2 # Copyright (C) 2016 Ian Kelling
4 # Licensed under the Apache License, Version 2.0 (the "License");
5 # you may not use this file except in compliance with the License.
6 # You may obtain a copy of the License at
8 # http://www.apache.org/licenses/LICENSE-2.0
10 # Unless required by applicable law or agreed to in writing, software
11 # distributed under the License is distributed on an "AS IS" BASIS,
12 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
13 # See the License for the specific language governing permissions and
14 # limitations under the License.
17 # todo: if we cancel in the middle of a btrfs send, then run again
18 # immediately, the received subvolume doesn't get a Received UUID:
19 # field, and we won't mount it. Need to figure out a solution that will
23 [[ $EUID == 0 ]] ||
exec sudo
-E "${BASH_SOURCE[0]}" "$@"
25 source /usr
/local
/lib
/err
29 btrbk-run [OPTIONS] [run|resume|archive]
30 usually -t TARGET_HOST or -s SOURCE_HOST
32 Note, at source location, intentionally not executable, run and read
36 echo "top of script file:"
37 sed -n '1,/^[# ]*end command line/{p;b};q' "$0"
43 script_name
="${BASH_SOURCE[0]}"
44 script_name
="${script_name##*/}"
45 pre
="${SSH_CLIENT:+$HOSTNAME} $script_name:"
46 m
() { if $verbose; then printf "$pre%s\n" "$*"; fi; "$@"; }
47 e
() { printf "$pre%s\n" "$*"; }
48 die
() { printf "$pre%s\n" "$*" >&2; echo "exiting with status 1" >&2; exit 1; }
49 mexit
() { echo "$pre: exiting with status $1"; exit $1; }
52 if [[ -e /b
/bash_unpublished
/source-state
]]; then
53 source /b
/bash_unpublished
/source-state
56 # note q is owned by root:1000
65 dry_run
=false
# mostly for testing
67 verbose
=true
; verbose_arg
=-v
68 if [[ $INVOCATION_ID ]]; then
69 # INVOCATION_ID means running as a systemd service. we cant show progress in this case,
70 # but if we pass the arg, it will insert mbuffer into the command.
73 progress_arg
="--progress"
75 incremental_strict
=false
78 default_args_file
=/etc
/btrbk-run.conf
79 if [[ -s $default_args_file ]]; then
80 # shellcheck disable=SC2046 # we want word splitting
81 set -- $
(< $default_args_file) "$@"
82 # i havent used this feature yet, so warn about it
83 echo "$0: warning: default btrbk-run options set in $default_args_file (sleeping 5 seconds):"
84 cat $default_args_file
94 temp
=$
(getopt
-l cron
,fast
,pull-reexec
,help 23ceikl
:m
:npqrs
:t
:vh
"$@") || usage
1
98 # some behaviors specific to running under cron:
99 # - skip hosts where xprintidle haven't been idle recently
100 # - if we can't ssh to 1 or more hosts, still do the rest
101 # - if we aren't MAIL_HOST and no -m or -s, just exit
106 # for the rare case we want to run multiple instances at the same time
109 # only creates the config file, does not run btrbk
110 -c) conf_only
=true
;;
111 # quit early, just btrbk, no extra remounting etc.
113 # skip various checks. when we run twice in a row for
114 # switch mail-host, no need to repeat the same checks again.
116 -i) incremental_strict
=true
;;
117 # note this implies resume
118 -k) kd_spread_maybe
=true
;;
119 # bytes per second, suffix k m g
120 -l) rate_limit
=$2; shift ;;
121 # Comma separated mountpoints to backup. This has defaults set below.
122 -m) IFS
=, mountpoints
=($2); unset IFS
; shift ;;
126 # internal option for rerunning under newer SOURCE_HOST version.
127 --pull-reexec) pull_reexec
=true
;;
129 -q) verbose
=false
; verbose_arg
=; progress_arg
= ;;
130 # source host to receive a backup from
134 if [[ $source == *:* ]]; then
135 bbksource
="[$source]"
139 # target hosts to send to. empty is valid for just doing local
140 # snapshot. we have default hosts we will populate.
