default_args_file=/etc/btrbk-run.conf
if [[ -s $default_args_file ]]; then
+ # shellcheck disable=SC2046 # we want word splitting
set -- $(< $default_args_file) "$@"
# i havent used this feature yet, so warn about it
echo "$0: warning: default btrbk-run options set in $default_args_file (sleeping 5 seconds):"
targets=()
early=false
cron=false
+fast=false
+kd_spread_maybe=false
orig_args=("$@")
-temp=$(getopt -l cron,pull-reexec,help 23ceil:m:npqrs:t:vh "$@") || usage 1
+temp=$(getopt -l cron,fast,pull-reexec,help 23ceikl:m:npqrs:t:vh "$@") || usage 1
eval set -- "$temp"
while true; do
case $1 in
-c) conf_only=true ;;
# quit early, just btrbk, no extra remounting etc.
-e) early=true ;;
+ # skip various checks. when we run twice in a row for
+ # switch mail-host, no need to repeat the same checks again.
+ --fast) fast=true ;;
-i) incremental_strict=true ;;
+ # note this implies resume
+ -k) kd_spread_maybe=true ;;
# bytes per second, suffix k m g
-l) rate_limit=$2; shift ;;
# Comma separated mountpoints to backup. This has defaults set below.
shift
done
-cmd_arg=${1:-run}
+cmd_arg="$1"
+
+if $kd_spread_maybe; then
+ if [[ $cmd_arg && $cmd_arg != resume ]]; then
+ die "dont pass -k without resume or empty run arg"
+ fi
+ cmd_arg=resume
+fi
+
+if [[ ! $cmd_arg ]]; then
+ cmd_arg=run
+fi
std_preserve="36h 14d 8w 24m"
die "error: -t and -s are mutually exclusive"
fi
-if $verbose; then
- printf "$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"
-fi
### end options parsing
# remove path from earlier version of btrbk
if [[ ! -v targets && ! $source ]]; then
if $cron; then
if [[ $HOSTNAME != "$MAIL_HOST" ]]; then
- if [[ $HOSTNAME == kd && $MAIL_HOST = x2 ]]; then
- kd_spread=true
+ if $kd_spread_maybe && [[ $HOSTNAME == kd && $MAIL_HOST == x3 ]]; then
+ if ping -q -c1 -w1 x3.office.fsf.org &>/dev/null; then
+ work_host=x3.office.fsf.org
+ elif ping -q -c1 -w1 x3wg.b8.nz &>/dev/null; then
+ work_host=x3wg.b8.nz
+ fi
+ if [[ $work_host ]]; then
+ source_state="$(ssh $work_host cat /a/bin/bash_unpublished/source-state)"
+ eval "$source_state"
+ if [[ $MAIL_HOST == x3 ]]; then
+ kd_spread=true
+ else
+ # x3 was the mail host, but it moved to some other machine
+ # without updating us yet.
