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"
44 m
() { if $verbose; then printf "$pre%s\n" "$*"; fi; "$@"; }
45 e
() { printf "$pre%s\n" "$*"; }
46 die
() { printf "$pre%s\n" "$*" >&2; echo "exiting with status 1" >&2; exit 1; }
47 mexit
() { echo "$pre: exiting with status $1"; exit $1; }
50 if [[ -e /b
/bash_unpublished
/source-state
]]; then
51 source /b
/bash_unpublished
/source-state
54 # note q is owned by root:1000
63 dry_run
=false
# mostly for testing
65 verbose
=true
; verbose_arg
=-v
66 if [[ $INVOCATION_ID ]]; then
67 # INVOCATION_ID means running as a systemd service. we cant show progress in this case,
68 # but if we pass the arg, it will insert mbuffer into the command.
71 progress_arg
="--progress"
73 incremental_strict
=false
76 default_args_file
=/etc
/btrbk-run.conf
77 if [[ -s $default_args_file ]]; then
78 set -- $
(< $default_args_file) "$@"
79 # i havent used this feature yet, so warn about it
80 echo "$0: warning: default btrbk-run options set in $default_args_file (sleeping 5 seconds):"
81 cat $default_args_file
88 temp
=$
(getopt
-l cron
,pull-reexec
,help ceil
:m
:npqrs
:t
:vh
"$@") || usage
1
92 # some behaviors specific to running under cron:
93 # - skip hosts where xprintidle haven't been idle recently
94 # - if we can't ssh to 1 or more hosts, still do the rest
95 # - if we aren't MAIL_HOST and no -m or -s, just exit
101 # only creates the config file, does not run btrbk
102 -c) conf_only
=true
; shift ;;
103 # quit early, just btrbk, no extra remounting etc.
104 -e) early
=true
; shift ;;
105 -i) incremental_strict
=true
; shift ;;
106 # bytes per second, suffix k m g
107 -l) rate_limit
=$2; shift 2 ;;
108 # Comma separated mountpoints to backup. This has defaults set below.
109 -m) IFS
=, mountpoints
=($2); unset IFS
; shift 2 ;;
110 -n) dry_run
=true
; dry_run_arg
=-n; shift ;;
112 -p) progress_arg
=; shift ;;
113 # internal option for rerunning under newer SOURCE_HOST version.
114 --pull-reexec) pull_reexec
=true
; shift ;;
116 -q) verbose
=false
; verbose_arg
=; progress_arg
=; shift ;;
117 # source host to receive a backup from
121 if [[ $source == *:* ]]; then
122 bbksource
="[$source]"
126 # target hosts to send to. empty is valid for just doing local
127 # snapshot. we have default hosts we will populate.
128 -t) IFS
=, targets
=($2); unset IFS
; shift 2 ;;
130 -v) verbose
=true
; verbose_arg
=-v; shift ;;
133 *) die
"Internal error!" ;;
140 std_preserve
="36h 14d 8w 24m"
146 # This works better than the normal archive command. We have to
147 # specify the mount points, but that is what we are used to doing and
148 # we prefer it. Another difference is that archive works recursively
149 # and we don't care about that. Sometimes we may still want to run
150 # btrbk archive, but it doesn't even use the config file, so just
151 # run it directly, eg:
152 # time s btrbk -v archive /mnt/r7/amy/boot/btrbk ssh://bo/mnt/boot2/btrbk
155 std_preserve
="999h 999d 999w 999m"
156 q_preserve
="$std_preserve"
159 *) die
"untested command arg" ;;
162 if (( $# > 1 )); then
163 die
: "only 1 nonoption arg is supported"
166 if [[ -v targets
&& $source ]]; then
167 # note, this doesnt need to be the case, but
168 # we would need to think about it.