141 -t) IFS
=, targets
=($2); unset IFS
; shift ;;
143 -v) verbose
=true
; verbose_arg
=-v ;;
146 *) die
"Internal error!" ;;
153 if $kd_spread_maybe; then
154 if [[ $cmd_arg && $cmd_arg != resume
]]; then
155 die
"dont pass -k without resume or empty run arg"
160 if [[ ! $cmd_arg ]]; then
165 std_preserve
="36h 14d 8w 24m"
166 q_preserve
="18h 14d 8w"
171 # This works better than the normal archive command. We have to
172 # specify the mount points, but that is what we are used to doing and
173 # we prefer it. Another difference is that archive works recursively
174 # and we don't care about that. Sometimes we may still want to run
175 # btrbk archive, but it doesn't even use the config file, so just
176 # run it directly, eg:
177 # time s btrbk -v archive /mnt/r7/amy/boot/btrbk ssh://bo/mnt/boot2/btrbk
180 std_preserve
="999h 999d 999w 999m"
181 q_preserve
="$std_preserve"
184 *) die
"untested command arg" ;;
187 if (( $# > 1 )); then
188 die
: "only 1 nonoption arg is supported"
191 if [[ -v targets
&& $source ]]; then
192 # note, this doesnt need to be the case, but
193 # we would need to think about it.
194 die
"error: -t and -s are mutually exclusive"
197 ### end options parsing
199 # remove path from earlier version of btrbk
200 rm -f /usr
/sbin
/btrbk
201 # note, this still works as intended if there is no /usr/bin/btrbk
202 if [[ /a
/opt
/btrbk
/btrbk
-nt /usr
/bin
/btrbk
]]; then
203 if [[ -e /b
/distro-functions
/src
/package-manager-abstractions
]]; then
204 .
/b
/distro-functions
/src
/package-manager-abstractions
211 # TODO: i wonder if there should be an option to send to the default
212 # targets, plus any given on the command line.
217 # set default targets
218 if [[ ! -v targets
&& ! $source ]]; then
220 if [[ $HOSTNAME != "$MAIL_HOST" ]]; then
221 if $kd_spread_maybe && [[ $HOSTNAME == kd
&& $MAIL_HOST == x3
]]; then
222 if ping -q -c1 -w1 x3.office.fsf.org
&>/dev
/null
; then
223 work_host
=x3.office.fsf.org
224 elif ping -q -c1 -w1 x3wg.b8.nz
&>/dev
/null
; then
227 if [[ $work_host ]]; then
228 source_state
="$(ssh $work_host cat /a/bin/bash_unpublished/source-state)"
230 if [[ $MAIL_HOST == x3
]]; then
233 # x3 was the mail host, but it moved to some other machine
234 # without updating us yet.
235 echo "MAIL_HOST=$MAIL_HOST, nothing to do"
239 echo "MAIL_HOST=$MAIL_HOST, nothing to do"
243 echo "MAIL_HOST=$MAIL_HOST, nothing to do"
253 kw|kd|frodo|x2|x3|sy
) : ;;
255 die
"error: no default targets for this host, use -t"
267 if [[ $
(dig +short @
10.2.0.1 -x 10.2.0.2 2>&1 ||
:) == kd.b8.nz.
]] \
268 && ip n show
10.2.0.1 |
grep .
&>/dev
/null
; then
270 elif ping -q -c1 -w1 hal.office.fsf.org
&>/dev
/null \
271 && ip n show
192.168.0.26 |
grep .