+ echo "MAIL_HOST=$MAIL_HOST, nothing to do"
+ mexit 0
+ fi
+ else
+ echo "MAIL_HOST=$MAIL_HOST, nothing to do"
+ mexit 0
+ fi
else
echo "MAIL_HOST=$MAIL_HOST, nothing to do"
mexit 0
fi
fi
- # x2 at home atm
- kd_spread=false
-
at_work=false
at_home=false
;;&
*)
if $at_home; then
- # main work machine
- if ping -q -c1 -w1 x3.office.fsf.org &>/dev/null; then
- targets+=(x3.office.fsf.org)
- else
- targets+=(x3wg.b8.nz)
+ if ! $kd_spread && [[ $HOSTNAME != x3 ]]; then
+ # main work machine
+ if ping -q -c1 -w1 x3.office.fsf.org &>/dev/null; then
+ targets+=(x3.office.fsf.org)
+ elif ping -q -c1 -w1 $h.b8.nz &>/dev/null; then
+ # in case we took it home
+ targets+=(x3.b8.nz)
+ else
+ targets+=(x3wg.b8.nz)
+ fi
fi
- for h in frodo kd; do
+ # temporarily disabled while doing recovery
+ # for h in frodo kd; do
+ for h in kd; do
if [[ $HOSTNAME == "$h" ]]; then
continue
fi
targets+=($h.b8.nz)
done
- for h in x2 x3 sy; do
+ for h in x2 sy; do
if [[ $HOSTNAME == "$h" ]]; then
continue
fi
fi
done
elif $at_work; then
- if ping -q -c1 -w1 iank.vpn.office.fsf.org &>/dev/null; then
- targets+=(iank.vpn.office.fsf.org)
- else
- targets+=(i.b8.nz)
- fi
+ targets+=(i.b8.nz)
for h in x2 x3 kw; do
if [[ $HOSTNAME == "$h" ]]; then
continue
esac
fi
-if [[ -v targets ]]; then
- echo "targets: ${targets[*]}"
-fi
-
-if [[ $source ]]; then
- echo "source: $source"
-fi
-
if [[ ${mountpoints[0]} ]]; then
for mp in ${mountpoints[@]}; do
if [[ -e /nocow/btrfs-stale/$mp ]]; then
if [[ $HOSTNAME == "$HOST2" ]]; then
prospective_mps+=(/a /ar /qr /q)
fi
+ if $kd_spread; then
+ prospective_mps=(/a /ar /o /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.
done
fi
-echo "mountpoints: ${mountpoints[*]}"
+if (( ! ${#mountpoints[@]} )); then
+ die didnt get mountpoint arg and had no defaults
+fi
##### end command line parsing ########
-if [[ $source ]]; then
+#### begin pre-checks #####
+
+# todo: this has a timing problem, since btrbk.timer could activate the service after this check.
+if ! $fast && [[ $source ]]; then
if [[ $(ssh $source ps --no-headers -o comm 1) == systemd ]]; then
status=$(ssh $source systemctl is-active btrbk.service) || : # normally returns 3
case $status in
fi
fi
+if ! command -v btrbk &>/dev/null; then
+ die "error: no btrbk binary found"
+fi
+
+
+#### end pre-checks #####
+
+mkdir -p /var/log/btrbk
+# The journal doesnt go back to my oldest backups, and I've found myself
+# wanting older logs. Not going to bother expiring old logs, since it is
+# fine if they go back years.
+log_path=/var/log/btrbk/$(date +%F_%T%:::z).log
+echo copying output to $log_path
+exec &> >(ts "%F %T" | tee -a $log_path)
+
+
+if $verbose; then
+ printf "$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"
+fi
+
+if [[ -v targets ]]; then
+ echo "targets: ${targets[*]}"
+fi
+
+if [[ $source ]]; then
+ echo "source: $source"
+fi
+
+echo "mountpoints: ${mountpoints[*]}"
+
+
+
# pull_reexec stops us from getting into an infinite loop if there is some
# kind of weird problem
pulla=false
fi
-if ! command -v btrbk &>/dev/null; then
- die "error: no btrbk binary found"
-fi
-# if our mountpoints are from stale snapshots,
-# it doesn't make sense to do a backup.
-m check-subvol-stale ${mountpoints[@]} || die "found stale mountpoints in ${mountpoints[*]}"
+if ! $fast; then
+ # if our mountpoints are from stale snapshots,
+ # it doesn't make sense to do a backup.
+ m check-subvol-stale ${mountpoints[@]} || die "found stale mountpoints in ${mountpoints[*]}"
-# for an initial run, btrbk requires the dir to exist.
-mkdir -p /mnt/{root,o}/btrbk
+ # for an initial run, btrbk requires the dir to exist.
