usage() {
cat <<'EOF'
-btrbk-run [OPTIONS]
+btrbk-run [OPTIONS] [run|resume|archive]
usually -t TARGET_HOST or -s SOURCE_HOST
Note, at source location, intentionally not executable, run and read
dry_run=false # mostly for testing
rate_limit=no
verbose=true; verbose_arg=-v
-progress_arg="--progress"
+if [[ $INVOCATION_ID ]]; then
+ # INVOCATION_ID means running as a systemd service. we cant show progress in this case,
+ # but if we pass the arg, it will insert mbuffer into the command.
+ progress_arg=
+else
+ progress_arg="--progress"
+fi
incremental_strict=false
pull_reexec=false
sleep 5
fi
-early=false # quit early, just btrbk, no extra remounting etc.
+early=false
cron=false
orig_args=("$@")
-temp=$(getopt -l cron,pull-reexec,help ceil:m:npqs:t:vh "$@") || usage 1
+temp=$(getopt -l cron,pull-reexec,help ceil:m:npqrs:t:vh "$@") || usage 1
eval set -- "$temp"
while true; do
case $1 in
+ # some behaviors specific to running under cron:
+ # - skip hosts where xprintidle haven't been idle recently
+ # - if we can't ssh to 1 or more hosts, still do the rest
+ # - if we aren't MAIL_HOST and no -m or -s, just exit
--cron)
cron=true
pre=
;;
# only creates the config file, does not run btrbk
-c) conf_only=true; shift ;;
+ # quit early, just btrbk, no extra remounting etc.
-e) early=true; shift ;;
-i) incremental_strict=true; shift ;;
# bytes per second, suffix k m g
# Comma separated mountpoints to backup. This has defaults set below.
-m) IFS=, mountpoints=($2); unset IFS; shift 2 ;;
-n) dry_run=true; dry_run_arg=-n; shift ;;
- -p) progress_arg="--progress"; shift ;;
+ # hide progress
+ -p) progress_arg=; shift ;;
+ # internal option for rerunning under newer SOURCE_HOST version.
--pull-reexec) pull_reexec=true; shift ;;
+ # quiet
-q) verbose=false; verbose_arg=; progress_arg=; shift ;;
# source host to receive a backup from
-s)
# target hosts to send to. empty is valid for just doing local
# snapshot. we have default hosts we will populate.
-t) IFS=, targets=($2); unset IFS; shift 2 ;;
+ # verbose.
-v) verbose=true; verbose_arg=-v; shift ;;
-h|--help) usage ;;
--) shift; break ;;
esac
done
-# only tested commands are resume and archive
cmd_arg=${1:-run}
+
+std_preserve="36h 14d 8w 24m"
+q_preserve="18h 14d"
+
case $cmd_arg in
- run|resume|archive) : ;;
+ run|resume) : ;;
+
+ # This works better than the normal archive command. We have to
+ # specify the mount points, but that is what we are used to doing and
+ # we prefer it. Another difference is that archive works recursively
+ # and we don't care about that. Sometimes we may still want to run
+ # btrbk archive, but it doesn't even use the config file, so just
+ # run it directly, eg:
+ # time s btrbk -v archive /mnt/r7/amy/boot/btrbk ssh://bo/mnt/boot2/btrbk
+ archive)
+ cmd_arg=resume
+ std_preserve="999h 999d 999w 999m"
+ q_preserve="$std_preserve"
+ preserve_arg=-p
+ ;;
*) die "untested command arg" ;;
esac
at_work=false
- targets=(frodo.b8.nz)
+ # todo, fix this up once frodo is back
+ # targets=(frodo.b8.nz)
case $HOSTNAME in
- kw)
+ x2|kw)
at_work=true
;;&
- x2|x3|sy)
+ x2|x3|sy|bo)
if ping -q -c1 -w1 hal.office.fsf.org \
&& ip n show 192.168.0.26 | grep . &>/dev/null; then
at_work=true
fi
;;&
- kw|x2|x3|sy)
+ kw|x2|x3|sy|bo)
if $at_work; then
if ping -q -c1 -w1 iank.vpn.office.fsf.org &>/dev/null; then
home=iank.vpn.office.fsf.org
home=b8.nz
fi
;;&
+ x2)
+ targets+=($home)
+ ;;
kw)
- targets+=($home x3)
+ targets+=($home x2.office.fsf.org)
;;
- x2|x3|sy)
+ x3|sy|bo)
targets+=($home)