&>/dev
/null
; then
277 if ! $kd_spread && [[ $HOSTNAME != x3
]]; then
279 if ping -q -c1 -w1 x3.office.fsf.org
&>/dev
/null
; then
280 targets
+=(x3.office.fsf.org
)
281 elif ping -q -c1 -w1 $h.b8.nz
&>/dev
/null
; then
282 # in case we took it home
285 targets
+=(x3wg.b8.nz
)
288 # temporarily disabled while doing recovery
289 # for h in frodo kd; do
291 if [[ $HOSTNAME == "$h" ]]; then
297 if [[ $HOSTNAME == "$h" ]]; then
300 if ping -q -c1 -w1 $h.b8.nz
&>/dev
/null
; then
302 elif ping -q -c1 -w1 ${h}w.b8.nz
&>/dev
/null
; then
303 targets
+=(${h}w.b8.nz
)
308 for h
in x2 x3 kw
; do
309 if [[ $HOSTNAME == "$h" ]]; then
312 if ping -q -c1 -w1 $h.office.fsf.org
&>/dev
/null
; then
313 targets
+=($h.office.fsf.org
)
323 if [[ ${mountpoints[0]} ]]; then
324 for mp
in ${mountpoints[@]}; do
325 if [[ -e /nocow
/btrfs-stale
/$mp ]]; then
326 die
"error: $mp is stale, mount-latest-subvol first"
330 # set default mountpoints
331 if [[ ${targets[0]} == tp
]]; then
337 if [[ $source ]]; then
338 source_state
="$(ssh $source cat /a/bin/bash_unpublished/source-state)"
340 source_host
="$(ssh $source cat /etc/hostname)"
341 if [[ $source_host == "$MAIL_HOST" ]]; then
342 prospective_mps
+=(/o
)
344 if [[ $source_host == "$HOST2" ]]; then
345 prospective_mps
+=(/a
/ar /qr
/q
)
348 if [[ $HOSTNAME == "$MAIL_HOST" ]]; then
349 prospective_mps
+=(/o
)
351 if [[ $HOSTNAME == "$HOST2" ]]; then
352 prospective_mps
+=(/a
/ar /qr
/q
)
355 prospective_mps
=(/a
/ar /o
/qr
/q
)
358 # note: put q last just in case its specific retention options were to
359 # affect other config sections. I havent tested if that is the case.
363 for mp
in ${prospective_mps[@]}; do # default mountpoints to sync
364 if [[ -e /nocow
/btrfs-stale
/$mp ]]; then
365 e
"warning: $mp stale, not adding to default mountpoints"
368 if awk '{print $2}' /etc
/fstab |
grep -xF $mp &>/dev
/null
; then
374 if (( ! ${#mountpoints[@]} )); then
375 die didnt get mountpoint arg and had no defaults
378 ##### end command line parsing ########
380 #### begin pre-checks #####
382 # todo: this has a timing problem, since btrbk.timer could activate the service after this check.
383 if ! $fast && [[ $source ]]; then
384 if [[ $
(ssh $source ps
--no-headers -o comm 1) == systemd
]]; then
385 status
=$
(ssh $source systemctl is-active btrbk.service
) ||
: # normally returns 3
387 inactive|failed
) : ;;
389 echo "$0: error: cron btrbk is running on source. exiting out of caution"
395 if ! command -v btrbk
&>/dev
/null
; then
396 die
"error: no btrbk binary found"
400 #### end pre-checks #####
402 mkdir
-p /var
/log
/btrbk
403 # The journal doesnt go back to my oldest backups, and I've found myself
404 # wanting older logs. Not going to bother expiring old logs, since it is
405 # fine if they go back years.
406 log_path
=/var
/log
/btrbk
/$
(date +%F_
%T
%:::z
).log
407 echo copying output to
$log_path
408 exec &> >(tee -a $log_path)
412 printf "$pre options: conf_only=%s\ndry_run=%s\nrate_limit=%s\nverbose=%s\ncmd_arg=%s" "$conf_only" "$dry_run" "$rate_limit" "$verbose" "$cmd_arg"
415 if [[ -v targets
]]; then
416 echo "targets: ${targets[*]}"
419 if [[ $source ]]; then
420 echo "source: $source"
423 echo "mountpoints: ${mountpoints[*]}"
427 # pull_reexec stops us from getting into an infinite loop if there is some
428 # kind of weird problem
430 for m
in "${mountpoints[@]}"; do
431 if [[ $m == /a
]]; then
436 if ! $pull_reexec && [[ $source ]] && $pulla ; then
438 m rsync
-ra $source:/usr
/local
/bin
/{mount-latest-subvol
,check-subvol-stale
} /usr
/local
/bin
439 m rsync
-ra $source:/usr
/local
/lib
/err
/usr
/local
/lib
440 m scp
$source:/a
/bin
/distro-setup
/btrbk-run
$tmpf
441 if ! diff -q $tmpf ${BASH_SOURCE[0]}; then
442 e
"found different version on host $source. reexecing"
443 install -T $tmpf /usr
/local
/bin
/btrbk-run
444 m
/usr
/local
/bin
/btrbk-run
--pull-reexec "${orig_args[@]}"
451 # if our mountpoints are from stale snapshots,
452 # it doesn't make sense to do a backup.
