#!/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.

set -eE -o pipefail
trap 'echo "$0:$LINENO:error: \"$BASH_COMMAND\" returned $?" >&2' ERR

[[ $EUID == 0 ]] || exec sudo -E "$BASH_SOURCE" "$@"
usage() {
    echo "top of script file:"
    sed -n '1,/^[# ]*end command line/{p;b};q' "$0"
    exit $1
}

script_dir=$(dirname $(readlink -f "$BASH_SOURCE"))

# note q is owned by root:1000
# note p/m is owned 1000:1000 and chmod 700
mountpoints=(/a)
private_mountpoints=(/q)
rsync_mountpoint=/q

conf_only=false
dry_run=false # mostly for testing
resume_arg=
do_i=true
do_o=true

temp=$(getopt -l help hcinoprt: "$@") || usage 1
eval set -- "$temp"
while true; do
    case $1 in
        -c) conf_only=true; shift ;;
        -i) do_i=false; shift ;;
        -n) dry_run=true; dry_run_arg=-n; shift ;;
        -o) do_o=false; shift ;;
        -p) progress_arg="--progress"; shift ;;
        # btrbk arg: Resume only. Skips snapshot creation.
        -r) resume_arg=-r; shift ;;
        -t) IFS=, targets=($2); unset IFS; shift 2 ;;
        -h|--help) usage ;;
        --) shift; break ;;
        *) echo "$0: Internal error!" ; exit 1 ;;
    esac
done

if $do_o; then
    private_mountpoints+=(/o)
fi
read primary <<<"$@" # not yet used

##### end command line parsing ########

rsync-dirs() {
    local host=$1
    local path=$2
    m rsync $dry_run_arg -ahi --relative --delete "$path" "root@$host:/"
}

vol-conf() {
    cat >>/etc/btrbk.conf <<EOF
volume $vol
EOF
}
sub-conf() {
    cat >>/etc/btrbk.conf <<EOF
subvolume $sub
EOF
}
tg-conf() {
    cat >>/etc/btrbk.conf <<EOF
target send-receive ssh://$tg$vol/btrbk
EOF
}
m() { printf "%s: %s\n" "${0##*/}" "$*";  "$@"; }


if ! which btrbk &>/dev/null; then
    echo "$0: error: no btrbk binary found"
fi

cat >/etc/btrbk.conf <<'EOF'
ssh_identity /root/.ssh/id_rsa
# Just a guess that local7 is a good facility to pick.
# It's a bit odd that the transaction log has to be logged to
# a file or syslog, while other output is sent to std out.
# The man does not mention a way for them to be together, but
# I dunno if setting a log level like warn might also output
# transaction info.
transaction_syslog local7

# so we only run one at a time
lockfile                   /var/lock/btrbk.lock

# default format of short does not accomidate hourly preservation setting
timestamp_format long-iso

# only make a snapshot if things have changed
snapshot_create onchange
# I could make this different from target_preserve,
# if one disk had less space.
# for now, keeping them equal.
snapshot_preserve 36h 14d 8w 24m
snapshot_preserve_min 4h
snapshot_dir btrbk

# so, total backups = ~89
target_preserve 36h 14d 8w 24m
target_preserve_min 4h

# if something fails and it's not obvious, try doing
# btrbk -l debug -v dryrun
EOF

for mp in ${private_mountpoints[@]}; do # private mountpoints
    if awk '{print $2}' /etc/fstab | grep -xF $mp &>/dev/null; then
        mountpoints+=($mp)
    fi
done

# if our mountpoints are from stale snapshots,
# it doesn't make sense to do a backup.
check-subvol-stale ${mountpoints[@]} || exit 1

if [[ ! $targets ]]; then
    case $HOSTNAME in
        tp|x2)
            if ! timeout -s 9 10 ssh frodo :; then
                targets=($HOME_DOMAIN)
            fi
            ;;
        treetowl)
            targets=(frodo)
            if timeout -s 9 10 ssh x2 :; then
                targets+=(x2)
            fi
            ;;
        *)
            targets=(frodo)
            ;;
    esac
    echo "targets: ${targets[*]}"
fi


# for i, we just do a 1 way sync from master to backup,
# and manually manage any changes to that.
i_possible=false
for tg in ${targets[@]}; do
    # for an initial run, btrbk requires the dir to exist
    ssh root@$tg mkdir -p /mnt/root/btrbk
    if [[ $tg == frodo && $HOSTNAME == treetowl ]]; then
        i_possible=true
    fi
done
if ! $i_possible; then
    do_i=false
fi


vol=/mnt/root
vol-conf
for m in ${mountpoints[@]}; do
    sub=${m##*/}
    sub-conf
    for tg in ${targets[@]}; do
        tg-conf
    done
done

if $do_i; then
    vol=/mnt/iroot
    vol-conf
    sub=i
    sub-conf
    tg=frodo
    vol=/mnt/root
    tg-conf
fi



# todo: umount first to ensure we don't have any errors
# todo: do some kill fuser stuff to make umount more reliable
# todo: run this on a systemd timer on $primary, once per hour,
# and if primary is, change that timer over to primary, and make
# sure we mount the latest



if $conf_only; then
    exit
fi

if $dry_run; then
    m btrbk -n $resume_arg run
else
    # -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 $progress_arg $resume_arg run
fi

# if we have it, sync to systems which don't
if mountpoint $rsync_mountpoint >/dev/null; then
    for tg in ${targets[@]}; do
        case $tg in
            tp|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 ! $dry_run; then
    m $script_dir/mount-latest-remote ${targets[@]}
fi


# 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