if $at_work; then
- targets+=(kw.office.fsf.org x2.b8.nz)
+ targets+=(x2.office.fsf.org x2.b8.nz)
else
- targets+=(kw.b8.nz)
+ targets+=(x2wg.b8.nz)
fi
;;
kd)
- targets+=(x2.b8.nz kw.b8.nz)
- ## sy is unused
- # if ping -q -c1 -w1 sy.b8.nz &>/dev/null; then
- # targets+=(sy.b8.nz)
- # else
- # targets+=(syw.b8.nz)
- # fi
+ if ping -q -c1 -w1 x2.office.fsf.org &>/dev/null; then
+ targets+=(x2.office.fsf.org)
+ else
+ targets+=(x2wg.b8.nz)
+ fi
+ if ping -q -c1 -w1 sy.b8.nz &>/dev/null; then
+ targets+=(sy.b8.nz)
+ else
+ targets+=(syw.b8.nz)
+ fi
+ if ping -q -c1 -w1 x3.b8.nz &>/dev/null; then
+ targets+=(x3.b8.nz)
+ else
+ targets+=(x3w.b8.nz)
+ fi
;;
frodo)
# no targets
prospective_mps=(/i)
;;
*)
- prospective_mps=(/a /q)
+ prospective_mps=()
if [[ $source ]]; then
source_state="$(ssh $source cat /a/bin/bash_unpublished/source-state)"
eval "$source_state"
if [[ $source_host == "$MAIL_HOST" ]]; then
prospective_mps+=(/o)
fi
+ if [[ $source_host == "$HOST2" ]]; then
+ prospective_mps+=(/a /ar /qr /q)
+ fi
else
if [[ $HOSTNAME == "$MAIL_HOST" ]]; then
- # HOST2 is really the mail host if it exists
- if [[ $HOST2 && $HOST2 != "$HOSTNAME" ]]; then
- echo "skipping /o because HOST2 is not us"
- else
- prospective_mps+=(/o)
- fi
+ prospective_mps+=(/o)
+ fi
+ if [[ $HOSTNAME == "$HOST2" ]]; then
+ prospective_mps+=(/a /ar /qr /q)
fi
fi
+ # note: put q last just in case its specific retention options were to
+ # affect other config sections. I havent tested if that is the case.
;;
esac
fi
fi
-if ! which btrbk &>/dev/null; then
+if ! command -v btrbk &>/dev/null; then
die "error: no btrbk binary found"
fi
# if our mountpoints are from stale snapshots,
check-subvol-stale ${mountpoints[@]} || die "found stale mountpoints in ${mountpoints[*]}"
# for an initial run, btrbk requires the dir to exist.
-mkdir -p /mnt/root/btrbk
+mkdir -p /mnt/{root,o}/btrbk
local_zone=$(date +%z)
if [[ $source ]]; then
if ! zone=$(ssh root@$source date +%z); then
- die failed to ssh to root@$source
+ if $conf_only; then
+ echo "$0: warning: failed to ssh to root@$source"
+ else
+ die failed to ssh to root@$source
+ fi
fi
if [[ $zone != "$local_zone" ]]; then
die "error: dont confuse yourself with multiple time zones. $h has different timezone than localhost"
sshfail=()
min_idle_ms=$((1000 * 60 * 15))
for h in ${targets[@]}; do
- if remote_info=( $(timeout -s 9 6 ssh root@$h "mkdir -p /mnt/root/btrbk && date +%z && df --output=size,pcent / | tail -n1") ); then
+ 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
zone=${remote_info[0]}
root_size=${remote_info[1]}
percent_used=${remote_info[2]%%%}
if (( ${#remote_info[@]} != 3 )); then
die "error: didnt get 3 fields in test ssh to target $h. investigate"
fi
+ elif $conf_only; then
+ # Use some typical values in this case
+ root_size=$(( 1024 * 1024 * 2000 )) #2tb
+ percent_used=10
+ zone=$(date +%z)
+ else
+ sshfail+=($h)
+ continue
+ fi
+ # we may be booted into a bootstrap fs or something
+ min_root_kb=$(( 1024 * 1024 * 200 )) # 200 gb
+ if (( root_size < min_root_kb )); then
+ continue
+ fi
- # we may be booted into a bootstrap fs or something
- min_root_kb=$(( 1024 * 1024 * 200 )) # 200 gb
- if (( root_size < min_root_kb )); then
- continue
- fi
-
- if (( percent_used >= 98 )); then
- die "error: filesystem on target $h is $percent_used % full"
- fi
+ if (( percent_used >= 98 )); then
+ die "error: filesystem on target $h is $percent_used % full"
+ fi
- # This is a separate ssh because xprintidle can fail and thats ok.