169 die
"error: -t and -s are mutually exclusive"
173 printf "options: conf_only=%s\ndry_run=%s\nrate_limit=%s\nverbose=%s\ncmd_arg=%s" "$conf_only" "$dry_run" "$rate_limit" "$verbose" "$cmd_arg"
175 ### end options parsing
177 # remove path from earlier version of btrbk
178 rm -f /usr
/sbin
/btrbk
179 # note, this still works as intended if there is no /usr/bin/btrbk
180 if [[ /a
/opt
/btrbk
/btrbk
-nt /usr
/bin
/btrbk
]]; then
181 if [[ -e /b
/distro-functions
/src
/package-manager-abstractions
]]; then
182 .
/b
/distro-functions
/src
/package-manager-abstractions
189 # TODO: i wonder if there should be an option to send to the default
190 # targets, plus any given on the command line.
195 # set default targets
196 if [[ ! -v targets
&& ! $source ]]; then
198 if [[ $HOSTNAME != "$MAIL_HOST" ]]; then
199 if [[ $HOSTNAME == kd
&& $MAIL_HOST = x2
]]; then
202 echo "MAIL_HOST=$MAIL_HOST, nothing to do"
213 # todo, fix this up once frodo is back
214 # targets=(frodo.b8.nz)
220 if ping -q -c1 -w1 hal.office.fsf.org \
221 && ip n show
192.168.0.26 |
grep .
&>/dev
/null
; then
227 if ping -q -c1 -w1 iank.vpn.office.fsf.org
&>/dev
/null
; then
228 home
=iank.vpn.office.fsf.org
233 if ping -q -c1 -w1 b8.nz
&>/dev
/null
; then
244 targets
+=($home x2.office.fsf.org
)
249 targets
+=(x2.office.fsf.org x2.b8.nz
)
251 targets
+=(x2wg.b8.nz
)
255 if ! $kd_spread; then
256 if ping -q -c1 -w1 x2.office.fsf.org
&>/dev
/null
; then
257 targets
+=(x2.office.fsf.org
)
259 targets
+=(x2wg.b8.nz
)
262 if ping -q -c1 -w1 sy.b8.nz
&>/dev
/null
; then
267 if ping -q -c1 -w1 x3.b8.nz
&>/dev
/null
; then
278 die
"error: no default targets for this host, use -t"
283 if [[ -v targets
]]; then
284 echo "targets: ${targets[*]}"
287 if [[ $source ]]; then
288 echo "source: $source"
291 if [[ ${mountpoints[0]} ]]; then
292 for mp
in ${mountpoints[@]}; do
293 if [[ -e /nocow
/btrfs-stale
/$mp ]]; then
294 die
"error: $mp is stale, mount-latest-subvol first"
298 # set default mountpoints
299 if [[ ${targets[0]} == tp
]]; then
308 if [[ $source ]]; then
309 source_state
="$(ssh $source cat /a/bin/bash_unpublished/source-state)"
311 source_host
="$(ssh $source cat /etc/hostname)"
312 if [[ $source_host == "$MAIL_HOST" ]]; then
313 prospective_mps
+=(/o
)
315 if [[ $source_host == "$HOST2" ]]; then
316 prospective_mps
+=(/a
/ar /qr
/q
)
319 if [[ $HOSTNAME == "$MAIL_HOST" ]]; then
320 prospective_mps
+=(/o
)
322 if [[ $HOSTNAME == "$HOST2" ]]; then
323 prospective_mps
+=(/a
/ar /qr
/q
)
326 # note: put q last just in case its specific retention options were to
327 # affect other config sections. I havent tested if that is the case.