+ mkdir -p /mnt/{root,o}/btrbk
+fi
local_zone=$(date +%z)
if [[ $source ]]; then
- if ! zone=$(ssh root@$source date +%z); then
- if $conf_only; then
- echo "$0: warning: failed to ssh to root@$source"
- else
- die failed to ssh to root@$source
+ if $fast; then
+ zone=$local_zone
+ else
+ if ! zone=$(ssh root@$source date +%z); then
+ 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"
fi
fi
- if [[ $zone != "$local_zone" ]]; then
- die "error: dont confuse yourself with multiple time zones. $h has different timezone than localhost"
- fi
-
else
sshable=()
sshfail=()
- min_idle_ms=$((1000 * 60 * 15))
for h in ${targets[@]}; do
- if remote_info=( $(timeout -s 9 6 ssh root@$h "mkdir -p /mnt/root/btrbk /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
+ if $fast || $conf_only; then
# Use some typical values in this case
root_size=$(( 1024 * 1024 * 2000 )) #2tb
percent_used=10
zone=$(date +%z)
+ elif remote_str=$(timeout -s 9 6 ssh root@$h "mkdir -p /mnt/root/btrbk /mnt/o/btrbk && date +%z && df --output=size,pcent / | tail -n1"); then
+ mapfile -t tmp_array <<<"$remote_str"
+ zone="${tmp_array[0]}"
+ IFS=" " read -r root_size percent_used <<<"${tmp_array[1]}"
+ percent_used=${percent_used%%%}
+
+ if (( ${#tmp_array[@]} != 2 )); then
+ die "error: didnt get 2 lines in test ssh to target $h. investigate"
+ fi
+ case $percent_used in
+ [0-9]|[1-9][0-9]) : ;;
+ *)
+ die "error: didnt get percent disk use in test ssh to target $h. investigate"
+ ;;
+ esac
else
sshfail+=($h)
continue
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
-
+ # on sy, xprintidle is resetting every 12 seconds even when not
+ # idle, i dunno why, instead we are checking if the screen is locked,
+ # which is good enough.
+ #
+ # This is a separate ssh because the command can fail and thatis ok.
+ if $cron; then
+ locked=false
+ if lock_info=$(timeout -s 9 6 ssh $h DISPLAY=:0 xscreensaver-command -time); then
+ if [[ $lock_info != *non-blanked* ]]; then
+ locked=true
+ fi
+ else
+ locked=true
+ fi
+ if ! $locked; 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"
+ e "warning: $h: seems to be actively in use, skipping for now"
continue
fi
fi
# if one disk had less space.
# for now, keeping them equal.
snapshot_preserve $std_preserve
-snapshot_preserve_min 2h
+snapshot_preserve_min 6h
snapshot_dir btrbk
# so, total backups = ~58
target_preserve $std_preserve
-target_preserve_min 2h
+target_preserve_min 6h
+
+# i tried this when investigating: clone no source subvolume found error
+#incremental_prefs sro:1 srn:1 sao san:1 aro:1 arn:1
# if something fails and it's not obvious, try doing
# btrbk -l debug -v dryrun
mp_count=${#mountpoints[@]}
for (( i=0; i < mp_count - 1 ; i++ )); do
if [[ ${mountpoints[i]} == /q ]]; then
- unset mountpoints[i]
+ unset "mountpoints[i]"
mountpoints+=(/q)
fi
done
done
fi
+subvols=()
+for mp in "${mountpoints[@]}"; do
+ subvols+=("${mp##*/}")
+done
if [[ $source ]]; then
- m mount-latest-subvol
+ m mount-latest-subvol "${subvols[@]}"
else
- m /a/exe/mount-latest-remote ${targets[@]}
+ for tg in ${targets[@]}; do
+ m /a/exe/mount-latest-remote "$tg" "${subvols[@]}" || ret=$?
+ done
fi
if [[ $ret == 0 ]]; then
for tg in ${targets[@]}; do
- :
- #ssh root@$tg /a/exe/mail-backup-clean
+ h=$(ssh root@$tg hostname)
+ rsync -a /var/log/btrbk $tg:/var/log/btrbk/$tg
+ ssh root@$tg /usr/local/bin/mail-backup-clean
done
+ if [[ $source ]]; then
+ rsync -a $source:/var/log/btrbk /var/log/btrbk/$source
+ fi
fi
mexit $ret