3; shift 3 |
| 737 | ## try_backup FS DATE COMMAND ARGS ... |
| 738 | ## |
| 739 | ## Run COMMAND ARGS to back up filesystem FS on the current host, |
| 740 | ## maintaining a log, and checking whether it worked. The caller has |
| 741 | ## usually worked out the DATE in order to set up the filesystem, and we |
| 742 | ## need it to name the log file properly. |
| 743 | |
| 744 | ## Find a name for the log file. In unusual circumstances, we may have |
| 745 | ## deleted old logs from today, so just checking for an unused sequence |
| 746 | ## number is insufficient. Instead, check all of the logfiles for today, |
| 747 | ## and use a sequence number that's larger than any of them. |
| 748 | case $dryrun in |
| 749 | t) |
| 750 | log=/dev/null |
| 751 | ;; |
| 752 | nil) |
| 753 | seq=1 |
| 754 | for i in "$logdir/$host/$fs.$date#"*; do |
| 755 | tail=${i##*#} |
| 756 | case "$tail" in [!1-9]* | *[!0-9]*) continue ;; esac |
| 757 | if [ -f "$i" -a $tail -ge $seq ]; then seq=$(( tail + 1 )); fi |
| 758 | done |
| 759 | log="$logdir/$host/$fs.$date#$seq" |
| 760 | ;; |
| 761 | esac |
| 762 | |
| 763 | ## Run the backup command. |
| 764 | case $dryrun in nil) mkdir -p $logdir/$host ;; esac |
| 765 | if ! "$cmd" "$@" 9>$log 1>&9; then |
| 766 | echo >&2 |
| 767 | echo >&2 "$quis: backup of $host:$fs FAILED!" |
| 768 | bkprc=1 |
| 769 | fi |
| 770 | |
| 771 | ## Clear away any old logfiles. |
| 772 | remove_old_logfiles "$logdir/$host/$fs" |
| 773 | } |
| 774 | |
| 775 | backup () { |
| 776 | ## backup FS[:ARG] ... |
| 777 | ## |
| 778 | ## Back up the filesystems on the currently selected host using the |
| 779 | ## currently selected snapshot type. |
| 780 | |
| 781 | ## Make sure that there's a store volume. We must do this here rather than |
| 782 | ## in the main body of the script, since the configuration file needs a |
| 783 | ## chance to override STOREDIR. |
| 784 | if ! [ -r $STOREDIR/.rsync-backup-store ]; then |
| 785 | echo >&2 "$quis: no backup volume mounted" |
| 786 | exit 15 |
| 787 | fi |
| 788 | |
| 789 | ## Read the volume name if we don't have one already. Again, this allows |
| 790 | ## the configuration file to provide a volume name. |
| 791 | case "${VOLUME+t}${VOLUME-nil}" in |
| 792 | nil) VOLUME=$(cat $METADIR/volume) ;; |
| 793 | esac |
| 794 | |
| 795 | ## Back up each requested file system in turn. |
| 796 | for fs in "$@"; do |
| 797 | |
| 798 | ## Parse the argument. |
| 799 | case $fs in |
| 800 | *:*) fsarg=${fs#*:} fs=${fs%%:*} ;; |
| 801 | *) fsarg="" ;; |
| 802 | esac |
| 803 | $verbose " filesystem $fs" |
| 804 | |
| 805 | ## Move to the store directory and set up somewhere to put this backup. |
| 806 | cd $STOREDIR |
| 807 | case $dryrun in |
| 808 | nil) |
| 809 | if [ ! -d $host ]; then |
| 810 | mkdir -m755 $host |
| 811 | chown root:root $host |
| 812 | fi |
| 813 | if [ ! -d $host/$fs ]; then |
| 814 | mkdir -m750 $host/$fs |
| 815 | chown root:backup $host/$fs |
| 816 | fi |
| 817 | ;; |
| 818 | esac |
| 819 | cd $host/$fs |
| 820 | |
| 821 | ## Find out if we've already copied this filesystem today. |
| 822 | date=$(date +%Y-%m-%d) |
| 823 | if [ $dryrun = nil ] && [ -d $date ]; then |
| 824 | $verbose " already dumped" |
| 825 | continue |
| 826 | fi |
| 827 | |
| 828 | ## Do the backup of this filesystem. |
| 829 | run_backup_cmd $fs $date do_backup $date $fs $fsarg |
| 830 | done |
| 831 | } |
| 832 | |
| 833 | ###-------------------------------------------------------------------------- |
| 834 | ### Configuration functions. |
| 835 | |
| 836 | defhook start |
| 837 | defhook end |
| 838 | |
| 839 | done_first_host_p=nil |
| 840 | |
| 841 | host () { |
| 842 | host=$1 |
| 843 | like= userat= |
| 844 | case $done_first_host_p in |
| 845 | nil) runhook start; done_first_host_p=t ;; |
| 846 | esac |
| 847 | case "${expire_policy+t},${default_policy+t}" in |
| 848 | t,) default_policy=$expire_policy ;; |
| 849 | esac |
| 850 | unset expire_policy |
| 851 | $verbose "host $host" |
| 852 | } |
| 853 | |
| 854 | snaptype () { snap=$1; shift; snapargs="$*"; retry=0; } |
| 855 | rsyncargs () { rsyncargs="$*"; } |
| 856 | like () { like="$*"; } |
| 857 | retry () { retry="$*"; } |
| 858 | user () { userat="$*@"; } |
| 859 | |
| 860 | retain () { |
| 861 | case $clear_policy in t) unset expire_policy; clear_policy=nil ;; esac |
| 862 | expire_policy="${expire_policy+$expire_policy |
| 863 | }$*" |
| 864 | } |
| 865 | |
| 866 | ###-------------------------------------------------------------------------- |
| 867 | ### Read the configuration and we're done. |
| 868 | |
| 869 | usage () { |
| 870 | echo "usage: $quis [-nv] [-c CONF]" |
| 871 | } |
| 872 | |
| 873 | version () { |
| 874 | echo "$quis version $VERSION" |
| 875 | } |
| 876 | |
| 877 | whine () { echo >&8 "$@"; } |
| 878 | |
| 879 | while getopts "hVvc:n" opt; do |
| 880 | case "$opt" in |
| 881 | h) usage; exit 0 ;; |
| 882 | V) version; config; exit 0 ;; |
| 883 | v) verbose=whine ;; |
| 884 | c) conf=$OPTARG ;; |
| 885 | n) dryrun=t ;; |
| 886 | *) exit 1 ;; |
| 887 | esac |
| 888 | done |
| 889 | shift $((OPTIND - 1)) |
| 890 | case $# in 0) ;; *) usage >&2; exit 1 ;; esac |
| 891 | exec 8>&1 |
| 892 | |
| 893 | . "$conf" |
| 894 | |
| 895 | runhook end $bkprc |
| 896 | case "$bkprc" in |
| 897 | 0) $verbose "All backups successful" ;; |
| 898 | *) $verbose "Backups FAILED" ;; |
| 899 | esac |
| 900 | |
| 901 | ###----- That's all, folks -------------------------------------------------- |
| 902 | |
| 903 | exit $bkprc |