331 for mp
in ${prospective_mps[@]}; do # default mountpoints to sync
332 if [[ -e /nocow
/btrfs-stale
/$mp ]]; then
333 e
"warning: $mp stale, not adding to default mountpoints"
336 if awk '{print $2}' /etc
/fstab |
grep -xF $mp &>/dev
/null
; then
342 echo "mountpoints: ${mountpoints[*]}"
344 ##### end command line parsing ########
346 if [[ $source ]]; then
347 if [[ $
(ssh $source ps
--no-headers -o comm 1) == systemd
]]; then
348 status
=$
(ssh $source systemctl is-active btrbk.service
) ||
: # normally returns 3
350 inactive|failed
) : ;;
352 echo "$0: error: cron btrbk is running on source. exiting out of caution"
358 # pull_reexec stops us from getting into an infinite loop if there is some
359 # kind of weird problem
361 for m
in "${mountpoints[@]}"; do
362 if [[ $m == /a
]]; then
367 if ! $pull_reexec && [[ $source ]] && $pulla ; then
369 m rsync
-ra $source:/usr
/local
/bin
/{mount-latest-subvol
,check-subvol-stale
} /usr
/local
/bin
370 m rsync
-ra $source:/usr
/local
/lib
/err
/usr
/local
/lib
371 m scp
$source:/a
/bin
/distro-setup
/btrbk-run
$tmpf
372 if ! diff -q $tmpf ${BASH_SOURCE[0]}; then
373 e
"found different version on host $source. reexecing"
374 install -T $tmpf /usr
/local
/bin
/btrbk-run
375 m
/usr
/local
/bin
/btrbk-run
--pull-reexec "${orig_args[@]}"
381 if ! command -v btrbk
&>/dev
/null
; then
382 die
"error: no btrbk binary found"
384 # if our mountpoints are from stale snapshots,
385 # it doesn't make sense to do a backup.
386 check-subvol-stale
${mountpoints[@]} || die
"found stale mountpoints in ${mountpoints[*]}"
388 # for an initial run, btrbk requires the dir to exist.
389 mkdir
-p /mnt
/{root
,o
}/btrbk
390 local_zone
=$
(date +%z
)
392 if [[ $source ]]; then
393 if ! zone
=$
(ssh root@
$source date +%z
); then
395 echo "$0: warning: failed to ssh to root@$source"
397 die failed to
ssh to root@
$source
400 if [[ $zone != "$local_zone" ]]; then
401 die
"error: dont confuse yourself with multiple time zones. $h has different timezone than localhost"
408 min_idle_ms
=$
((1000 * 60 * 15))
409 for h
in ${targets[@]}; do
410 if remote_info
=( $
(timeout
-s 9 6 ssh root@
$h "mkdir -p /mnt/root/btrbk /mnt/o/btrbk && date +%z && df --output=size,pcent / | tail -n1") ); then
411 zone
=${remote_info[0]}
412 root_size
=${remote_info[1]}
413 percent_used
=${remote_info[2]%%%}
415 if (( ${#remote_info[@]} != 3 )); then
416 die
"error: didnt get 3 fields in test ssh to target $h. investigate"
418 elif $conf_only; then
419 # Use some typical values in this case
420 root_size
=$
(( 1024 * 1024 * 2000 )) #2tb
428 # we may be booted into a bootstrap fs or something
429 min_root_kb
=$
(( 1024 * 1024 * 200 )) # 200 gb
430 if (( root_size
< min_root_kb
)); then
434 if (( percent_used
>= 98 )); then
435 die
"error: filesystem on target $h is $percent_used % full"
438 # This is a separate ssh because xprintidle can fail and thats ok.
439 if $cron && idle_ms
=$
(timeout
-s 9 6 ssh $h DISPLAY
=:0 xprintidle
); then
440 if (( idle_ms
< min_idle_ms
)); then
442 # Ignore this host. i sometimes use a non-main machine for
443 # testing or web browsing, knowing that everything will be wiped
444 # by the next backup, but I dont want it to happen as Im using
446 e
"warning: $h: active X session in the last 15 minutes, skipping for now"
451 if [[ $zone != "$local_zone" ]]; then
452 die
"error: dont confuse yourself with multiple time zones. $h has different timezone than localhost"
455 if [[ ! ${sshable[*]} ]] || { ! $cron && [[ ${sshfail[*]} ]]; }; then
456 die
"failed to ssh to hosts: ${sshfail[*]}"
458 if [[ ${sshfail[*]} ]]; then
460 e
"error: failed to ssh to ${sshfail[*]} but continuing with other hosts"
462 targets
=(${sshable[@]})
467 cat >/etc
/btrbk.conf
<<EOF
468 ssh_identity /q/root/h
469 #ssh_identity /root/.ssh/home
471 # Just a guess that local7 is a good facility to pick.