453 m check-subvol-stale
${mountpoints[@]} || die
"found stale mountpoints in ${mountpoints[*]}"
455 # for an initial run, btrbk requires the dir to exist.
456 mkdir
-p /mnt
/{root
,o
}/btrbk
458 local_zone
=$
(date +%z
)
460 if [[ $source ]]; then
464 if ! zone
=$
(ssh root@
$source date +%z
); then
466 echo "$0: warning: failed to ssh to root@$source"
468 die failed to
ssh to root@
$source
471 if [[ $zone != "$local_zone" ]]; then
472 die
"error: dont confuse yourself with multiple time zones. $h has different timezone than localhost"
479 for h
in ${targets[@]}; do
480 if $fast ||
$conf_only; then
481 # Use some typical values in this case
482 root_size
=$
(( 1024 * 1024 * 2000 )) #2tb
485 elif remote_str
=$
(timeout
-s 9 6 ssh root@
$h "mkdir -p /mnt/root/btrbk /mnt/o/btrbk && date +%z && df --output=size,pcent / | tail -n1"); then
486 mapfile
-t tmp_array
<<<"$remote_str"
487 zone
="${tmp_array[0]}"
488 IFS
=" " read -r root_size percent_used
<<<"${tmp_array[1]}"
489 percent_used
=${percent_used%%%}
491 if (( ${#tmp_array[@]} != 2 )); then
492 die
"error: didnt get 2 lines in test ssh to target $h. investigate"
494 case $percent_used in
495 [0-9]|
[1-9][0-9]) : ;;
497 die
"error: didnt get percent disk use in test ssh to target $h. investigate"
505 # we may be booted into a bootstrap fs or something
506 min_root_kb
=$
(( 1024 * 1024 * 200 )) # 200 gb
507 if (( root_size
< min_root_kb
)); then
511 if (( percent_used
>= 98 )); then
512 die
"error: filesystem on target $h is $percent_used % full"
515 # on sy, xprintidle is resetting every 12 seconds even when not
516 # idle, i dunno why, instead we are checking if the screen is locked,
517 # which is good enough.
519 # This is a separate ssh because the command can fail and thatis ok.
522 if lock_info
=$
(timeout
-s 9 6 ssh $h DISPLAY
=:0 xscreensaver-command
-time); then
523 if [[ $lock_info != *non-blanked
* ]]; then
530 # Ignore this host. i sometimes use a non-main machine for
531 # testing or web browsing, knowing that everything will be wiped
532 # by the next backup, but I dont want it to happen as Im using
534 e
"warning: $h: seems to be actively in use, skipping for now"
539 if [[ $zone != "$local_zone" ]]; then
540 die
"error: dont confuse yourself with multiple time zones. $h has different timezone than localhost"
543 if [[ ! ${sshable[*]} ]] || { ! $cron && [[ ${sshfail[*]} ]]; }; then
544 die
"failed to ssh to hosts: ${sshfail[*]}"
546 if [[ ${sshfail[*]} ]]; then
548 e
"error: failed to ssh to ${sshfail[*]} but continuing with other hosts"
550 targets
=(${sshable[@]})
555 cat >/etc
/btrbk
$conf_suf.conf
<<EOF
556 ssh_identity /q/root/h
557 #ssh_identity /root/.ssh/home
559 # Just a guess that local7 is a good facility to pick.
560 # It's a bit odd that the transaction log has to be logged to
561 # a file or syslog, while other output is sent to std out.
562 # The man does not mention a way for them to be together, but
563 # I dunno if setting a log level like warn might also output
565 transaction_syslog local7
568 #stream_compress zstd
570 # so we only run one at a time
571 lockfile /var/lock/btrbk$conf_suf.lock
573 # default format of short does not accomidate hourly preservation setting
574 timestamp_format long-iso
576 # only make a snapshot if things have changed
577 snapshot_create onchange
578 # I could make this different from target_preserve,
579 # if one disk had less space.
580 # for now, keeping them equal.
