X-Git-Url: https://iankelling.org/git/?a=blobdiff_plain;f=btrbk-run;h=86d29e0c7cae496ebbd20f80e6d9be83dab833bd;hb=d6def754cd241538c61456536b52ee51cbd85b42;hp=b3a32071fabaea9803808743b305e0c6de024d99;hpb=ce4cacd36c5b5babeea85d0f93771017e6169180;p=distro-setup diff --git a/btrbk-run b/btrbk-run index b3a3207..86d29e0 100644 --- a/btrbk-run +++ b/btrbk-run @@ -26,7 +26,7 @@ source /usr/local/lib/err 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 @@ -63,7 +63,13 @@ conf_only=false 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 @@ -76,13 +82,17 @@ if [[ -s $default_args_file ]]; then sleep 5 fi -early=false # quit early, just btrbk, no extra remounting etc. +early=false cron=false orig_args=("$@") temp=$(getopt -l cron,pull-reexec,help ceil:m:npqrs:t:vh "$@") || 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= @@ -90,6 +100,7 @@ while true; do ;; # 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 @@ -97,9 +108,11 @@ while true; do # 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 ;; - -r) archive=false; shift ;; + # quiet -q) verbose=false; verbose_arg=; progress_arg=; shift ;; # source host to receive a backup from -s) @@ -113,6 +126,7 @@ while true; do # target hosts to send to. empty is valid for just doing local # snapshot. we have default hosts we will populate. -t) IFS=, targets=($2); unset IFS; shift 2 ;; + # verbose. -v) verbose=true; verbose_arg=-v; shift ;; -h|--help) usage ;; --) shift; break ;; @@ -120,11 +134,28 @@ while true; do esac done -# only tested commands are resume and archive cmd_arg=${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 @@ -159,21 +190,30 @@ fi # targets, plus any given on the command line. -amy=false + +kd_spread=false # set default targets if [[ ! -v targets && ! $source ]]; then - if [[ $HOSTNAME != "$MAIL_HOST" ]] && $cron ; then - echo "MAIL_HOST=$MAIL_HOST, nothing to do" - mexit 0 - else - amy=true + if $cron; then + if [[ $HOSTNAME != "$MAIL_HOST" ]]; then + if [[ $HOSTNAME == kd && $MAIL_HOST = x2 ]]; then + kd_spread=true + else + echo "MAIL_HOST=$MAIL_HOST, nothing to do" + mexit 0 + fi + fi fi + # x2 at home atm + kd_spread=false + at_work=false - targets=(frodo.b8.nz) + # todo, fix this up once frodo is back + # targets=(frodo.b8.nz) case $HOSTNAME in - x2|kw) + kw) at_work=true ;;& x2|x3|sy|bo) @@ -190,13 +230,20 @@ if [[ ! -v targets && ! $source ]]; then home=i.b8.nz fi else - home=b8.nz + if ping -q -c1 -w1 b8.nz &>/dev/null; then + home=b8.nz + else + home=i.b8.nz + fi fi ;;& + x2) + targets+=($home) + ;; kw) targets+=($home x2.office.fsf.org) ;; - x2|x3|sy|bo) + x3|sy|bo) targets+=($home) if $at_work; then targets+=(x2.office.fsf.org x2.b8.nz) @@ -205,12 +252,23 @@ if [[ ! -v targets && ! $source ]]; then fi ;; kd) - targets+=(x2wg.b8.nz x3.b8.nz) + if ! $kd_spread; then + if ping -q -c1 -w1 x2.office.fsf.org &>/dev/null; then + targets+=(x2.office.fsf.org) + else + targets+=(x2wg.b8.nz) + fi + 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 @@ -254,19 +312,19 @@ else 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. - prospective_mps+=(/a /ar /qr /q) ;; esac fi @@ -333,7 +391,11 @@ 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" @@ -353,36 +415,41 @@ else 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 @@ -410,7 +477,7 @@ ssh_identity /q/root/h transaction_syslog local7 # trying this out -stream_compress zstd +#stream_compress zstd # so we only run one at a time lockfile /var/lock/btrbk.lock @@ -423,11 +490,11 @@ snapshot_create onchange # 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 @@ -469,10 +536,10 @@ EOF q) # q has sensitive data i dont want to backup for so long cat >>/etc/btrbk.conf <>/etc/btrbk.conf <<'EOF' -# volume ssh://amy/mnt/root -# subvolume root_ubuntubionic -# target send-receive /mnt/root/btrbk -# EOF -# fi - # todo: umount first to ensure we don't have any errors # todo: do some kill fuser stuff to make umount more reliable @@ -512,19 +569,10 @@ fi 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