- if $cron && idle_ms=$(timeout -s 9 6 ssh $h DISPLAY=:0 xprintidle); then
- if (( idle_ms < min_idle_ms )); then
+ # This is a separate ssh because xprintidle can fail and thats ok.
+ if $cron && idle_ms=$(timeout -s 9 6 ssh $h DISPLAY=:0 xprintidle); then
+ if (( idle_ms < min_idle_ms )); then
- # Ignore this host. i sometimes use a non-main machine for
- # testing or web browsing, knowing that everything will be wiped
- # by the next backup, but I dont want it to happen as Im using
- # it from cronjob.
- e "warning: $h: active X session in the last 15 minutes, skipping for now"
- continue
- fi
- fi
- sshable+=($h)
- if [[ $zone != "$local_zone" ]]; then
- die "error: dont confuse yourself with multiple time zones. $h has different timezone than localhost"
+ # Ignore this host. i sometimes use a non-main machine for
+ # testing or web browsing, knowing that everything will be wiped
+ # by the next backup, but I dont want it to happen as Im using
+ # it from cronjob.
+ e "warning: $h: active X session in the last 15 minutes, skipping for now"
+ continue
fi
- else
- sshfail+=($h)
+ fi
+ sshable+=($h)
+ if [[ $zone != "$local_zone" ]]; then
+ die "error: dont confuse yourself with multiple time zones. $h has different timezone than localhost"
fi
done
if [[ ! ${sshable[*]} ]] || { ! $cron && [[ ${sshfail[*]} ]]; }; then
# I could make this different from target_preserve,
# if one disk had less space.
# for now, keeping them equal.
-snapshot_preserve 18h 14d 8w 24m
+snapshot_preserve $std_preserve
snapshot_preserve_min 2h
snapshot_dir btrbk
-
# so, total backups = ~58
-target_preserve 18h 14d 8w 24m
+target_preserve $std_preserve
target_preserve_min 2h
# if something fails and it's not obvious, try doing
EOF
fi
-vol=/mnt/root
for m in ${mountpoints[@]}; do
+ case $m in
+ /o)
+ vol=/mnt/o
+ ;;
+ *)
+ vol=/mnt/root
+ ;;
+ esac
+
sub=${m#/}
if [[ $source ]]; then
cat >>/etc/btrbk.conf <<EOF
volume $vol
subvolume $sub
EOF
+ case $sub in
+ q)
+ # q has sensitive data i dont want to backup for so long
+ cat >>/etc/btrbk.conf <<EOF
+snapshot_preserve $std_preserve
+snapshot_preserve_min 2h
+snapshot_dir btrbk
+target_preserve $std_preserve
+target_preserve_min 2h
+EOF
+ ;;
+ esac
for tg in ${targets[@]}; do
# handle ipv6
if [[ $tg == *:* ]]; then
if $dry_run; then
m btrbk -v -n $cmd_arg
mexit 0
-elif [[ $cmd_arg == archive ]]; then
- if [[ $source ]]; then
- m btrbk $verbose_arg $progress_arg $cmd_arg ssh://$source$vol $vol
- else
- for tg in ${targets[@]}; do
- m btrbk $verbose_arg $progress_arg $cmd_arg $vol ssh://$tg$vol
- done
- fi
- mexit 0
fi
# -q and just using the syslog option seemed nice,
# but it doesn't show when a send has a parent and when it doesn't.
-m btrbk $verbose_arg $progress_arg $cmd_arg
+m btrbk $preserve_arg $verbose_arg $progress_arg $cmd_arg
# todo: tp not valid anymore.
# if we have it, sync to systems which don't