581 snapshot_preserve $std_preserve
582 snapshot_preserve_min 6h
584 # so, total backups = ~58
585 target_preserve $std_preserve
586 target_preserve_min 6h
588 # i tried this when investigating: clone no source subvolume found error
589 #incremental_prefs sro:1 srn:1 sao san:1 aro:1 arn:1
591 # if something fails and it's not obvious, try doing
592 # btrbk -l debug -v dryrun
594 rate_limit $rate_limit
597 if $incremental_strict; then
598 cat >>/etc
/btrbk
$conf_suf.conf
<<EOF
606 # q has sensitive data i dont want to backup for so long
607 cat >>/etc
/btrbk
$conf_suf.conf
<<EOF
608 snapshot_preserve $q_preserve
609 snapshot_preserve_min 2h
611 target_preserve $q_preserve
612 target_preserve_min 2h
620 mp_count
=${#mountpoints[@]}
621 for (( i
=0; i
< mp_count
- 1 ; i
++ )); do
622 if [[ ${mountpoints[i]} == /q
]]; then
623 unset "mountpoints[i]"
628 for m
in ${mountpoints[@]}; do
639 if [[ $source ]]; then
640 cat >>/etc
/btrbk
$conf_suf.conf
<<EOF
641 volume ssh://$bbksource$vol
645 cat >>/etc
/btrbk
$conf_suf.conf
<<EOF
646 target send-receive $vol/btrbk
649 if (( ${#targets[@]} )); then
650 cat >>/etc
/btrbk
$conf_suf.conf
<<EOF
655 for tg
in ${targets[@]}; do
657 if [[ $tg == *:* ]]; then
660 cat >>/etc
/btrbk
$conf_suf.conf
<<EOF
661 target send-receive ssh://$tg$vol/btrbk
667 # todo: umount first to ensure we don't have any errors
668 # todo: do some kill fuser stuff to make umount more reliable
678 m btrbk
-c /etc
/btrbk
$conf_suf.conf
-v -n $cmd_arg
681 # -q and just using the syslog option seemed nice,
682 # but it doesn't show when a send has a parent and when it doesn't.
683 m btrbk
-c /etc
/btrbk
$conf_suf.conf
$preserve_arg $verbose_arg $progress_arg $cmd_arg
689 # todo: tp not valid anymore.
690 # if we have it, sync to systems which don't
691 if mountpoint
$rsync_mountpoint >/dev
/null
; then
692 for tg
in ${targets[@]}; do
695 dirs=(/p
/c
/machine_specific
/tp
)
696 for x
in /p
/c
/machine_specific
/*.hosts
; do
697 if grep -qxF $tg $x; then
701 m rsync
-aSAXPH --specials --devices --delete --relative ${dirs[@]} root@
$tg:/
708 for mp
in "${mountpoints[@]}"; do
709 subvols
+=("${mp##*/}")
711 if [[ $source ]]; then
712 m mount-latest-subvol
"${subvols[@]}"
714 for tg
in ${targets[@]}; do
715 m
/a
/exe
/mount-latest-remote
"$tg" "${subvols[@]}" || ret
=$?
719 if [[ $ret == 0 ]]; then
720 for tg
in ${targets[@]}; do
721 ssh root@
$tg /usr
/local
/bin
/mail-backup-clean
727 # todo: move variable data we don't care about backing up
728 # to /nocow and symlink it.
731 # background on btrbk timezones. with short/long, timestamps use local time.
732 # for long, if your local time moves backwards, by moving timezones or
733 # for an hour when daylight savings changes it, you will temporarily get
734 # a more aggressive retention policy for the overlapping period, and
735 # vice versa for the opposite timezone move. The alternative is using
736 # long-iso, which puts timezone info into the timestamp, which means
737 # that instead of shifting time, you shift the start of day/week/month
738 # which is used for retention to your new local time, which means for
739 # example, if you moved forward by 8 hours, the daily/weekly/monthly
740 # retention will be 8 hours more aggressive since midnight is at a new
741 # time, unless you fake the timzeone using the TZ env variable.
742 # However, in the short term, there will be no inconsistencies.
743 # I don't see any problem with shifting when the day starts for
744 # retention, so I'm using long-iso.
746 # note to create a long-iso timestamp: date +%Y%m%dT%H%M%S%z