X-Git-Url: https://iankelling.org/git/?p=distro-setup;a=blobdiff_plain;f=btrbk-run;h=6908018f61fbafda774339f4f299699e479a388b;hp=9ca4eb3333eaffbaa8b176a71d73d240c5530d6b;hb=8a6b446c7e336596af614c853e1c6177e55a7983;hpb=0dec66207e287c8d5d1d5a5319a7795a3bdec637 diff --git a/btrbk-run b/btrbk-run old mode 100755 new mode 100644 index 9ca4eb3..6908018 --- a/btrbk-run +++ b/btrbk-run @@ -1,38 +1,303 @@ -#!/bin/bash -l +#!/bin/bash +# Copyright (C) 2016 Ian Kelling + +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at + +# http://www.apache.org/licenses/LICENSE-2.0 + +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + + +# todo: if we cancel in the middle of a btrfs send, then run again immediately, the received subvolume doesn't get a Received UUID: field, and we won't mount it. Need to figure out a solution that will fix this. set -eE -o pipefail trap 'echo "$0:$LINENO:error: \"$BASH_COMMAND\" returned $?" >&2' ERR [[ $EUID == 0 ]] || exec sudo -E "$BASH_SOURCE" "$@" + +usage() { + cat <<'EOF' +btrbk-run [OPTIONS] +usually -t TARGET_HOST or -s SOURCE_HOST + +Note, at source location, intentionally not executable, run and read +install-my-scripts. + +EOF + echo "top of script file:" + sed -n '1,/^[# ]*end command line/{p;b};q' "$0" + exit $1 +} + +rsync-dirs() { + local host=$1 + local path=$2 + m rsync $dry_run_arg -ahi --relative --delete "$path" "root@$host:/" +} + + +m() { if $verbose; then printf "$pre %s\n" "$*"; fi; "$@"; } +die() { printf "$pre %s\n" "$*" >&2; exit 1; } + +# latest $MAIL_HOST +if [[ -e /b/bash_unpublished/source-semi-priv ]]; then + source /b/bash_unpublished/source-semi-priv +fi + +# note q is owned by root:1000 + +mountpoints=() + +rsync_mountpoint=/q + +# default options conf_only=false -dry_run=false -# mostly for testing -case $1 in - -c) conf_only=true ;; - -n) dry_run=true ;; - ?*) echo "$0: error: unsupported arg"; exit 1 ;; -esac - -# background on timezones. with short/long, timestamps use local time. -# for long, if your local time moves backwards, by moving timezones or -# for an hour when daylight savings changes it, you will temporarily get -# a more aggressive retention policy for the overlapping period, and -# vice versa for the opposite timezone move. The alternative is using -# long-iso, which puts timezone info into the timestamp, which means -# that instead of shifting time, you shift the start of day/week/month -# which is used for retention to your new local time, which means for -# example, if you moved forward by 8 hours, the daily/weekly/monthly -# retention will be 8 hours more aggressive since midnight is at a new -# time, unless you fake the timzeone using the TZ env variable. -# However, in the short term, there will be no inconsistencies. -# I don't see any problem with shifting when the day starts for -# retention, so I'm using long-iso. +dry_run=false # mostly for testing +rate_limit=no +verbose=true; verbose_arg=-v +progress_arg="--progress" +pull_reexec=false + +default_args_file=/etc/btrbk-run.conf +if [[ -s $default_args_file ]]; then + 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):" + cat $default_args_file + sleep 5 +fi + +pre="${0##*/}:" +cron=false +orig_args=("$@") +temp=$(getopt -l cron,pull-reexec,help cl:m:npqs:t:vh "$@") || usage 1 +eval set -- "$temp" +while true; do + case $1 in + --cron) + cron=true + pre= + shift + ;; + # only creates the config file, does not run btrbk + -c) conf_only=true; shift ;; + # bytes per second, suffix k m g + -l) rate_limit=$2; shift 2 ;; + # 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 ;; + --pull-reexec) pull_reexec=true; shift ;; + -q) verbose=false; verbose_arg=; progress_arg=; shift ;; + # source host to receive a backup from + -s) source=$2; shift 2 ;; + # target hosts to send to. empty is valid for just doing local + # snapshot. we have default hosts we will populate. + -t) IFS=, targets=($2); unset IFS; shift 2 ;; + -v) verbose=true; verbose_arg=-v; shift ;; + -h|--help) usage ;; + --) shift; break ;; + *) die "Internal error!" ;; + esac +done + +# usefull commands are resume and archive +cmd_arg=${1:-run} + +if [[ -v targets && $source ]]; then + 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 + +# TODO: i wonder if there should be an option to send to the default +# targets, plus any given on the command line. + +# set default targets +if [[ ! -v targets && ! $source ]]; then + if [[ $HOSTNAME != "$MAIL_HOST" ]] && $cron ; then + echo "We are not MAIL_HOST, nothing to do" + exit 0 + fi + case $HOSTNAME in + kw|x2) + if ping -q -c1 -w1 iank.vpn.office.fsf.org &>/dev/null; then + home=iank.vpn.office.fsf.org + else + home=$HOME_DOMAIN + fi + ;;& + kw) + targets=($home x2) + ;; + x2) + targets=($home kw) + ;; + tp) + targets=(frodo kd) + # might not be connected to the vpn + if timeout -s 9 10 ssh kw :; then + targets+=(kw) + fi + ;; + kd) + targets=(frodo tp) + # might not be connected to the vpn + if timeout -s 9 10 ssh kw :; then + targets+=(kw) + fi + ;; + *) + die "error: no default targets for this host, use -t" + ;; + esac +fi + +if [[ -v targets ]]; then + echo "targets: ${targets[*]}" +fi + +if [[ $source ]]; then + echo "source: $source" +fi + + +if [[ $mountpoints ]]; then + for mp in ${mountpoints[@]}; do # default mountpoints to sync + if [[ -e /nocow/btrfs-stale/$mp ]]; then + die "error: $mp is stale, mount-latest-subvol first" + fi + done +else + # set default mountpoints + case $HOSTNAME in + # no remote backups atm. note, if we do enable this, configuration below will need some changes. + # frodo) + # prospective_mps=(/i) + # ;; + *) + prospective_mps=(/a /q) + if [[ $HOSTNAME == "$MAIL_HOST" ]]; then + prospective_mps+=(/o) + fi + ;; + esac + for mp in ${prospective_mps[@]}; do # default mountpoints to sync + if [[ -e /nocow/btrfs-stale/$mp ]]; then + echo "$pre warning: $mp stale, not adding to default mountpoints" + continue + fi + if awk '{print $2}' /etc/fstab | grep -xF $mp &>/dev/null; then + mountpoints+=($mp) + fi + done +fi + +echo "mountpoints: ${mountpoints[*]}" + + -target_host=frodo -cat >/etc/btrbk.conf <<'EOF' -ssh_identity /root/.ssh/id_rsa -transaction_syslog daemon +# pull_reexec stops us from getting into an infinite loop if there is some +# kind of weird problem +pulla=false +for m in "${mountpoints[@]}"; do + if [[ $m == /a ]]; then + pulla=true + break + fi +done +if ! $pull_reexec && [[ $source ]] && $pulla ; then + tmpf=$(mktemp) + scp $source:/a/bin/distro-setup/btrbk-run $tmpf + if ! diff -q $tmpf $BASH_SOURCE; then + echo "$pre found newer version on host $source. reexecing" + install -T $tmpf /usr/local/bin/btrbk-run + m /usr/local/bin/btrbk-run --pull-reexec "${orig_args[@]}" + exit + fi +fi + + +##### end command line parsing ######## + + + +if ! which 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. +check-subvol-stale ${mountpoints[@]} || die "found stale mountpoints in ${mountpoints[*]}" + +# for an initial run, btrbk requires the dir to exist. +mkdir -p /mnt/root/btrbk +local_zone=$(date +%z) + +if [[ $source ]]; then + if ! zone=$(ssh root@$source date +%z); then + die failed to ssh to root@$source + 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 ${targets[@]}; do + if zone=$(ssh root@$h "mkdir -p /mnt/root/btrbk && date +%z"); then + if $cron && DISPLAY=:0 xprintidle; then + # This is a separate ssh because xprintidle can fail and thats ok. + # Ignore this host. i sometimes use a non-main machine for testing or web browsing, knowing that + # everything will be wiped by the next backup, but I dont want it to happen as Im using + # it from cronjob. + continue + fi + sshable+=($h) + if [[ $zone != $local_zone ]]; then + die "error: dont confuse yourself with multiple time zones. $h has different timezone than localhost" + fi + else + sshfail+=($h) + fi + done + if [[ ! $sshable ]] || { ! $cron && [[ $sshfail ]]; }; then + die "failed to ssh to hosts: ${sshfail[*]}" + else + if [[ $sshfail ]]; then + ret=1 + echo "$pre error: failed to ssh to ${sshfail[*]} but continuing with other hosts" + fi + targets=(${sshable[@]}) + fi +fi + + +cat >/etc/btrbk.conf <>/etc/btrbk.conf </dev/null || return - cat >>/etc/btrbk.conf <>/etc/btrbk.conf <>/etc/btrbk.conf <>/etc/btrbk.conf </dev/null; then + for tg in ${targets[@]}; do + case $tg in + li|lk) + for x in /p/c/machine_specific/*.hosts; do + if grep -qxF $tg $x; then + dir=${x%.hosts} + rsync-dirs $tg $dir + fi + done + ;; + esac + done +fi + +if [[ $source ]]; then + m mount-latest-subvol $verbose_arg else - btrbk -q run + m /a/exe/mount-latest-remote ${targets[@]} fi + +exit $ret + +# todo: move variable data we don't care about backing up +# to /nocow and symlink it. + + +# background on btrbk timezones. with short/long, timestamps use local time. +# for long, if your local time moves backwards, by moving timezones or +# for an hour when daylight savings changes it, you will temporarily get +# a more aggressive retention policy for the overlapping period, and +# vice versa for the opposite timezone move. The alternative is using +# long-iso, which puts timezone info into the timestamp, which means +# that instead of shifting time, you shift the start of day/week/month +# which is used for retention to your new local time, which means for +# example, if you moved forward by 8 hours, the daily/weekly/monthly +# retention will be 8 hours more aggressive since midnight is at a new +# time, unless you fake the timzeone using the TZ env variable. +# However, in the short term, there will be no inconsistencies. +# I don't see any problem with shifting when the day starts for +# retention, so I'm using long-iso. + +# note to create a long-iso timestamp: date +%Y%m%dT%H%M%S%z