472 # It's a bit odd that the transaction log has to be logged to
473 # a file or syslog, while other output is sent to std out.
474 # The man does not mention a way for them to be together, but
475 # I dunno if setting a log level like warn might also output
477 transaction_syslog local7
480 #stream_compress zstd
482 # so we only run one at a time
483 lockfile /var/lock/btrbk.lock
485 # default format of short does not accomidate hourly preservation setting
486 timestamp_format long-iso
488 # only make a snapshot if things have changed
489 snapshot_create onchange
490 # I could make this different from target_preserve,
491 # if one disk had less space.
492 # for now, keeping them equal.
493 snapshot_preserve $std_preserve
494 snapshot_preserve_min 2h
496 # so, total backups = ~58
497 target_preserve $std_preserve
498 target_preserve_min 2h
500 # if something fails and it's not obvious, try doing
501 # btrbk -l debug -v dryrun
503 rate_limit $rate_limit
506 if $incremental_strict; then
507 cat >>/etc
/btrbk.conf
<<EOF
512 for m
in ${mountpoints[@]}; do
523 if [[ $source ]]; then
524 cat >>/etc
/btrbk.conf
<<EOF
525 volume ssh://$bbksource$vol
527 target send-receive $vol/btrbk
530 if (( ${#targets[@]} )); then
531 cat >>/etc
/btrbk.conf
<<EOF
537 # q has sensitive data i dont want to backup for so long
538 cat >>/etc
/btrbk.conf
<<EOF
539 snapshot_preserve $std_preserve
540 snapshot_preserve_min 2h
542 target_preserve $std_preserve
543 target_preserve_min 2h
547 for tg
in ${targets[@]}; do
549 if [[ $tg == *:* ]]; then
552 cat >>/etc
/btrbk.conf
<<EOF
553 target send-receive ssh://$tg$vol/btrbk
559 # todo: umount first to ensure we don't have any errors
560 # todo: do some kill fuser stuff to make umount more reliable
570 m btrbk
-v -n $cmd_arg
573 # -q and just using the syslog option seemed nice,
574 # but it doesn't show when a send has a parent and when it doesn't.
575 m btrbk
$preserve_arg $verbose_arg $progress_arg $cmd_arg
577 # todo: tp not valid anymore.
578 # if we have it, sync to systems which don't
579 if mountpoint
$rsync_mountpoint >/dev
/null
; then
580 for tg
in ${targets[@]}; do
583 dirs=(/p
/c
/machine_specific
/tp
)
584 for x
in /p
/c
/machine_specific
/*.hosts
; do
585 if grep -qxF $tg $x; then
589 m rsync
-aSAXPH --specials --devices --delete --relative ${dirs[@]} root@
$tg:/
595 if [[ $source ]]; then
596 m mount-latest-subvol
598 m
/a
/exe
/mount-latest-remote
${targets[@]}
601 if [[ $ret == 0 ]]; then
602 for tg
in ${targets[@]}; do
604 #ssh root@$tg /a/exe/mail-backup-clean
610 # todo: move variable data we don't care about backing up
611 # to /nocow and symlink it.
614 # background on btrbk timezones. with short/long, timestamps use local time.
615 # for long, if your local time moves backwards, by moving timezones or
616 # for an hour when daylight savings changes it, you will temporarily get
617 # a more aggressive retention policy for the overlapping period, and
618 # vice versa for the opposite timezone move. The alternative is using
619 # long-iso, which puts timezone info into the timestamp, which means
620 # that instead of shifting time, you shift the start of day/week/month
621 # which is used for retention to your new local time, which means for
622 # example, if you moved forward by 8 hours, the daily/weekly/monthly
623 # retention will be 8 hours more aggressive since midnight is at a new
624 # time, unless you fake the timzeone using the TZ env variable.
625 # However, in the short term, there will be no inconsistencies.
626 # I don't see any problem with shifting when the day starts for
627 # retention, so I'm using long-iso.
629 # note to create a long-iso timestamp: date +%Y%m%dT%H%M%S%z